Update go.mod and go.sum (#34)

- Add server-side detection of SDK processor's internal system prompt
  in handleSessionInit, since CLAUDE_MNEMONIC_INTERNAL env var is not
  propagated by Claude Code to hook subprocesses
- Add cross-session duplicate detection (FindRecentPromptByTextGlobal)
  to catch same prompt text arriving from different session IDs
- Add hooks, mcpServers, and commands references to plugin.json per
  Claude Code plugin spec
- Remove MCP server injection from register-plugin.sh (now in plugin.json)
- Use ${CLAUDE_PLUGIN_ROOT} for statusline path instead of hardcoded path
- Add python3 fallback for plugin registration when jq is unavailable
- Replace hardcoded 1.0.0 version in findWorkerBinary with glob lookup
- Add cache copy verification in register-plugin.sh
- Add update-version Makefile target to keep metadata in sync

Makefile

@@ -14,7 +14,7 @@ GOARCH ?= $(shell go env GOARCH)
 export CGO_ENABLED=1
 BUILD_TAGS := -tags "fts5"
 
-.PHONY: all build clean test install lint hooks worker mcp stop-worker start-worker restart-worker dashboard website dev-website setup-libs
+.PHONY: all build clean test install lint hooks worker mcp stop-worker start-worker restart-worker dashboard website dev-website setup-libs update-version
 
 all: build
 
@@ -22,8 +22,20 @@ all: build
 setup-libs:
 	@./scripts/download-onnx-libs.sh all
 
+# Update root plugin metadata with current version
+update-version:
+	@mkdir -p .claude-plugin
+	@sed 's/{{ .Version }}/$(VERSION)/g; s/{{.Version}}/$(VERSION)/g' $(PLUGIN_DIR)/.claude-plugin/plugin.json.tpl > .claude-plugin/plugin.json
+	@echo "Updated .claude-plugin/plugin.json to version $(VERSION)"
+	@# marketplace.json contains release-specific data (URLs, SHA256 hashes) that requires manual update per release.
+	@# Only the top-level version field is updated here.
+	@if [ -f marketplace.json ]; then \
+		sed 's/"version": "[^"]*"/"version": "$(VERSION)"/' marketplace.json > marketplace.json.tmp && mv marketplace.json.tmp marketplace.json; \
+		echo "Updated marketplace.json version fields to $(VERSION)"; \
+	fi
+
 # Build all binaries
-build: setup-libs dashboard worker hooks mcp
+build: setup-libs update-version dashboard worker hooks mcp
 
 # Build Vue dashboard
 dashboard:
@@ -108,6 +120,9 @@ build-windows:
 # Stop any running worker
 stop-worker:
 	@echo "Stopping worker..."
+	@-pkill -TERM -f 'claude-mnemonic.*worker' 2>/dev/null || true
+	@-pkill -TERM -f '\.claude/plugins/.*/worker' 2>/dev/null || true
+	@sleep 1
 	@-pkill -9 -f 'claude-mnemonic.*worker' 2>/dev/null || true
 	@-pkill -9 -f '\.claude/plugins/.*/worker' 2>/dev/null || true
 	@-lsof -ti :37777 | xargs kill -9 2>/dev/null || true
@@ -135,6 +150,10 @@ restart-worker: stop-worker start-worker
 # Install to Claude plugins directory
 install: build stop-worker
 	@echo "Installing to Claude plugins directory..."
+	@# Verify build output binaries exist
+	@test -f $(BUILD_DIR)/worker || { echo "ERROR: $(BUILD_DIR)/worker not found. Build may have failed."; exit 1; }
+	@test -f $(BUILD_DIR)/mcp-server || { echo "ERROR: $(BUILD_DIR)/mcp-server not found. Build may have failed."; exit 1; }
+	@test -d $(BUILD_DIR)/hooks || { echo "ERROR: $(BUILD_DIR)/hooks not found. Build may have failed."; exit 1; }
 	@# Install to marketplaces directory (for direct installs)
 	@mkdir -p $(HOME)/.claude/plugins/marketplaces/claude-mnemonic/hooks
 	@mkdir -p $(HOME)/.claude/plugins/marketplaces/claude-mnemonic/.claude-plugin
@@ -228,6 +247,7 @@ help:
 	@echo "  make fmt            - Format code"
 	@echo "  make clean          - Clean build artifacts"
 	@echo "  make dev            - Run worker in development mode"
+	@echo "  make update-version - Update version in plugin metadata files"
 	@echo "  make deps           - Download dependencies"
 	@echo "  make website        - Build website for production"
 	@echo "  make dev-website    - Run website dev server"

cmd/hooks/post-tool-use/main.go

@@ -2,8 +2,10 @@
 package main
 
 import (
+	"context"
 	"fmt"
 	"os"
+	"time"
 
 	"github.com/lukaszraczylo/claude-mnemonic/pkg/hooks"
 )
@@ -51,6 +53,10 @@ var skipTools = map[string]bool{
 }
 
 func main() {
+	if !hooks.IsWorkerAvailable() {
+		hooks.WriteResponse("PostToolUse", true)
+		return
+	}
 	hooks.RunHook("PostToolUse", handlePostToolUse)
 }
 
@@ -63,15 +69,31 @@ func handlePostToolUse(ctx *hooks.HookContext, input *Input) (string, error) {
 
 	fmt.Fprintf(os.Stderr, "[post-tool-use] %s\n", input.ToolName)
 
-	// Send observation to worker
-	_, err := hooks.POST(ctx.Port, "/api/sessions/observations", map[string]interface{}{
-		"claudeSessionId": ctx.SessionID,
-		"project":         ctx.Project,
-		"tool_name":       input.ToolName,
-		"tool_input":      input.ToolInput,
-		"tool_response":   input.ToolResponse,
-		"cwd":             ctx.CWD,
-	})
+	// Fire-and-forget: send the observation without waiting for the response.
+	// The worker just queues it -- we don't need the response data.
+	// Use a short-lived context to ensure the request body is at least sent
+	// before this process exits.
+	done := make(chan struct{})
+	go func() {
+		defer close(done)
+		sendCtx, cancel := context.WithTimeout(context.Background(), 500*time.Millisecond)
+		defer cancel()
+		_ = hooks.POSTWithContext(sendCtx, ctx.Port, "/api/sessions/observations", map[string]interface{}{
+			"claudeSessionId": ctx.SessionID,
+			"project":         ctx.Project,
+			"tool_name":       input.ToolName,
+			"tool_input":      input.ToolInput,
+			"tool_response":   input.ToolResponse,
+			"cwd":             ctx.CWD,
+		})
+	}()
 
-	return "", err
+	// Wait briefly for the TCP connection to be established and request sent,
+	// but don't block the hook for the full response.
+	select {
+	case <-done:
+	case <-time.After(100 * time.Millisecond):
+	}
+
+	return "", nil
 }

cmd/hooks/session-start/main.go

@@ -6,6 +6,7 @@ import (
 	"net/url"
 	"os"
 	"strings"
+	"time"
 
 	"github.com/lukaszraczylo/claude-mnemonic/pkg/hooks"
 )
@@ -27,10 +28,17 @@ type Observation struct {
 }
 
 func main() {
+	if !hooks.IsWorkerAvailable() {
+		hooks.WriteResponse("SessionStart", true)
+		return
+	}
 	hooks.RunHook("SessionStart", handleSessionStart)
 }
 
 func handleSessionStart(ctx *hooks.HookContext, input *Input) (string, error) {
+	deadline, cancel := hooks.HookDeadline(30 * time.Second)
+	defer cancel()
+
 	// Fetch observations for context injection
 	endpoint := fmt.Sprintf("/api/context/inject?project=%s&cwd=%s",
 		url.QueryEscape(ctx.Project),
@@ -59,12 +67,21 @@ func handleSessionStart(ctx *hooks.HookContext, input *Input) (string, error) {
 	fmt.Fprintf(os.Stderr, "[claude-mnemonic] Injecting %d observations from project memory (%d detailed, %d condensed)\n",
 		len(obsData), min(fullCount, len(obsData)), max(0, len(obsData)-fullCount))
 
+	// Token budget for context injection
+	maxTokens := 16000 // default; could be made configurable via worker config endpoint
+	currentTokens := 0
+
 	// Build context string
-	contextBuilder := "<claude-mnemonic-context>\n"
-	contextBuilder += fmt.Sprintf("# Project Memory (%d observations)\n", len(obsData))
-	contextBuilder += "Use this knowledge to answer questions without re-exploring the codebase.\n\n"
+	header := fmt.Sprintf("<claude-mnemonic-context>\n# Project Memory (%d observations)\nUse this knowledge to answer questions without re-exploring the codebase.\n\n", len(obsData))
+	currentTokens += estimateTokens(header)
+	contextBuilder := header
 
 	for i, o := range obsData {
+		if deadline.Err() != nil {
+			contextBuilder += "\n... (returning early due to time limit)\n"
+			break
+		}
+
 		obs, ok := o.(map[string]interface{})
 		if !ok {
 			continue
@@ -73,40 +90,94 @@ func handleSessionStart(ctx *hooks.HookContext, input *Input) (string, error) {
 		title := getString(obs, "title")
 		obsType := getString(obs, "type")
 
+		var obsText string
+
 		// First `fullCount` observations get full detail, rest are condensed
 		if i < fullCount {
 			// Full detail: include narrative and facts
 			narrative := getString(obs, "narrative")
 
-			contextBuilder += fmt.Sprintf("## %d. [%s] %s\n", i+1, strings.ToUpper(obsType), title)
+			obsText = fmt.Sprintf("## %d. [%s] %s\n", i+1, strings.ToUpper(obsType), title)
 			if narrative != "" {
-				contextBuilder += narrative + "\n"
+				obsText += narrative + "\n"
 			}
 
 			if facts, ok := obs["facts"].([]interface{}); ok && len(facts) > 0 {
-				contextBuilder += "Key facts:\n"
+				obsText += "Key facts:\n"
 				for _, f := range facts {
 					if fact, ok := f.(string); ok && fact != "" {
-						contextBuilder += fmt.Sprintf("- %s\n", fact)
+						obsText += fmt.Sprintf("- %s\n", fact)
 					}
 				}
 			}
-			contextBuilder += "\n"
+			obsText += "\n"
 		} else {
 			// Condensed: just title and subtitle (one line)
 			subtitle := getString(obs, "subtitle")
 			if subtitle != "" {
-				contextBuilder += fmt.Sprintf("- [%s] %s: %s\n", strings.ToUpper(obsType), title, subtitle)
+				obsText = fmt.Sprintf("- [%s] %s: %s\n", strings.ToUpper(obsType), title, subtitle)
 			} else {
-				contextBuilder += fmt.Sprintf("- [%s] %s\n", strings.ToUpper(obsType), title)
+				obsText = fmt.Sprintf("- [%s] %s\n", strings.ToUpper(obsType), title)
 			}
 		}
+
+		obsTokens := estimateTokens(obsText)
+		if currentTokens+obsTokens > maxTokens {
+			contextBuilder += fmt.Sprintf("\n... (%d more observations omitted due to token budget)\n", len(obsData)-i)
+			break
+		}
+
+		contextBuilder += obsText
+		currentTokens += obsTokens
 	}
 
 	contextBuilder += "</claude-mnemonic-context>\n"
 	return contextBuilder, nil
 }
 
+// estimateTokens provides a more accurate token count estimate.
+// Uses word count * 1.3 as base, with adjustments for code and non-ASCII.
+func estimateTokens(s string) int {
+	if len(s) == 0 {
+		return 0
+	}
+
+	// Count words (split on whitespace)
+	words := len(strings.Fields(s))
+	if words == 0 {
+		// No whitespace = probably a single token or code blob
+		return (len(s) + 3) / 4
+	}
+
+	// Base estimate: ~1.3 tokens per word for English text
+	estimate := int(float64(words) * 1.3)
+
+	// Detect code-heavy content (high non-alpha ratio)
+	nonAlpha := 0
+	nonASCII := 0
+	for _, r := range s {
+		if r > 127 {
+			nonASCII++
+		} else if !('a' <= r && r <= 'z') && !('A' <= r && r <= 'Z') && !('0' <= r && r <= '9') && r != ' ' {
+			nonAlpha++
+		}
+	}
+
+	totalChars := len(s)
+
+	// Code adjustment: more special chars = more tokens per word
+	if totalChars > 0 && float64(nonAlpha)/float64(totalChars) > 0.15 {
+		estimate = int(float64(estimate) * 1.3)
+	}
+
+	// Non-ASCII adjustment: CJK and other scripts use more tokens
+	if totalChars > 0 && float64(nonASCII)/float64(totalChars) > 0.1 {
+		estimate += nonASCII // Roughly 1 extra token per non-ASCII char
+	}
+
+	return estimate
+}
+
 func getString(m map[string]interface{}, key string) string {
 	if v, ok := m[key].(string); ok {
 		return v

cmd/hooks/statusline/main.go

@@ -117,7 +117,7 @@ func getWorkerStats(port int, project string) *WorkerStats {
 	if err != nil {
 		return nil
 	}
-	defer resp.Body.Close()
+	defer func() { _ = resp.Body.Close() }()
 
 	if resp.StatusCode != http.StatusOK {
 		return nil

cmd/hooks/stop/main.go

@@ -5,12 +5,16 @@ import (
 	"bufio"
 	"encoding/json"
 	"fmt"
+	"io"
 	"os"
 	"strings"
+	"time"
 
 	"github.com/lukaszraczylo/claude-mnemonic/pkg/hooks"
 )
 
+var debug = os.Getenv("CLAUDE_MNEMONIC_DEBUG") != ""
+
 // Input is the hook input from Claude Code.
 type Input struct {
 	hooks.BaseInput
@@ -62,7 +66,19 @@ func parseTranscript(path string) (lastUser, lastAssistant string) {
 	if err != nil {
 		return "", ""
 	}
-	defer file.Close()
+	defer func() { _ = file.Close() }()
+
+	// For large transcripts, seek to the last 256KB for efficiency.
+	// We only need the last user/assistant messages, not the entire history.
+	const tailSize = 256 * 1024
+	info, err := file.Stat()
+	if err == nil && info.Size() > tailSize {
+		if _, seekErr := file.Seek(-tailSize, io.SeekEnd); seekErr == nil {
+			// Discard partial first line after seek
+			discardScanner := bufio.NewScanner(file)
+			discardScanner.Scan()
+		}
+	}
 
 	scanner := bufio.NewScanner(file)
 	// Increase buffer size for large messages
@@ -97,12 +113,20 @@ func parseTranscript(path string) (lastUser, lastAssistant string) {
 }
 
 func main() {
+	if !hooks.IsWorkerAvailable() {
+		hooks.WriteResponse("Stop", true)
+		return
+	}
 	hooks.RunHook("Stop", handleStop)
 }
 
 func handleStop(ctx *hooks.HookContext, input *Input) (string, error) {
-	// Debug: dump raw input
-	fmt.Fprintf(os.Stderr, "[stop] Raw input: %s\n", string(ctx.RawInput))
+	deadline, cancel := hooks.HookDeadline(30 * time.Second)
+	defer cancel()
+
+	if debug {
+		fmt.Fprintf(os.Stderr, "[stop] Raw input: %s\n", string(ctx.RawInput))
+	}
 
 	// Find session
 	result, err := hooks.GET(ctx.Port, fmt.Sprintf("/api/sessions?claudeSessionId=%s", ctx.SessionID))
@@ -122,18 +146,33 @@ func handleStop(ctx *hooks.HookContext, input *Input) (string, error) {
 		lastUser, lastAssistant = parseTranscript(input.TranscriptPath)
 	}
 
-	// Debug: log what we extracted
-	fmt.Fprintf(os.Stderr, "[stop] Transcript path: %s\n", input.TranscriptPath)
-	fmt.Fprintf(os.Stderr, "[stop] Last user message length: %d\n", len(lastUser))
-	fmt.Fprintf(os.Stderr, "[stop] Last assistant message length: %d\n", len(lastAssistant))
-	if len(lastAssistant) > 0 {
-		preview := lastAssistant
-		if len(preview) > 300 {
-			preview = preview[:300] + "..."
+	// Truncate messages to avoid sending excessive data to the worker
+	if len(lastAssistant) > 10000 {
+		lastAssistant = lastAssistant[:10000]
+	}
+	if len(lastUser) > 5000 {
+		lastUser = lastUser[:5000]
+	}
+
+	if debug {
+		fmt.Fprintf(os.Stderr, "[stop] Transcript path: %s\n", input.TranscriptPath)
+		fmt.Fprintf(os.Stderr, "[stop] Last user message length: %d\n", len(lastUser))
+		fmt.Fprintf(os.Stderr, "[stop] Last assistant message length: %d\n", len(lastAssistant))
+		if len(lastAssistant) > 0 {
+			preview := lastAssistant
+			if len(preview) > 300 {
+				preview = preview[:300] + "..."
+			}
+			fmt.Fprintf(os.Stderr, "[stop] Last assistant preview: %s\n", preview)
 		}
-		fmt.Fprintf(os.Stderr, "[stop] Last assistant preview: %s\n", preview)
+		fmt.Fprintf(os.Stderr, "[stop] Requesting summary for session %d (transcript: %v)\n", int64(sessionID), input.TranscriptPath != "")
+	}
+
+	// Check deadline before expensive summary request
+	if deadline.Err() != nil {
+		fmt.Fprintf(os.Stderr, "[stop] Returning early due to time limit\n")
+		return "", nil
 	}
-	fmt.Fprintf(os.Stderr, "[stop] Requesting summary for session %d (transcript: %v)\n", int64(sessionID), input.TranscriptPath != "")
 
 	// Request summary with message context from transcript
 	_, err = hooks.POST(ctx.Port, fmt.Sprintf("/sessions/%d/summarize", int64(sessionID)), map[string]interface{}{

cmd/hooks/subagent-stop/main.go

@@ -16,6 +16,10 @@ type Input struct {
 }
 
 func main() {
+	if !hooks.IsWorkerAvailable() {
+		hooks.WriteResponse("SubagentStop", true)
+		return
+	}
 	hooks.RunHook("SubagentStop", handleSubagentStop)
 }
 

cmd/hooks/user-prompt/main.go

@@ -5,8 +5,12 @@ import (
 	"fmt"
 	"net/url"
 	"os"
+	"strings"
+	"sync"
+	"time"
 
 	"github.com/lukaszraczylo/claude-mnemonic/pkg/hooks"
+	"github.com/lukaszraczylo/claude-mnemonic/pkg/sanitize"
 )
 
 // Input is the hook input from Claude Code.
@@ -15,31 +19,117 @@ type Input struct {
 	Prompt string `json:"prompt"`
 }
 
+// estimateTokens provides a more accurate token count estimate.
+// Uses word count * 1.3 as base, with adjustments for code and non-ASCII.
+func estimateTokens(s string) int {
+	if len(s) == 0 {
+		return 0
+	}
+
+	// Count words (split on whitespace)
+	words := len(strings.Fields(s))
+	if words == 0 {
+		// No whitespace = probably a single token or code blob
+		return (len(s) + 3) / 4
+	}
+
+	// Base estimate: ~1.3 tokens per word for English text
+	estimate := int(float64(words) * 1.3)
+
+	// Detect code-heavy content (high non-alpha ratio)
+	nonAlpha := 0
+	nonASCII := 0
+	for _, r := range s {
+		if r > 127 {
+			nonASCII++
+		} else if !('a' <= r && r <= 'z') && !('A' <= r && r <= 'Z') && !('0' <= r && r <= '9') && r != ' ' {
+			nonAlpha++
+		}
+	}
+
+	totalChars := len(s)
+
+	// Code adjustment: more special chars = more tokens per word
+	if totalChars > 0 && float64(nonAlpha)/float64(totalChars) > 0.15 {
+		estimate = int(float64(estimate) * 1.3)
+	}
+
+	// Non-ASCII adjustment: CJK and other scripts use more tokens
+	if totalChars > 0 && float64(nonASCII)/float64(totalChars) > 0.1 {
+		estimate += nonASCII // Roughly 1 extra token per non-ASCII char
+	}
+
+	return estimate
+}
+
 func main() {
+	if !hooks.IsWorkerAvailable() {
+		hooks.WriteResponse("UserPromptSubmit", true)
+		return
+	}
 	hooks.RunHook("UserPromptSubmit", handleUserPrompt)
 }
 
 func handleUserPrompt(ctx *hooks.HookContext, input *Input) (string, error) {
-	// Search for relevant observations based on the prompt
+	deadline, cancel := hooks.HookDeadline(10 * time.Second)
+	defer cancel()
+
 	searchURL := fmt.Sprintf("/api/context/search?project=%s&query=%s&cwd=%s",
 		url.QueryEscape(ctx.Project),
 		url.QueryEscape(input.Prompt),
 		url.QueryEscape(ctx.CWD))
 
-	var contextToInject string
-	var observationCount int
+	// Run search and session init concurrently.
+	// Session init doesn't strictly depend on search results -- the observation
+	// count passed is approximate (0) and acceptable.
+	var (
+		wg               sync.WaitGroup
+		searchResult     map[string]interface{}
+		initResult       map[string]interface{}
+		initErr          error
+		contextToInject  string
+		observationCount int
+	)
 
-	searchResult, _ := hooks.GET(ctx.Port, searchURL)
+	// Start search in background
+	wg.Add(1)
+	go func() {
+		defer wg.Done()
+		searchResult, _ = hooks.GET(ctx.Port, searchURL)
+	}()
+
+	// Start session init in parallel (with observationCount=0; approximate is fine)
+	wg.Add(1)
+	go func() {
+		defer wg.Done()
+		initResult, initErr = hooks.POST(ctx.Port, "/api/sessions/init", map[string]interface{}{
+			"claudeSessionId":     ctx.SessionID,
+			"project":             ctx.Project,
+			"prompt":              input.Prompt,
+			"matchedObservations": 0,
+		})
+	}()
+
+	// Wait for both to complete
+	wg.Wait()
+
+	// Check deadline after network calls
+	if deadline.Err() != nil {
+		return "", nil
+	}
+
+	// Process search results
 	if observations, ok := searchResult["observations"].([]interface{}); ok && len(observations) > 0 {
-		// Results are already filtered by relevance threshold and capped by max_results
-		// from the server-side config (ContextRelevanceThreshold, ContextMaxPromptResults)
 		observationCount = len(observations)
 
-		// Build context from search results
+		// Token budget for prompt context injection
+		maxTokens := 8000
+		currentTokens := 0
+
+		header := "<relevant-memory>\n# Relevant Knowledge From Previous Sessions\nIMPORTANT: Use this information to answer the question directly. Do NOT explore the codebase if the answer is here.\n\n"
+		currentTokens += estimateTokens(header)
 		var contextBuilder string
-		contextBuilder = "<relevant-memory>\n"
-		contextBuilder += "# Relevant Knowledge From Previous Sessions\n"
-		contextBuilder += "IMPORTANT: Use this information to answer the question directly. Do NOT explore the codebase if the answer is here.\n\n"
+		contextBuilder = header
 
 		for i, obs := range observations {
 			if obsMap, ok := obs.(map[string]interface{}); ok {
@@ -52,24 +142,30 @@ func handleUserPrompt(ctx *hooks.HookContext, input *Input) (string, error) {
 					obsType = t
 				}
 
-				// Start observation block
-				contextBuilder += fmt.Sprintf("## %d. [%s] %s\n", i+1, obsType, title)
+				var obsText string
+				obsText = fmt.Sprintf("## %d. [%s] %s\n", i+1, obsType, title)
 
-				// Add facts first (most concise answers)
 				if facts, ok := obsMap["facts"].([]interface{}); ok && len(facts) > 0 {
-					contextBuilder += "Key facts:\n"
+					obsText += "Key facts:\n"
 					for _, fact := range facts {
 						if factStr, ok := fact.(string); ok {
-							contextBuilder += fmt.Sprintf("- %s\n", factStr)
+							obsText += fmt.Sprintf("- %s\n", sanitize.StripSystemXML(factStr))
 						}
 					}
-					contextBuilder += "\n"
+					obsText += "\n"
 				}
 
-				// Add narrative if present
 				if narrative, ok := obsMap["narrative"].(string); ok && narrative != "" {
-					contextBuilder += narrative + "\n\n"
+					obsText += sanitize.StripSystemXML(narrative) + "\n\n"
 				}
+
+				obsTokens := estimateTokens(obsText)
+				if currentTokens+obsTokens > maxTokens {
+					break
+				}
+
+				contextBuilder += obsText
+				currentTokens += obsTokens
 			}
 		}
 
@@ -77,40 +173,24 @@ func handleUserPrompt(ctx *hooks.HookContext, input *Input) (string, error) {
 		contextToInject = contextBuilder
 	}
 
-	// Initialize session with matched observations count
-	result, err := hooks.POST(ctx.Port, "/api/sessions/init", map[string]interface{}{
-		"claudeSessionId":     ctx.SessionID,
-		"project":             ctx.Project,
-		"prompt":              input.Prompt,
-		"matchedObservations": observationCount,
-	})
-	if err != nil {
-		return "", err
+	// Check session init result
+	if initErr != nil {
+		return "", initErr
 	}
 
 	// Check if skipped due to privacy
-	if skipped, ok := result["skipped"].(bool); ok && skipped {
+	if skipped, ok := initResult["skipped"].(bool); ok && skipped {
 		fmt.Fprintf(os.Stderr, "[user-prompt] Session skipped (private)\n")
 		return "", nil
 	}
 
-	// Safely extract session ID and prompt number with type checking
-	sessionDbIdRaw, ok := result["sessionDbId"].(float64)
-	if !ok {
-		return "", fmt.Errorf("invalid or missing sessionDbId in response")
-	}
-	sessionID := int64(sessionDbIdRaw)
-
-	promptNumberRaw, ok := result["promptNumber"].(float64)
-	if !ok {
-		return "", fmt.Errorf("invalid or missing promptNumber in response")
-	}
-	promptNumber := int(promptNumberRaw)
+	sessionID := int64(initResult["sessionDbId"].(float64))
+	promptNumber := int(initResult["promptNumber"].(float64))
 
 	fmt.Fprintf(os.Stderr, "[user-prompt] Session %d, prompt #%d\n", sessionID, promptNumber)
 
-	// Start SDK agent
-	_, err = hooks.POST(ctx.Port, fmt.Sprintf("/sessions/%d/init", sessionID), map[string]interface{}{
+	// Start SDK agent (depends on session init result, so kept sequential)
+	_, err := hooks.POST(ctx.Port, fmt.Sprintf("/sessions/%d/init", sessionID), map[string]interface{}{
 		"userPrompt":   input.Prompt,
 		"promptNumber": promptNumber,
 	})
@@ -120,7 +200,6 @@ func handleUserPrompt(ctx *hooks.HookContext, input *Input) (string, error) {
 
 	// Return context if we found relevant observations
 	if observationCount > 0 {
-		// Show match count to user via stderr
 		fmt.Fprintf(os.Stderr, "[claude-mnemonic] Found %d relevant memories for this prompt\n", observationCount)
 		return contextToInject, nil
 	}

cmd/mcp/main.go

@@ -4,20 +4,16 @@ package main
 import (
 	"context"
 	"flag"
+	"fmt"
+	"net/http"
 	"os"
 	"os/signal"
 	"syscall"
 	"time"
 
 	"github.com/lukaszraczylo/claude-mnemonic/internal/config"
-	"github.com/lukaszraczylo/claude-mnemonic/internal/db/gorm"
-	"github.com/lukaszraczylo/claude-mnemonic/internal/embedding"
 	"github.com/lukaszraczylo/claude-mnemonic/internal/mcp"
-	"github.com/lukaszraczylo/claude-mnemonic/internal/scoring"
-	"github.com/lukaszraczylo/claude-mnemonic/internal/search"
-	"github.com/lukaszraczylo/claude-mnemonic/internal/vector/sqlitevec"
 	"github.com/lukaszraczylo/claude-mnemonic/internal/watcher"
-	"github.com/lukaszraczylo/claude-mnemonic/pkg/models"
 	"github.com/rs/zerolog"
 	"github.com/rs/zerolog/log"
 )
@@ -28,7 +24,6 @@ var Version = "dev"
 func main() {
 	// Parse flags
 	project := flag.String("project", "", "Project name (required)")
-	dataDir := flag.String("data-dir", "", "Data directory (default: ~/.claude-mnemonic)")
 	debug := flag.Bool("debug", false, "Enable debug logging")
 	flag.Parse()
 
@@ -43,23 +38,12 @@ func main() {
 		log.Fatal().Msg("--project is required")
 	}
 
-	// Ensure data directory and settings exist
-	if err := config.EnsureAll(); err != nil {
-		log.Fatal().Err(err).Msg("Failed to ensure data directories")
-	}
+	// Get worker port from config
+	port := config.GetWorkerPort()
+	workerURL := fmt.Sprintf("http://localhost:%d", port)
 
-	// Load config
-	cfg, err := config.Load()
-	if err != nil {
-		log.Warn().Err(err).Msg("Failed to load config, using defaults")
-		cfg = config.Default()
-	}
-
-	// Override data directory if specified
-	dbPath := cfg.DBPath
-	if *dataDir != "" {
-		dbPath = *dataDir + "/claude-mnemonic.db"
-	}
+	// Create HTTP client for worker
+	client := &http.Client{Timeout: 30 * time.Second}
 
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
@@ -73,69 +57,12 @@ func main() {
 		cancel()
 	}()
 
-	// Initialize database store (migrations run automatically)
-	storeCfg := gorm.Config{
-		Path:     dbPath,
-		MaxConns: cfg.MaxConns,
-		// WALMode is enabled automatically by GORM
-	}
-	store, err := gorm.NewStore(storeCfg)
-	if err != nil {
-		log.Fatal().Err(err).Msg("Failed to initialize database store")
-	}
-	defer store.Close()
+	// Start file watchers for config changes
+	startWatchers()
 
-	// Initialize stores
-	observationStore := gorm.NewObservationStore(store, nil, nil, nil)
-	summaryStore := gorm.NewSummaryStore(store)
-	promptStore := gorm.NewPromptStore(store, nil)
-	patternStore := gorm.NewPatternStore(store)
-	relationStore := gorm.NewRelationStore(store)
-	sessionStore := gorm.NewSessionStore(store)
-
-	// Initialize embedding service and vector client
-	var vectorClient *sqlitevec.Client
-	embedSvc, err := embedding.NewService()
-	if err != nil {
-		log.Warn().Err(err).Msg("Embedding service unavailable, vector search disabled")
-	} else {
-		defer embedSvc.Close()
-		vectorClient, err = sqlitevec.NewClient(sqlitevec.Config{DB: store.GetRawDB()}, embedSvc)
-		if err != nil {
-			log.Warn().Err(err).Msg("Vector client unavailable, vector search disabled")
-		} else {
-			log.Info().Msg("Vector search enabled via sqlite-vec")
-		}
-	}
-
-	// Initialize scoring components
-	scoreConfig := models.DefaultScoringConfig()
-	scoreCalculator := scoring.NewCalculator(scoreConfig)
-	recalculator := scoring.NewRecalculator(observationStore, scoreCalculator, log.Logger)
-	go recalculator.Start(ctx)
-	defer recalculator.Stop()
-
-	// Initialize search manager
-	searchMgr := search.NewManager(observationStore, summaryStore, promptStore, vectorClient)
-
-	// Start file watchers
-	startWatchers(ctx, dbPath)
-
-	// Create and run MCP server with all dependencies
-	// Note: maintenanceService is nil because it runs in the worker process
-	server := mcp.NewServer(
-		searchMgr,
-		Version,
-		observationStore,
-		patternStore,
-		relationStore,
-		sessionStore,
-		vectorClient,
-		scoreCalculator,
-		recalculator,
-		nil, // maintenanceService - handled by worker
-	)
-	log.Info().Str("project", *project).Str("version", Version).Msg("Starting MCP server")
+	// Create and run MCP server
+	server := mcp.NewServer(client, workerURL, *project, Version)
+	log.Info().Str("project", *project).Str("version", Version).Str("worker", workerURL).Msg("Starting MCP server")
 
 	if err := server.Run(ctx); err != nil {
 		log.Fatal().Err(err).Msg("MCP server error")
@@ -143,7 +70,7 @@ func main() {
 }
 
 // startWatchers initializes file watchers for config.
-func startWatchers(ctx context.Context, dbPath string) {
+func startWatchers() {
 	// Watch config file for changes (triggers process exit for restart)
 	configPath := config.SettingsPath()
 	configWatcher, err := watcher.New(configPath, func() {

go.mod

@@ -7,16 +7,16 @@ replace github.com/sugarme/tokenizer => github.com/clems4ever/tokenizer v0.0.0-2
 require (
 	github.com/asg017/sqlite-vec-go-bindings v0.1.6
 	github.com/fsnotify/fsnotify v1.9.0
-	github.com/go-chi/chi/v5 v5.2.3
+	github.com/go-chi/chi/v5 v5.2.5
 	github.com/go-gormigrate/gormigrate/v2 v2.1.5
-	github.com/goccy/go-json v0.10.5
-	github.com/mattn/go-sqlite3 v1.14.33
+	github.com/goccy/go-json v0.10.6
+	github.com/mattn/go-sqlite3 v1.14.34
 	github.com/rs/zerolog v1.34.0
 	github.com/smacker/go-tree-sitter v0.0.0-20240827094217-dd81d9e9be82
 	github.com/stretchr/testify v1.11.1
 	github.com/sugarme/tokenizer v0.3.0
-	github.com/yalue/onnxruntime_go v1.25.0
-	golang.org/x/sync v0.19.0
+	github.com/yalue/onnxruntime_go v1.27.0
+	golang.org/x/sync v0.20.0
 	gorm.io/driver/sqlite v1.6.0
 	gorm.io/gorm v1.31.1
 )
@@ -34,7 +34,7 @@ require (
 	github.com/rivo/uniseg v0.4.7 // indirect
 	github.com/schollz/progressbar/v2 v2.15.0 // indirect
 	github.com/sugarme/regexpset v0.0.0-20200920021344-4d4ec8eaf93c // indirect
-	golang.org/x/sys v0.40.0 // indirect
-	golang.org/x/text v0.33.0 // indirect
+	golang.org/x/sys v0.42.0 // indirect
+	golang.org/x/text v0.35.0 // indirect
 	gopkg.in/yaml.v3 v3.0.1 // indirect
 )

go.sum

@@ -10,12 +10,12 @@ github.com/emirpasic/gods v1.18.1 h1:FXtiHYKDGKCW2KzwZKx0iC0PQmdlorYgdFG9jPXJ1Bc
 github.com/emirpasic/gods v1.18.1/go.mod h1:8tpGGwCnJ5H4r6BWwaV6OrWmMoPhUl5jm/FMNAnJvWQ=
 github.com/fsnotify/fsnotify v1.9.0 h1:2Ml+OJNzbYCTzsxtv8vKSFD9PbJjmhYF14k/jKC7S9k=
 github.com/fsnotify/fsnotify v1.9.0/go.mod h1:8jBTzvmWwFyi3Pb8djgCCO5IBqzKJ/Jwo8TRcHyHii0=
-github.com/go-chi/chi/v5 v5.2.3 h1:WQIt9uxdsAbgIYgid+BpYc+liqQZGMHRaUwp0JUcvdE=
-github.com/go-chi/chi/v5 v5.2.3/go.mod h1:L2yAIGWB3H+phAw1NxKwWM+7eUH/lU8pOMm5hHcoops=
+github.com/go-chi/chi/v5 v5.2.5 h1:Eg4myHZBjyvJmAFjFvWgrqDTXFyOzjj7YIm3L3mu6Ug=
+github.com/go-chi/chi/v5 v5.2.5/go.mod h1:X7Gx4mteadT3eDOMTsXzmI4/rwUpOwBHLpAfupzFJP0=
 github.com/go-gormigrate/gormigrate/v2 v2.1.5 h1:1OyorA5LtdQw12cyJDEHuTrEV3GiXiIhS4/QTTa/SM8=
 github.com/go-gormigrate/gormigrate/v2 v2.1.5/go.mod h1:mj9ekk/7CPF3VjopaFvWKN2v7fN3D9d3eEOAXRhi/+M=
-github.com/goccy/go-json v0.10.5 h1:Fq85nIqj+gXn/S5ahsiTlK3TmC85qgirsdTP/+DeaC4=
-github.com/goccy/go-json v0.10.5/go.mod h1:oq7eo15ShAhp70Anwd5lgX2pLfOS3QCiwU/PULtXL6M=
+github.com/goccy/go-json v0.10.6 h1:p8HrPJzOakx/mn/bQtjgNjdTcN+/S6FcG2CTtQOrHVU=
+github.com/goccy/go-json v0.10.6/go.mod h1:oq7eo15ShAhp70Anwd5lgX2pLfOS3QCiwU/PULtXL6M=
 github.com/godbus/dbus/v5 v5.0.4/go.mod h1:xhWf0FNVPg57R7Z0UbKHbJfkEywrmjJnf7w5xrFpKfA=
 github.com/jinzhu/inflection v1.0.0 h1:K317FqzuhWc8YvSVlFMCCUb36O/S9MCKRDI7QkRKD/E=
 github.com/jinzhu/inflection v1.0.0/go.mod h1:h+uFLlag+Qp1Va5pdKtLDYj+kHp5pxUVkryuEj+Srlc=
@@ -28,8 +28,8 @@ github.com/mattn/go-isatty v0.0.16/go.mod h1:kYGgaQfpe5nmfYZH+SKPsOc2e4SrIfOl2e/
 github.com/mattn/go-isatty v0.0.19/go.mod h1:W+V8PltTTMOvKvAeJH7IuucS94S2C6jfK/D7dTCTo3Y=
 github.com/mattn/go-isatty v0.0.20 h1:xfD0iDuEKnDkl03q4limB+vH+GxLEtL/jb4xVJSWWEY=
 github.com/mattn/go-isatty v0.0.20/go.mod h1:W+V8PltTTMOvKvAeJH7IuucS94S2C6jfK/D7dTCTo3Y=
-github.com/mattn/go-sqlite3 v1.14.33 h1:A5blZ5ulQo2AtayQ9/limgHEkFreKj1Dv226a1K73s0=
-github.com/mattn/go-sqlite3 v1.14.33/go.mod h1:Uh1q+B4BYcTPb+yiD3kU8Ct7aC0hY9fxUwlHK0RXw+Y=
+github.com/mattn/go-sqlite3 v1.14.34 h1:3NtcvcUnFBPsuRcno8pUtupspG/GM+9nZ88zgJcp6Zk=
+github.com/mattn/go-sqlite3 v1.14.34/go.mod h1:Uh1q+B4BYcTPb+yiD3kU8Ct7aC0hY9fxUwlHK0RXw+Y=
 github.com/mitchellh/colorstring v0.0.0-20190213212951-d06e56a500db h1:62I3jR2EmQ4l5rM/4FEfDWcRD+abF5XlKShorW5LRoQ=
 github.com/mitchellh/colorstring v0.0.0-20190213212951-d06e56a500db/go.mod h1:l0dey0ia/Uv7NcFFVbCLtqEBQbrT4OCwCSKTEv6enCw=
 github.com/patrickmn/go-cache v2.1.0+incompatible h1:HRMgzkcYKYpi3C8ajMPV8OFXaaRUnok+kx1WdO15EQc=
@@ -52,17 +52,17 @@ github.com/stretchr/testify v1.11.1 h1:7s2iGBzp5EwR7/aIZr8ao5+dra3wiQyKjjFuvgVKu
 github.com/stretchr/testify v1.11.1/go.mod h1:wZwfW3scLgRK+23gO65QZefKpKQRnfz6sD981Nm4B6U=
 github.com/sugarme/regexpset v0.0.0-20200920021344-4d4ec8eaf93c h1:pwb4kNSHb4K89ymCaN+5lPH/MwnfSVg4rzGDh4d+iy4=
 github.com/sugarme/regexpset v0.0.0-20200920021344-4d4ec8eaf93c/go.mod h1:2gwkXLWbDGUQWeL3RtpCmcY4mzCtU13kb9UsAg9xMaw=
-github.com/yalue/onnxruntime_go v1.25.0 h1:nlhVau1BpLZ/BYr+WpPZCJRD/WES0qo6dK7aKyyAs3g=
-github.com/yalue/onnxruntime_go v1.25.0/go.mod h1:b4X26A8pekNb1ACJ58wAXgNKeUCGEAQ9dmACut9Sm/4=
-golang.org/x/sync v0.19.0 h1:vV+1eWNmZ5geRlYjzm2adRgW2/mcpevXNg50YZtPCE4=
-golang.org/x/sync v0.19.0/go.mod h1:9KTHXmSnoGruLpwFjVSX0lNNA75CykiMECbovNTZqGI=
+github.com/yalue/onnxruntime_go v1.27.0 h1:c1YSgDNtpf0WGtxj3YeRIb8VC5LmM1J+Ve3uHdteC1U=
+github.com/yalue/onnxruntime_go v1.27.0/go.mod h1:b4X26A8pekNb1ACJ58wAXgNKeUCGEAQ9dmACut9Sm/4=
+golang.org/x/sync v0.20.0 h1:e0PTpb7pjO8GAtTs2dQ6jYa5BWYlMuX047Dco/pItO4=
+golang.org/x/sync v0.20.0/go.mod h1:9xrNwdLfx4jkKbNva9FpL6vEN7evnE43NNNJQ2LF3+0=
 golang.org/x/sys v0.0.0-20220811171246-fbc7d0a398ab/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.6.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.12.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
-golang.org/x/sys v0.40.0 h1:DBZZqJ2Rkml6QMQsZywtnjnnGvHza6BTfYFWY9kjEWQ=
-golang.org/x/sys v0.40.0/go.mod h1:OgkHotnGiDImocRcuBABYBEXf8A9a87e/uXjp9XT3ks=
-golang.org/x/text v0.33.0 h1:B3njUFyqtHDUI5jMn1YIr5B0IE2U0qck04r6d4KPAxE=
-golang.org/x/text v0.33.0/go.mod h1:LuMebE6+rBincTi9+xWTY8TztLzKHc/9C1uBCG27+q8=
+golang.org/x/sys v0.42.0 h1:omrd2nAlyT5ESRdCLYdm3+fMfNFE/+Rf4bDIQImRJeo=
+golang.org/x/sys v0.42.0/go.mod h1:4GL1E5IUh+htKOUEOaiffhrAeqysfVGipDYzABqnCmw=
+golang.org/x/text v0.35.0 h1:JOVx6vVDFokkpaq1AEptVzLTpDe9KGpj5tR4/X+ybL8=
+golang.org/x/text v0.35.0/go.mod h1:khi/HExzZJ2pGnjenulevKNX1W67CUy0AsXcNubPGCA=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405 h1:yhCVgyC4o1eVCa2tZl7eS0r+SDo693bJlVdllGtEeKM=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
 gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=

internal/chunking/python/chunker_test.go

@@ -0,0 +1,298 @@
+package python
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/lukaszraczylo/claude-mnemonic/internal/chunking"
+)
+
+// =============================================================================
+// TEST HELPERS
+// =============================================================================
+
+func createTempPythonFile(t *testing.T, content string) string {
+	t.Helper()
+
+	tmpDir := t.TempDir()
+	filePath := filepath.Join(tmpDir, "test.py")
+
+	err := os.WriteFile(filePath, []byte(content), 0600)
+	require.NoError(t, err)
+
+	return filePath
+}
+
+// =============================================================================
+// TESTS FOR Chunker
+// =============================================================================
+
+func TestNewChunker(t *testing.T) {
+	t.Parallel()
+
+	opts := chunking.DefaultChunkOptions()
+	c := NewChunker(opts)
+
+	assert.NotNil(t, c)
+	assert.NotNil(t, c.parser)
+}
+
+func TestChunker_Language(t *testing.T) {
+	t.Parallel()
+
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	assert.Equal(t, chunking.LanguagePython, c.Language())
+}
+
+func TestChunker_SupportedExtensions(t *testing.T) {
+	t.Parallel()
+
+	c := NewChunker(chunking.DefaultChunkOptions())
+	exts := c.SupportedExtensions()
+
+	assert.Contains(t, exts, ".py")
+}
+
+func TestChunker_Chunk_SimpleFunction(t *testing.T) {
+	t.Parallel()
+
+	code := `def greet(name):
+    """Greets a person by name."""
+    return f"Hello, {name}!"
+`
+
+	filePath := createTempPythonFile(t, code)
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	require.NotEmpty(t, chunks)
+
+	// Should find the greet function
+	var foundGreet bool
+	for _, chunk := range chunks {
+		if chunk.Name == "greet" {
+			foundGreet = true
+			assert.Equal(t, chunking.ChunkTypeFunction, chunk.Type)
+			assert.Equal(t, chunking.LanguagePython, chunk.Language)
+			assert.Contains(t, chunk.Content, "def greet")
+		}
+	}
+	assert.True(t, foundGreet, "Should find 'greet' function")
+}
+
+func TestChunker_Chunk_ClassWithMethods(t *testing.T) {
+	t.Parallel()
+
+	code := `class Calculator:
+    """A simple calculator class."""
+
+    def add(self, a, b):
+        """Adds two numbers."""
+        return a + b
+
+    def multiply(self, a, b):
+        """Multiplies two numbers."""
+        return a * b
+`
+
+	filePath := createTempPythonFile(t, code)
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	require.NotEmpty(t, chunks)
+
+	// Should find the Calculator class and its methods
+	var foundClass, foundAdd, foundMultiply bool
+	for _, chunk := range chunks {
+		switch chunk.Name {
+		case "Calculator":
+			foundClass = true
+			assert.Equal(t, chunking.ChunkTypeClass, chunk.Type)
+		case "add":
+			foundAdd = true
+			assert.Equal(t, chunking.ChunkTypeMethod, chunk.Type)
+			assert.Equal(t, "Calculator", chunk.ParentName)
+		case "multiply":
+			foundMultiply = true
+			assert.Equal(t, chunking.ChunkTypeMethod, chunk.Type)
+			assert.Equal(t, "Calculator", chunk.ParentName)
+		}
+	}
+
+	assert.True(t, foundClass, "Should find 'Calculator' class")
+	assert.True(t, foundAdd, "Should find 'add' method")
+	assert.True(t, foundMultiply, "Should find 'multiply' method")
+}
+
+func TestChunker_Chunk_MultipleFunctions(t *testing.T) {
+	t.Parallel()
+
+	code := `def first_function():
+    pass
+
+def second_function(x, y):
+    return x + y
+
+def third_function():
+    """Has a docstring."""
+    return 42
+`
+
+	filePath := createTempPythonFile(t, code)
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+
+	// Should find all three functions
+	functionNames := make(map[string]bool)
+	for _, chunk := range chunks {
+		if chunk.Type == chunking.ChunkTypeFunction {
+			functionNames[chunk.Name] = true
+		}
+	}
+
+	assert.True(t, functionNames["first_function"])
+	assert.True(t, functionNames["second_function"])
+	assert.True(t, functionNames["third_function"])
+}
+
+func TestChunker_Chunk_FileNotFound(t *testing.T) {
+	t.Parallel()
+
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	_, err := c.Chunk(context.Background(), "/nonexistent/path/file.py")
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "read file")
+}
+
+func TestChunker_Chunk_EmptyFile(t *testing.T) {
+	t.Parallel()
+
+	filePath := createTempPythonFile(t, "")
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	assert.Empty(t, chunks)
+}
+
+func TestChunker_Chunk_OnlyComments(t *testing.T) {
+	t.Parallel()
+
+	code := `# This is a comment
+# Another comment
+"""
+This is a module docstring
+"""
+`
+
+	filePath := createTempPythonFile(t, code)
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	// Comments and docstrings without code should not produce chunks
+	assert.Empty(t, chunks)
+}
+
+func TestChunker_Chunk_NestedClass(t *testing.T) {
+	t.Parallel()
+
+	code := `class Outer:
+    class Inner:
+        def inner_method(self):
+            pass
+
+    def outer_method(self):
+        pass
+`
+
+	filePath := createTempPythonFile(t, code)
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	require.NotEmpty(t, chunks)
+
+	// Should find the Outer class at minimum
+	var foundOuter bool
+	for _, chunk := range chunks {
+		if chunk.Name == "Outer" {
+			foundOuter = true
+		}
+	}
+	assert.True(t, foundOuter, "Should find 'Outer' class")
+}
+
+func TestChunker_Chunk_Decorators(t *testing.T) {
+	t.Parallel()
+
+	code := `@staticmethod
+def static_func():
+    pass
+
+@classmethod
+def class_func(cls):
+    pass
+
+@property
+def my_property(self):
+    return self._value
+`
+
+	filePath := createTempPythonFile(t, code)
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	require.NotEmpty(t, chunks)
+
+	// Should find decorated functions
+	functionNames := make(map[string]bool)
+	for _, chunk := range chunks {
+		functionNames[chunk.Name] = true
+	}
+
+	assert.True(t, functionNames["static_func"])
+	assert.True(t, functionNames["class_func"])
+	assert.True(t, functionNames["my_property"])
+}
+
+func TestChunker_Chunk_AsyncFunction(t *testing.T) {
+	t.Parallel()
+
+	code := `async def fetch_data(url):
+    """Fetches data from URL asynchronously."""
+    pass
+
+async def process_items(items):
+    for item in items:
+        await process(item)
+`
+
+	filePath := createTempPythonFile(t, code)
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	require.NotEmpty(t, chunks)
+
+	// Should find async functions
+	functionNames := make(map[string]bool)
+	for _, chunk := range chunks {
+		functionNames[chunk.Name] = true
+	}
+
+	assert.True(t, functionNames["fetch_data"])
+	assert.True(t, functionNames["process_items"])
+}

internal/chunking/types_test.go

@@ -0,0 +1,213 @@
+package chunking
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+// =============================================================================
+// TESTS FOR Chunk METHODS
+// =============================================================================
+
+func TestChunk_Identifier(t *testing.T) {
+	tests := []struct {
+		name     string
+		expected string
+		chunk    Chunk
+	}{
+		// ===== GOOD CASES =====
+		{
+			name: "top-level function",
+			chunk: Chunk{
+				Name:       "MyFunction",
+				ParentName: "",
+			},
+			expected: "MyFunction",
+		},
+		{
+			name: "method with parent",
+			chunk: Chunk{
+				Name:       "Process",
+				ParentName: "Handler",
+			},
+			expected: "Handler.Process",
+		},
+		{
+			name: "nested method",
+			chunk: Chunk{
+				Name:       "Validate",
+				ParentName: "UserService",
+			},
+			expected: "UserService.Validate",
+		},
+
+		// ===== EDGE CASES =====
+		{
+			name: "empty name",
+			chunk: Chunk{
+				Name:       "",
+				ParentName: "",
+			},
+			expected: "",
+		},
+		{
+			name: "parent but no name",
+			chunk: Chunk{
+				Name:       "",
+				ParentName: "Parent",
+			},
+			expected: "Parent.",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			result := tt.chunk.Identifier()
+			assert.Equal(t, tt.expected, result)
+		})
+	}
+}
+
+func TestChunk_LineRange(t *testing.T) {
+	tests := []struct {
+		name     string
+		expected string
+		chunk    Chunk
+	}{
+		// ===== GOOD CASES =====
+		{
+			name: "single line",
+			chunk: Chunk{
+				StartLine: 10,
+				EndLine:   10,
+			},
+			expected: "L10-L10",
+		},
+		{
+			name: "multi-line",
+			chunk: Chunk{
+				StartLine: 25,
+				EndLine:   50,
+			},
+			expected: "L25-L50",
+		},
+
+		// ===== EDGE CASES =====
+		{
+			name: "line 1",
+			chunk: Chunk{
+				StartLine: 1,
+				EndLine:   5,
+			},
+			expected: "L1-L5",
+		},
+		{
+			name: "large line numbers",
+			chunk: Chunk{
+				StartLine: 1000,
+				EndLine:   2500,
+			},
+			expected: "L1000-L2500",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			result := tt.chunk.LineRange()
+			assert.Equal(t, tt.expected, result)
+		})
+	}
+}
+
+func TestChunk_SearchableContent(t *testing.T) {
+	tests := []struct {
+		name     string
+		contains []string
+		chunk    Chunk
+	}{
+		// ===== GOOD CASES =====
+		{
+			name: "full chunk with all fields",
+			chunk: Chunk{
+				Signature:  "func ProcessData(input []byte) error",
+				DocComment: "// ProcessData handles incoming data",
+				Content:    "func ProcessData(input []byte) error {\n\treturn nil\n}",
+			},
+			contains: []string{
+				"func ProcessData(input []byte) error",
+				"ProcessData handles incoming data",
+				"return nil",
+			},
+		},
+		{
+			name: "only signature",
+			chunk: Chunk{
+				Signature: "func Hello()",
+			},
+			contains: []string{"func Hello()"},
+		},
+		{
+			name: "only content",
+			chunk: Chunk{
+				Content: "some code here",
+			},
+			contains: []string{"some code here"},
+		},
+
+		// ===== EDGE CASES =====
+		{
+			name:     "empty chunk",
+			chunk:    Chunk{},
+			contains: []string{},
+		},
+		{
+			name: "only doc comment",
+			chunk: Chunk{
+				DocComment: "// Important documentation",
+			},
+			contains: []string{"Important documentation"},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			result := tt.chunk.SearchableContent()
+			for _, expected := range tt.contains {
+				assert.Contains(t, result, expected)
+			}
+		})
+	}
+}
+
+func TestDefaultChunkOptions(t *testing.T) {
+	opts := DefaultChunkOptions()
+
+	assert.Greater(t, opts.MaxChunkSize, 0, "MaxChunkSize should be positive")
+	assert.True(t, opts.IncludeDocComments, "IncludeDocComments should be true by default")
+	assert.True(t, opts.IncludePrivate, "IncludePrivate should be true by default")
+	assert.Equal(t, 0, opts.MinLines, "MinLines should be 0 by default")
+}
+
+// =============================================================================
+// TESTS FOR ChunkType AND Language CONSTANTS
+// =============================================================================
+
+func TestChunkType_Values(t *testing.T) {
+	// Ensure all chunk types have expected values
+	assert.Equal(t, ChunkType("function"), ChunkTypeFunction)
+	assert.Equal(t, ChunkType("method"), ChunkTypeMethod)
+	assert.Equal(t, ChunkType("class"), ChunkTypeClass)
+	assert.Equal(t, ChunkType("interface"), ChunkTypeInterface)
+	assert.Equal(t, ChunkType("type"), ChunkTypeType)
+	assert.Equal(t, ChunkType("const"), ChunkTypeConst)
+	assert.Equal(t, ChunkType("var"), ChunkTypeVar)
+}
+
+func TestLanguage_Values(t *testing.T) {
+	// Ensure all language types have expected values
+	assert.Equal(t, Language("go"), LanguageGo)
+	assert.Equal(t, Language("python"), LanguagePython)
+	assert.Equal(t, Language("typescript"), LanguageTypeScript)
+	assert.Equal(t, Language("javascript"), LanguageJavaScript)
+}

internal/chunking/typescript/chunker_test.go

@@ -0,0 +1,398 @@
+package typescript
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/lukaszraczylo/claude-mnemonic/internal/chunking"
+)
+
+// =============================================================================
+// TEST HELPERS
+// =============================================================================
+
+func createTempTSFile(t *testing.T, content string, ext string) string {
+	t.Helper()
+
+	tmpDir := t.TempDir()
+	filePath := filepath.Join(tmpDir, "test"+ext)
+
+	err := os.WriteFile(filePath, []byte(content), 0600)
+	require.NoError(t, err)
+
+	return filePath
+}
+
+// =============================================================================
+// TESTS FOR Chunker
+// =============================================================================
+
+func TestNewChunker(t *testing.T) {
+	t.Parallel()
+
+	opts := chunking.DefaultChunkOptions()
+	c := NewChunker(opts)
+
+	assert.NotNil(t, c)
+	assert.NotNil(t, c.parser)
+}
+
+func TestChunker_Language(t *testing.T) {
+	t.Parallel()
+
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	assert.Equal(t, chunking.LanguageTypeScript, c.Language())
+}
+
+func TestChunker_SupportedExtensions(t *testing.T) {
+	t.Parallel()
+
+	c := NewChunker(chunking.DefaultChunkOptions())
+	exts := c.SupportedExtensions()
+
+	assert.Contains(t, exts, ".ts")
+	assert.Contains(t, exts, ".tsx")
+	assert.Contains(t, exts, ".js")
+	assert.Contains(t, exts, ".jsx")
+}
+
+func TestChunker_Chunk_SimpleFunction(t *testing.T) {
+	t.Parallel()
+
+	code := `function greet(name: string): string {
+    return "Hello, " + name + "!";
+}
+`
+
+	filePath := createTempTSFile(t, code, ".ts")
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	require.NotEmpty(t, chunks)
+
+	// Should find the greet function
+	var foundGreet bool
+	for _, chunk := range chunks {
+		if chunk.Name == "greet" {
+			foundGreet = true
+			assert.Equal(t, chunking.ChunkTypeFunction, chunk.Type)
+			assert.Equal(t, chunking.LanguageTypeScript, chunk.Language)
+			assert.Contains(t, chunk.Content, "function greet")
+		}
+	}
+	assert.True(t, foundGreet, "Should find 'greet' function")
+}
+
+func TestChunker_Chunk_ClassWithMethods(t *testing.T) {
+	t.Parallel()
+
+	code := `class Calculator {
+    add(a: number, b: number): number {
+        return a + b;
+    }
+
+    multiply(a: number, b: number): number {
+        return a * b;
+    }
+}
+`
+
+	filePath := createTempTSFile(t, code, ".ts")
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	require.NotEmpty(t, chunks)
+
+	// Should find the Calculator class and its methods
+	var foundClass, foundAdd, foundMultiply bool
+	for _, chunk := range chunks {
+		switch chunk.Name {
+		case "Calculator":
+			foundClass = true
+			assert.Equal(t, chunking.ChunkTypeClass, chunk.Type)
+		case "add":
+			foundAdd = true
+			assert.Equal(t, chunking.ChunkTypeMethod, chunk.Type)
+			assert.Equal(t, "Calculator", chunk.ParentName)
+		case "multiply":
+			foundMultiply = true
+			assert.Equal(t, chunking.ChunkTypeMethod, chunk.Type)
+			assert.Equal(t, "Calculator", chunk.ParentName)
+		}
+	}
+
+	assert.True(t, foundClass, "Should find 'Calculator' class")
+	assert.True(t, foundAdd, "Should find 'add' method")
+	assert.True(t, foundMultiply, "Should find 'multiply' method")
+}
+
+func TestChunker_Chunk_Interface(t *testing.T) {
+	t.Parallel()
+
+	code := `interface User {
+    id: number;
+    name: string;
+    email: string;
+}
+
+interface Authenticator {
+    login(username: string, password: string): boolean;
+    logout(): void;
+}
+`
+
+	filePath := createTempTSFile(t, code, ".ts")
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	require.NotEmpty(t, chunks)
+
+	// Should find interfaces
+	interfaceNames := make(map[string]bool)
+	for _, chunk := range chunks {
+		if chunk.Type == chunking.ChunkTypeInterface {
+			interfaceNames[chunk.Name] = true
+		}
+	}
+
+	assert.True(t, interfaceNames["User"])
+	assert.True(t, interfaceNames["Authenticator"])
+}
+
+func TestChunker_Chunk_TypeAlias(t *testing.T) {
+	t.Parallel()
+
+	code := `type UserID = string;
+
+type Handler = (event: Event) => void;
+
+type Result<T> = { success: true; data: T } | { success: false; error: Error };
+`
+
+	filePath := createTempTSFile(t, code, ".ts")
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	require.NotEmpty(t, chunks)
+
+	// Should find type aliases
+	typeNames := make(map[string]bool)
+	for _, chunk := range chunks {
+		if chunk.Type == chunking.ChunkTypeType {
+			typeNames[chunk.Name] = true
+		}
+	}
+
+	assert.True(t, typeNames["UserID"])
+	assert.True(t, typeNames["Handler"])
+	assert.True(t, typeNames["Result"])
+}
+
+func TestChunker_Chunk_ArrowFunction(t *testing.T) {
+	t.Parallel()
+
+	code := `const add = (a: number, b: number): number => a + b;
+
+const greet = (name: string): string => {
+    return "Hello, " + name;
+};
+`
+
+	filePath := createTempTSFile(t, code, ".ts")
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	_, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	// Arrow functions may or may not be captured depending on AST structure
+	// At minimum, no error should occur
+}
+
+func TestChunker_Chunk_FileNotFound(t *testing.T) {
+	t.Parallel()
+
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	_, err := c.Chunk(context.Background(), "/nonexistent/path/file.ts")
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "read file")
+}
+
+func TestChunker_Chunk_EmptyFile(t *testing.T) {
+	t.Parallel()
+
+	filePath := createTempTSFile(t, "", ".ts")
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	assert.Empty(t, chunks)
+}
+
+func TestChunker_Chunk_OnlyComments(t *testing.T) {
+	t.Parallel()
+
+	code := `// This is a comment
+/* Another comment */
+/**
+ * JSDoc comment
+ */
+`
+
+	filePath := createTempTSFile(t, code, ".ts")
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	// Comments without code should not produce chunks
+	assert.Empty(t, chunks)
+}
+
+func TestChunker_Chunk_AsyncFunction(t *testing.T) {
+	t.Parallel()
+
+	code := `async function fetchData(url: string): Promise<any> {
+    const response = await fetch(url);
+    return response.json();
+}
+
+async function processItems(items: string[]): Promise<void> {
+    for (const item of items) {
+        await process(item);
+    }
+}
+`
+
+	filePath := createTempTSFile(t, code, ".ts")
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	require.NotEmpty(t, chunks)
+
+	// Should find async functions
+	functionNames := make(map[string]bool)
+	for _, chunk := range chunks {
+		if chunk.Type == chunking.ChunkTypeFunction {
+			functionNames[chunk.Name] = true
+		}
+	}
+
+	assert.True(t, functionNames["fetchData"])
+	assert.True(t, functionNames["processItems"])
+}
+
+func TestChunker_Chunk_ExportedFunction(t *testing.T) {
+	t.Parallel()
+
+	code := `export function publicFunction(): void {
+    console.log("public");
+}
+
+export default function defaultExport(): void {
+    console.log("default");
+}
+`
+
+	filePath := createTempTSFile(t, code, ".ts")
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	require.NotEmpty(t, chunks)
+
+	// Should find exported functions
+	functionNames := make(map[string]bool)
+	for _, chunk := range chunks {
+		if chunk.Type == chunking.ChunkTypeFunction {
+			functionNames[chunk.Name] = true
+		}
+	}
+
+	assert.True(t, functionNames["publicFunction"])
+	assert.True(t, functionNames["defaultExport"])
+}
+
+func TestChunker_Chunk_JSXFile(t *testing.T) {
+	t.Parallel()
+
+	code := `function Button({ label }: { label: string }) {
+    return <button>{label}</button>;
+}
+
+function App() {
+    return (
+        <div>
+            <Button label="Click me" />
+        </div>
+    );
+}
+`
+
+	filePath := createTempTSFile(t, code, ".tsx")
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	require.NotEmpty(t, chunks)
+
+	// Should find JSX components as functions
+	functionNames := make(map[string]bool)
+	for _, chunk := range chunks {
+		if chunk.Type == chunking.ChunkTypeFunction {
+			functionNames[chunk.Name] = true
+		}
+	}
+
+	assert.True(t, functionNames["Button"])
+	assert.True(t, functionNames["App"])
+}
+
+func TestChunker_Chunk_JavaScript(t *testing.T) {
+	t.Parallel()
+
+	code := `function simpleFunc() {
+    return 42;
+}
+
+class MyClass {
+    constructor() {
+        this.value = 0;
+    }
+
+    getValue() {
+        return this.value;
+    }
+}
+`
+
+	filePath := createTempTSFile(t, code, ".js")
+	c := NewChunker(chunking.DefaultChunkOptions())
+
+	chunks, err := c.Chunk(context.Background(), filePath)
+	require.NoError(t, err)
+	require.NotEmpty(t, chunks)
+
+	// Should find JavaScript functions and classes
+	var foundFunc, foundClass bool
+	for _, chunk := range chunks {
+		if chunk.Name == "simpleFunc" {
+			foundFunc = true
+		}
+		if chunk.Name == "MyClass" {
+			foundClass = true
+		}
+	}
+
+	assert.True(t, foundFunc, "Should find 'simpleFunc' function")
+	assert.True(t, foundClass, "Should find 'MyClass' class")
+}

internal/config/config.go

@@ -5,6 +5,7 @@ import (
 	"encoding/json"
 	"os"
 	"path/filepath"
+	"strconv"
 	"strings"
 	"sync"
 )
@@ -52,6 +53,7 @@ type Config struct {
 	ContextRelevanceThreshold float64  `json:"context_relevance_threshold"`
 	RerankingCandidates       int      `json:"reranking_candidates"`
 	WorkerPort                int      `json:"worker_port"`
+	DeduplicationThreshold    float64  `json:"deduplication_threshold"`
 	RerankingMinImprovement   float64  `json:"reranking_min_improvement"`
 	ContextObservations       int      `json:"context_observations"`
 	ContextMaxPromptResults   int      `json:"context_max_prompt_results"`
@@ -64,10 +66,13 @@ type Config struct {
 	HubThreshold              int      `json:"hub_threshold"`
 	ObservationRetentionDays  int      `json:"observation_retention_days"`
 	MaintenanceIntervalHours  int      `json:"maintenance_interval_hours"`
+	ContextMaxTokensStartup   int      `json:"context_max_tokens_startup"`
+	ContextMaxTokensPrompt    int      `json:"context_max_tokens_prompt"`
 	ContextShowWorkTokens     bool     `json:"context_show_work_tokens"`
 	ContextShowReadTokens     bool     `json:"context_show_read_tokens"`
 	RerankingPureMode         bool     `json:"reranking_pure_mode"`
 	GraphEnabled              bool     `json:"graph_enabled"`
+	DeduplicationEnabled      bool     `json:"deduplication_enabled"`
 	MaintenanceEnabled        bool     `json:"maintenance_enabled"`
 	RerankingEnabled          bool     `json:"reranking_enabled"`
 	ContextShowLastSummary    bool     `json:"context_show_last_summary"`
@@ -168,6 +173,10 @@ func Default() *Config {
 		ContextObsConcepts:        DefaultObservationConcepts,
 		ContextRelevanceThreshold: 0.3,   // Minimum 30% similarity to include
 		ContextMaxPromptResults:   10,    // Cap at 10 results max (0 = no cap, threshold only)
+		ContextMaxTokensStartup:   16000, // Max tokens for SessionStart context injection
+		ContextMaxTokensPrompt:    8000,  // Max tokens for UserPromptSubmit context injection
+		DeduplicationEnabled:      true,  // Enable write-time vector dedup
+		DeduplicationThreshold:    0.9,   // Similarity threshold for merging (0.9 = very similar)
 		MaintenanceEnabled:        true,  // Enable scheduled maintenance
 		MaintenanceIntervalHours:  6,     // Run every 6 hours
 		ObservationRetentionDays:  0,     // 0 = no age-based deletion (keep all)
@@ -269,6 +278,29 @@ func Load() (*Config, error) {
 	if v, ok := settings["CLAUDE_MNEMONIC_HUB_THRESHOLD"].(float64); ok && v > 0 {
 		cfg.HubThreshold = int(v)
 	}
+	if v, ok := settings["CLAUDE_MNEMONIC_CONTEXT_MAX_TOKENS_STARTUP"].(float64); ok && v > 0 {
+		cfg.ContextMaxTokensStartup = int(v)
+	}
+	if v, ok := settings["CLAUDE_MNEMONIC_CONTEXT_MAX_TOKENS_PROMPT"].(float64); ok && v > 0 {
+		cfg.ContextMaxTokensPrompt = int(v)
+	}
+	// Deduplication settings
+	if v, ok := settings["CLAUDE_MNEMONIC_DEDUP_ENABLED"].(bool); ok {
+		cfg.DeduplicationEnabled = v
+	}
+	if v, ok := settings["CLAUDE_MNEMONIC_DEDUP_THRESHOLD"].(float64); ok && v > 0 && v <= 1 {
+		cfg.DeduplicationThreshold = v
+	}
+
+	// Also support env vars for dedup settings
+	if v := os.Getenv("CLAUDE_MNEMONIC_DEDUP_ENABLED"); v != "" {
+		cfg.DeduplicationEnabled = v == "true" || v == "1"
+	}
+	if v := os.Getenv("CLAUDE_MNEMONIC_DEDUP_THRESHOLD"); v != "" {
+		if f, err := strconv.ParseFloat(v, 64); err == nil && f > 0 && f <= 1 {
+			cfg.DeduplicationThreshold = f
+		}
+	}
 
 	return cfg, nil
 }

internal/db/gorm/pattern_store.go

@@ -4,6 +4,8 @@ package gorm
 import (
 	"context"
 	"database/sql"
+	"fmt"
+	"sync"
 	"time"
 
 	"gorm.io/gorm"
@@ -18,6 +20,7 @@ type PatternCleanupFunc func(ctx context.Context, deletedIDs []int64)
 type PatternStore struct {
 	db          *gorm.DB
 	cleanupFunc PatternCleanupFunc
+	cleanupMu   sync.RWMutex
 }
 
 // NewPatternStore creates a new pattern store.
@@ -29,6 +32,8 @@ func NewPatternStore(store *Store) *PatternStore {
 
 // SetCleanupFunc sets the callback for when patterns are deleted.
 func (s *PatternStore) SetCleanupFunc(fn PatternCleanupFunc) {
+	s.cleanupMu.Lock()
+	defer s.cleanupMu.Unlock()
 	s.cleanupFunc = fn
 }
 
@@ -238,6 +243,9 @@ func (s *PatternStore) MarkPatternDeprecated(ctx context.Context, id int64) erro
 
 // MergePatterns merges a source pattern into a target pattern.
 func (s *PatternStore) MergePatterns(ctx context.Context, sourceID, targetID int64) error {
+	if sourceID == targetID {
+		return fmt.Errorf("cannot merge pattern into itself")
+	}
 	// Get both patterns
 	source, err := s.GetPatternByID(ctx, sourceID)
 	if err != nil {
@@ -294,8 +302,13 @@ func (s *PatternStore) MergePatterns(ctx context.Context, sourceID, targetID int
 func (s *PatternStore) DeletePattern(ctx context.Context, id int64) error {
 	result := s.db.WithContext(ctx).Delete(&Pattern{}, id)
 
-	if result.Error == nil && s.cleanupFunc != nil {
-		s.cleanupFunc(ctx, []int64{id})
+	if result.Error == nil {
+		s.cleanupMu.RLock()
+		fn := s.cleanupFunc
+		s.cleanupMu.RUnlock()
+		if fn != nil {
+			fn(ctx, []int64{id})
+		}
 	}
 
 	return result.Error

internal/db/gorm/prompt_store.go

@@ -4,8 +4,10 @@ package gorm
 import (
 	"context"
 	"database/sql"
+	"sync"
 	"time"
 
+	"github.com/rs/zerolog/log"
 	"gorm.io/gorm"
 	"gorm.io/gorm/clause"
 
@@ -23,6 +25,7 @@ const MaxPromptsGlobal = 500
 type PromptStore struct {
 	db          *gorm.DB
 	cleanupFunc PromptCleanupFunc
+	cleanupMu   sync.RWMutex
 }
 
 // NewPromptStore creates a new prompt store.
@@ -35,6 +38,8 @@ func NewPromptStore(store *Store, cleanupFunc PromptCleanupFunc) *PromptStore {
 
 // SetCleanupFunc sets the callback for when prompts are deleted during cleanup.
 func (s *PromptStore) SetCleanupFunc(fn PromptCleanupFunc) {
+	s.cleanupMu.Lock()
+	defer s.cleanupMu.Unlock()
 	s.cleanupFunc = fn
 }
 
@@ -81,9 +86,15 @@ func (s *PromptStore) SaveUserPromptWithMatches(ctx context.Context, claudeSessi
 	go func() {
 		cleanupCtx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
 		defer cancel()
-		deletedIDs, _ := s.CleanupOldPrompts(cleanupCtx)
-		if len(deletedIDs) > 0 && s.cleanupFunc != nil {
-			s.cleanupFunc(cleanupCtx, deletedIDs)
+		if deletedIDs, err := s.CleanupOldPrompts(cleanupCtx); err != nil {
+			log.Warn().Err(err).Msg("Background prompt cleanup failed")
+		} else if len(deletedIDs) > 0 {
+			s.cleanupMu.RLock()
+			fn := s.cleanupFunc
+			s.cleanupMu.RUnlock()
+			if fn != nil {
+				fn(cleanupCtx, deletedIDs)
+			}
 		}
 	}()
 
@@ -253,6 +264,28 @@ func (s *PromptStore) FindRecentPromptByText(ctx context.Context, claudeSessionI
 	return prompt.ID, prompt.PromptNumber, true
 }
 
+// FindRecentPromptByTextGlobal finds a recent prompt by exact text match within a time window
+// across ALL sessions (not scoped to a single claude session ID).
+// This catches duplicates when the same prompt arrives from different session IDs,
+// e.g. when the SDK processor's callClaudeCLI subprocess fires with a different session ID.
+// Returns (promptID, promptNumber, found).
+func (s *PromptStore) FindRecentPromptByTextGlobal(ctx context.Context, promptText string, withinSeconds int) (int64, int, bool) {
+	cutoffEpoch := time.Now().Add(-time.Duration(withinSeconds) * time.Second).UnixMilli()
+
+	var prompt UserPrompt
+	err := s.db.WithContext(ctx).
+		Where("prompt_text = ? AND created_at_epoch >= ?",
+			promptText, cutoffEpoch).
+		Order("created_at_epoch DESC").
+		First(&prompt).Error
+
+	if err != nil {
+		return 0, 0, false
+	}
+
+	return prompt.ID, prompt.PromptNumber, true
+}
+
 // GetRecentUserPromptsByProject retrieves recent user prompts for a specific project.
 func (s *PromptStore) GetRecentUserPromptsByProject(ctx context.Context, project string, limit int) ([]*models.UserPromptWithSession, error) {
 	var results []struct {

internal/db/gorm/prompt_store_test.go

@@ -333,6 +333,39 @@ func TestPromptStore_FindRecentPromptByText(t *testing.T) {
 	assert.False(t, notFound, "Should not find prompt outside time window")
 }
 
+func TestPromptStore_FindRecentPromptByTextGlobal(t *testing.T) {
+	promptStore, _, cleanup := testPromptStore(t)
+	defer cleanup()
+
+	ctx := context.Background()
+
+	// Save a prompt under session "claude-1"
+	id, err := promptStore.SaveUserPromptWithMatches(ctx, "claude-1", 1, "What is the architecture?", 3)
+	require.NoError(t, err)
+
+	// Global search should find it without specifying session ID
+	foundID, foundNumber, found := promptStore.FindRecentPromptByTextGlobal(ctx, "What is the architecture?", 60)
+	assert.True(t, found, "Should find the prompt globally")
+	assert.Equal(t, id, foundID)
+	assert.Equal(t, 1, foundNumber)
+
+	// Global search should find it even when a different session ID would have been used
+	// (This is the core cross-session dedup scenario)
+	foundID2, foundNumber2, found2 := promptStore.FindRecentPromptByTextGlobal(ctx, "What is the architecture?", 60)
+	assert.True(t, found2, "Should find the prompt from any session")
+	assert.Equal(t, id, foundID2)
+	assert.Equal(t, 1, foundNumber2)
+
+	// Different text should not match
+	_, _, notFound := promptStore.FindRecentPromptByTextGlobal(ctx, "Different text", 60)
+	assert.False(t, notFound, "Should not find a different prompt")
+
+	// Should not find prompt outside time window
+	time.Sleep(100 * time.Millisecond)
+	_, _, notFound = promptStore.FindRecentPromptByTextGlobal(ctx, "What is the architecture?", 0)
+	assert.False(t, notFound, "Should not find prompt outside time window")
+}
+
 func TestPromptStore_GetRecentUserPromptsByProject(t *testing.T) {
 	promptStore, store, cleanup := testPromptStore(t)
 	defer cleanup()

internal/db/interface.go

@@ -57,6 +57,7 @@ type PromptReader interface {
 	GetAllPrompts(ctx context.Context) ([]*models.UserPromptWithSession, error)
 	GetRecentUserPromptsByProject(ctx context.Context, project string, limit int) ([]*models.UserPromptWithSession, error)
 	FindRecentPromptByText(ctx context.Context, claudeSessionID, promptText string, withinSeconds int) (int64, int, bool)
+	FindRecentPromptByTextGlobal(ctx context.Context, promptText string, withinSeconds int) (int64, int, bool)
 }
 
 // PromptWriter defines write operations for prompts.

internal/embedding/assets/lib/darwin-arm64/.version

@@ -1 +1 @@
-1.23.2
+1.24.3

internal/embedding/assets/lib/linux-amd64/.version

@@ -1 +1 @@
-1.23.2
+1.24.3

internal/embedding/assets/lib/linux-arm64/.version

@@ -1 +1 @@
-1.23.2
+1.24.3

internal/embedding/assets/lib/windows-amd64/.version

@@ -1 +1 @@
-1.23.2
+1.24.3

internal/embedding/service.go

@@ -8,6 +8,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"strings"
 	"sync"
 
 	"github.com/sugarme/tokenizer"
@@ -69,8 +70,10 @@ func newBGEModel() (EmbeddingModel, error) {
 	libPath := filepath.Join(libDir, onnxRuntimeLibName)
 	ort.SetSharedLibraryPath(libPath)
 
-	// Initialize ONNX runtime
-	if err := ort.InitializeEnvironment(); err != nil {
+	// Initialize ONNX runtime (idempotent - ignore "already initialized" since
+	// the ONNX environment is process-global and shared with the reranking service)
+	err = ort.InitializeEnvironment()
+	if err != nil && !strings.Contains(err.Error(), "already been initialized") {
 		return nil, fmt.Errorf("initialize ONNX runtime: %w", err)
 	}
 

internal/pattern/detector.go

@@ -3,6 +3,7 @@ package pattern
 
 import (
 	"context"
+	"fmt"
 	"sort"
 	"strings"
 	"sync"
@@ -370,12 +371,16 @@ func (d *Detector) CandidateCount() int {
 	return len(d.candidates)
 }
 
+// GetPatternInsight returns a formatted insight string for a pattern.
 // GetPatternInsight returns a formatted insight string for a pattern.
 func (d *Detector) GetPatternInsight(ctx context.Context, patternID int64) (string, error) {
 	pattern, err := d.patternStore.GetPatternByID(ctx, patternID)
 	if err != nil {
 		return "", err
 	}
+	if pattern == nil {
+		return "", fmt.Errorf("pattern not found: %d", patternID)
+	}
 
 	return formatPatternInsight(pattern), nil
 }

internal/pattern/detector_test.go

@@ -328,6 +328,26 @@ func TestDefaultConfig(t *testing.T) {
 	}
 }
 
+func TestDefaultConfig_AllFieldsValid(t *testing.T) {
+	config := DefaultConfig()
+
+	if config.MinMatchScore != 0.3 {
+		t.Errorf("MinMatchScore = %f, want 0.3", config.MinMatchScore)
+	}
+	if config.MinFrequencyForPattern != 2 {
+		t.Errorf("MinFrequencyForPattern = %d, want 2", config.MinFrequencyForPattern)
+	}
+	if config.AnalysisInterval != 5*time.Minute {
+		t.Errorf("AnalysisInterval = %v, want 5m", config.AnalysisInterval)
+	}
+	if config.MaxPatternsToTrack != 1000 {
+		t.Errorf("MaxPatternsToTrack = %d, want 1000", config.MaxPatternsToTrack)
+	}
+	if config.MaxCandidates != 500 {
+		t.Errorf("MaxCandidates = %d, want 500", config.MaxCandidates)
+	}
+}
+
 func TestGeneratePatternName(t *testing.T) {
 	tests := []struct {
 		patternType models.PatternType
@@ -352,6 +372,85 @@ func TestGeneratePatternName(t *testing.T) {
 	}
 }
 
+func TestGeneratePatternName_EdgeCases(t *testing.T) {
+	tests := []struct {
+		name      string
+		ptype     models.PatternType
+		title     string
+		want      string
+		signature []string
+	}{
+		{
+			name:      "with title uses title directly",
+			ptype:     models.PatternTypeBug,
+			signature: []string{"ignored"},
+			title:     "Custom Title",
+			want:      "Custom Title",
+		},
+		{
+			name:      "long title generates from signature",
+			ptype:     models.PatternTypeBug,
+			signature: []string{"sig1", "sig2"},
+			title:     "This is a very long title that exceeds sixty characters and should be ignored",
+			want:      "Bug Pattern: sig1, sig2",
+		},
+		{
+			name:      "empty signature returns Unnamed",
+			ptype:     models.PatternTypeBug,
+			signature: []string{},
+			title:     "",
+			want:      "Bug Pattern: Unnamed",
+		},
+		{
+			name:      "single signature element",
+			ptype:     models.PatternTypeRefactor,
+			signature: []string{"single"},
+			title:     "",
+			want:      "Refactor Pattern: single",
+		},
+		{
+			name:      "more than 3 signature elements truncates",
+			ptype:     models.PatternTypeBestPractice,
+			signature: []string{"a", "b", "c", "d", "e"},
+			title:     "",
+			want:      "Best Practice: a, b, c",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got := generatePatternName(tt.ptype, tt.signature, tt.title)
+			if got != tt.want {
+				t.Errorf("generatePatternName() = %q, want %q", got, tt.want)
+			}
+		})
+	}
+}
+
+func TestGeneratePatternName_AllTypes(t *testing.T) {
+	tests := []struct {
+		ptype      models.PatternType
+		wantPrefix string
+	}{
+		{models.PatternTypeBug, "Bug Pattern:"},
+		{models.PatternTypeRefactor, "Refactor Pattern:"},
+		{models.PatternTypeArchitecture, "Architecture Pattern:"},
+		{models.PatternTypeAntiPattern, "Anti-Pattern:"},
+		{models.PatternTypeBestPractice, "Best Practice:"},
+		{models.PatternType("unknown"), "test"}, // Unknown type has empty prefix, starts with first signature element
+	}
+
+	for _, tt := range tests {
+		t.Run(string(tt.ptype), func(t *testing.T) {
+			name := generatePatternName(tt.ptype, []string{"test", "sig"}, "")
+			if !hasPrefix(name, tt.wantPrefix) {
+				t.Errorf("Expected prefix %q for type %s, got: %s",
+					tt.wantPrefix, tt.ptype, name)
+			}
+		})
+	}
+}
+
 func TestFormatPatternInsight(t *testing.T) {
 	// Pattern without recommendation
 	pattern1 := &models.Pattern{
@@ -386,6 +485,470 @@ func TestFormatPatternInsight(t *testing.T) {
 	}
 }
 
+func TestFormatPatternInsight_AllTypes(t *testing.T) {
+	types := []struct {
+		ptype    models.PatternType
+		contains string
+	}{
+		{models.PatternTypeBug, "bug pattern"},
+		{models.PatternTypeRefactor, "recognized pattern"},     // Falls to default case
+		{models.PatternTypeArchitecture, "recognized pattern"}, // Falls to default case
+		{models.PatternTypeAntiPattern, "anti-pattern"},
+		{models.PatternTypeBestPractice, "best practice"},
+		{models.PatternType("unknown"), "recognized pattern"}, // Falls to default case
+	}
+
+	for _, tt := range types {
+		t.Run(string(tt.ptype), func(t *testing.T) {
+			pattern := &models.Pattern{
+				Type:      tt.ptype,
+				Frequency: 3,
+				Projects:  []string{"proj1"},
+			}
+			insight := formatPatternInsight(pattern)
+			if !containsString(insight, tt.contains) {
+				t.Errorf("Expected insight to contain %q for type %s, got: %s",
+					tt.contains, tt.ptype, insight)
+			}
+		})
+	}
+}
+
+func TestFormatPatternInsight_MultiProject(t *testing.T) {
+	pattern := &models.Pattern{
+		Type:      models.PatternTypeBug,
+		Frequency: 10,
+		Projects:  []string{"proj1", "proj2", "proj3"},
+	}
+
+	insight := formatPatternInsight(pattern)
+
+	if !containsString(insight, "10 times") {
+		t.Error("Expected frequency in insight")
+	}
+	if !containsString(insight, "3 projects") {
+		t.Error("Expected project count in insight")
+	}
+}
+
+func TestFormatPatternInsight_SingleProject(t *testing.T) {
+	pattern := &models.Pattern{
+		Type:      models.PatternTypeBestPractice,
+		Frequency: 5,
+		Projects:  []string{"only-one"},
+	}
+
+	insight := formatPatternInsight(pattern)
+
+	if !containsString(insight, "5 times") {
+		t.Error("Expected frequency in insight")
+	}
+	// Single project should NOT mention "projects"
+	if containsString(insight, "projects") {
+		t.Error("Single project should not mention 'projects'")
+	}
+}
+
+func TestDetector_SetSyncFunc(t *testing.T) {
+	store := setupTestStore(t)
+	defer store.Close()
+
+	patternStore := gorm.NewPatternStore(store)
+	observationStore := gorm.NewObservationStore(store, nil, nil, nil)
+	config := DefaultConfig()
+
+	detector := NewDetector(patternStore, observationStore, config)
+
+	// Initially nil
+	if detector.syncFunc != nil {
+		t.Error("Expected syncFunc to be nil initially")
+	}
+
+	// Set sync func
+	var syncCalled bool
+	detector.SetSyncFunc(func(p *models.Pattern) {
+		syncCalled = true
+	})
+
+	if detector.syncFunc == nil {
+		t.Error("Expected syncFunc to be set")
+	}
+
+	// Verify it can be called
+	detector.syncFunc(&models.Pattern{})
+	if !syncCalled {
+		t.Error("Expected sync function to be called")
+	}
+}
+
+func TestDetector_CandidateCount(t *testing.T) {
+	store := setupTestStore(t)
+	defer store.Close()
+
+	patternStore := gorm.NewPatternStore(store)
+	observationStore := gorm.NewObservationStore(store, nil, nil, nil)
+	config := DefaultConfig()
+
+	detector := NewDetector(patternStore, observationStore, config)
+
+	// Initially zero
+	if count := detector.CandidateCount(); count != 0 {
+		t.Errorf("Expected 0 candidates, got %d", count)
+	}
+
+	// Add some candidates
+	detector.candidates["key1"] = &candidatePattern{}
+	detector.candidates["key2"] = &candidatePattern{}
+
+	if count := detector.CandidateCount(); count != 2 {
+		t.Errorf("Expected 2 candidates, got %d", count)
+	}
+}
+
+func TestDetector_AnalyzeRecentObservations(t *testing.T) {
+	store := setupTestStore(t)
+	defer store.Close()
+
+	patternStore := gorm.NewPatternStore(store)
+	observationStore := gorm.NewObservationStore(store, nil, nil, nil)
+	config := DefaultConfig()
+
+	detector := NewDetector(patternStore, observationStore, config)
+	ctx := context.Background()
+
+	// Should not error even with no observations
+	err := detector.AnalyzeRecentObservations(ctx)
+	if err != nil {
+		t.Fatalf("AnalyzeRecentObservations() error = %v", err)
+	}
+}
+
+func TestGenerateCandidateKey(t *testing.T) {
+	tests := []struct {
+		name      string
+		want      string
+		signature []string
+	}{
+		{
+			name:      "single element",
+			signature: []string{"error"},
+			want:      "error|",
+		},
+		{
+			name:      "multiple elements",
+			signature: []string{"error", "handling", "nil"},
+			want:      "error|handling|nil|",
+		},
+		{
+			name:      "empty signature",
+			signature: []string{},
+			want:      "",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got := generateCandidateKey(tt.signature)
+			if got != tt.want {
+				t.Errorf("generateCandidateKey() = %q, want %q", got, tt.want)
+			}
+		})
+	}
+}
+
+func TestGenerateCandidateKey_EdgeCases(t *testing.T) {
+	tests := []struct {
+		name      string
+		want      string
+		signature []string
+	}{
+		{
+			name:      "nil signature",
+			signature: nil,
+			want:      "",
+		},
+		{
+			name:      "empty strings in signature",
+			signature: []string{"", ""},
+			want:      "||",
+		},
+		{
+			name:      "special characters",
+			signature: []string{"error|handling", "nil"},
+			want:      "error|handling|nil|",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got := generateCandidateKey(tt.signature)
+			if got != tt.want {
+				t.Errorf("generateCandidateKey() = %q, want %q", got, tt.want)
+			}
+		})
+	}
+}
+
+func TestItoa(t *testing.T) {
+	tests := []struct {
+		want  string
+		input int
+	}{
+		{"0", 0},
+		{"1", 1},
+		{"10", 10},
+		{"123", 123},
+		{"-1", -1},
+		{"-123", -123},
+		{"1000000", 1000000},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.want, func(t *testing.T) {
+			got := itoa(tt.input)
+			if got != tt.want {
+				t.Errorf("itoa(%d) = %q, want %q", tt.input, got, tt.want)
+			}
+		})
+	}
+}
+
+func TestItoa_EdgeCases(t *testing.T) {
+	tests := []struct {
+		want  string
+		input int
+	}{
+		{"0", 0},
+		{"0", -0},
+		{"1", 1},
+		{"-1", -1},
+		{"9", 9},
+		{"10", 10},
+		{"99", 99},
+		{"100", 100},
+		{"999", 999},
+		{"1000", 1000},
+		{"-999", -999},
+		{"-1000", -1000},
+		{"2147483647", 2147483647},   // Max int32
+		{"-2147483647", -2147483647}, // Min int32 + 1
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.want, func(t *testing.T) {
+			got := itoa(tt.input)
+			if got != tt.want {
+				t.Errorf("itoa(%d) = %q, want %q", tt.input, got, tt.want)
+			}
+		})
+	}
+}
+
+func TestDetectionResult_ZeroValue(t *testing.T) {
+	result := &DetectionResult{}
+
+	if result.MatchedPattern != nil {
+		t.Error("Zero value should have nil MatchedPattern")
+	}
+	if result.MatchScore != 0 {
+		t.Error("Zero value should have 0 MatchScore")
+	}
+	if result.IsNewPattern {
+		t.Error("Zero value should have false IsNewPattern")
+	}
+}
+
+func TestCandidatePattern_Fields(t *testing.T) {
+	candidate := &candidatePattern{
+		patternType:    models.PatternTypeBug,
+		title:          "Test Title",
+		signature:      []string{"sig1", "sig2"},
+		observationIDs: []int64{1, 2, 3},
+		projects:       []string{"proj1", "proj2"},
+		lastSeenEpoch:  time.Now().UnixMilli(),
+	}
+
+	if candidate.patternType != models.PatternTypeBug {
+		t.Error("Wrong pattern type")
+	}
+	if candidate.title != "Test Title" {
+		t.Error("Wrong title")
+	}
+	if len(candidate.signature) != 2 {
+		t.Error("Wrong signature length")
+	}
+	if len(candidate.observationIDs) != 3 {
+		t.Error("Wrong observationIDs length")
+	}
+	if len(candidate.projects) != 2 {
+		t.Error("Wrong projects length")
+	}
+}
+
+func TestDetector_AnalyzeObservation_EmptySignature(t *testing.T) {
+	store := setupTestStore(t)
+	defer store.Close()
+
+	patternStore := gorm.NewPatternStore(store)
+	observationStore := gorm.NewObservationStore(store, nil, nil, nil)
+	config := DefaultConfig()
+
+	detector := NewDetector(patternStore, observationStore, config)
+	ctx := context.Background()
+
+	// Create observation with empty concepts/title/narrative
+	obs := &models.Observation{
+		ID:           1,
+		SDKSessionID: "test-session",
+		Project:      "test-project",
+		Scope:        models.ScopeProject,
+		Type:         models.ObsTypeBugfix,
+		// All fields that would create signature are empty
+	}
+
+	result, err := detector.AnalyzeObservation(ctx, obs)
+	if err != nil {
+		t.Fatalf("AnalyzeObservation() error = %v", err)
+	}
+
+	// Should return empty result for empty signature
+	if result.MatchedPattern != nil {
+		t.Error("Expected nil pattern for empty signature")
+	}
+}
+
+func TestDetector_AnalyzeObservation_CandidateEviction(t *testing.T) {
+	store := setupTestStore(t)
+	defer store.Close()
+
+	patternStore := gorm.NewPatternStore(store)
+	observationStore := gorm.NewObservationStore(store, nil, nil, nil)
+	config := DefaultConfig()
+	config.MaxCandidates = 2           // Very small for testing
+	config.MinFrequencyForPattern = 10 // High so nothing gets promoted
+
+	detector := NewDetector(patternStore, observationStore, config)
+	ctx := context.Background()
+
+	// Add observations with different signatures until we exceed MaxCandidates
+	obs1 := createTestObservation(1, "First", []string{"first", "unique"})
+	obs2 := createTestObservation(2, "Second", []string{"second", "unique"})
+	obs3 := createTestObservation(3, "Third", []string{"third", "unique"})
+
+	// Analyze all observations
+	_, _ = detector.AnalyzeObservation(ctx, obs1)
+	time.Sleep(10 * time.Millisecond) // Small delay so timestamps differ
+	_, _ = detector.AnalyzeObservation(ctx, obs2)
+	time.Sleep(10 * time.Millisecond)
+	_, _ = detector.AnalyzeObservation(ctx, obs3)
+
+	// Should have at most MaxCandidates
+	if count := detector.CandidateCount(); count > config.MaxCandidates {
+		t.Errorf("Expected at most %d candidates, got %d", config.MaxCandidates, count)
+	}
+}
+
+func TestDetector_PromoteCandidateWithSyncFunc(t *testing.T) {
+	store := setupTestStore(t)
+	defer store.Close()
+
+	patternStore := gorm.NewPatternStore(store)
+	observationStore := gorm.NewObservationStore(store, nil, nil, nil)
+	config := DefaultConfig()
+	config.MinFrequencyForPattern = 2
+
+	detector := NewDetector(patternStore, observationStore, config)
+	ctx := context.Background()
+
+	// Set up sync function to track calls
+	var syncedPattern *models.Pattern
+	detector.SetSyncFunc(func(p *models.Pattern) {
+		syncedPattern = p
+	})
+
+	// Create two similar observations to trigger pattern promotion
+	obs1 := createTestObservation(1, "Sync Test", []string{"sync", "test"})
+	obs2 := createTestObservation(2, "Sync Test", []string{"sync", "test"})
+
+	_, _ = detector.AnalyzeObservation(ctx, obs1)
+	result, _ := detector.AnalyzeObservation(ctx, obs2)
+
+	if result.MatchedPattern == nil {
+		t.Fatal("Expected pattern to be created")
+	}
+
+	if syncedPattern == nil {
+		t.Error("Expected sync function to be called")
+	}
+
+	if syncedPattern != nil && syncedPattern.Name != result.MatchedPattern.Name {
+		t.Errorf("Synced pattern name mismatch: got %s, want %s",
+			syncedPattern.Name, result.MatchedPattern.Name)
+	}
+}
+
+func TestDetector_AnalyzeObservation_UpdateExistingCandidate(t *testing.T) {
+	store := setupTestStore(t)
+	defer store.Close()
+
+	patternStore := gorm.NewPatternStore(store)
+	observationStore := gorm.NewObservationStore(store, nil, nil, nil)
+	config := DefaultConfig()
+	config.MinFrequencyForPattern = 5 // High enough that we don't promote
+
+	detector := NewDetector(patternStore, observationStore, config)
+	ctx := context.Background()
+
+	// Create observations with same signature
+	obs1 := createTestObservation(1, "Update Test", []string{"update", "test"})
+	obs2 := createTestObservation(2, "Update Test", []string{"update", "test"})
+	obs2.Project = "different-project"
+
+	// Analyze first observation
+	_, _ = detector.AnalyzeObservation(ctx, obs1)
+
+	// Check candidate count
+	if count := detector.CandidateCount(); count != 1 {
+		t.Errorf("Expected 1 candidate after first obs, got %d", count)
+	}
+
+	// Analyze second observation
+	_, _ = detector.AnalyzeObservation(ctx, obs2)
+
+	// Still 1 candidate (same signature)
+	if count := detector.CandidateCount(); count != 1 {
+		t.Errorf("Expected 1 candidate after second obs, got %d", count)
+	}
+
+	// Check that candidate has both projects
+	key := generateCandidateKey([]string{"update", "test"})
+	candidate := detector.candidates[key]
+	if candidate == nil {
+		t.Fatal("Expected candidate to exist")
+	}
+	if len(candidate.projects) != 2 {
+		t.Errorf("Expected 2 projects, got %d", len(candidate.projects))
+	}
+}
+
+func TestDetector_GetPatternInsight_NotFound(t *testing.T) {
+	store := setupTestStore(t)
+	defer store.Close()
+
+	patternStore := gorm.NewPatternStore(store)
+	observationStore := gorm.NewObservationStore(store, nil, nil, nil)
+	config := DefaultConfig()
+
+	detector := NewDetector(patternStore, observationStore, config)
+	ctx := context.Background()
+
+	// Try to get insight for non-existent pattern
+	_, err := detector.GetPatternInsight(ctx, 99999)
+	if err == nil {
+		t.Error("Expected error for non-existent pattern")
+	}
+}
+
 // Helper functions
 
 func setupTestStore(t *testing.T) *gorm.Store {

internal/worker/handlers.go

@@ -13,6 +13,7 @@ import (
 	"fmt"
 	"net/http"
 	"strconv"
+	"time"
 
 	"github.com/rs/zerolog/log"
 )
@@ -142,19 +143,55 @@ func formatWarning(format string, args ...any) string {
 }
 
 // handleHealth handles health check requests.
-// Returns 200 OK immediately (even during init) so hooks can connect quickly.
-// Use /api/ready for full readiness check.
+// Returns 200 when ready, 503 when initializing or degraded.
 func (s *Service) handleHealth(w http.ResponseWriter, r *http.Request) {
-	status := "starting"
-	if s.ready.Load() {
-		status = "ready"
-	} else if err := s.GetInitError(); err != nil {
-		status = "error"
+	status := "ready"
+	dbStatus := "ok"
+	embeddingStatus := "ok"
+
+	if !s.ready.Load() {
+		status = "initializing"
+		if err := s.GetInitError(); err != nil {
+			status = "error"
+		}
 	}
-	writeJSON(w, map[string]any{
-		"status":  status,
-		"version": s.version,
-	})
+
+	// Check embedding service
+	if s.embedSvc == nil {
+		embeddingStatus = "unavailable"
+		if status == "ready" {
+			status = "degraded"
+		}
+	}
+
+	// Check DB
+	if s.store == nil {
+		dbStatus = "unavailable"
+		if status == "ready" {
+			status = "degraded"
+		}
+	}
+
+	activeSessions := 0
+	if s.sessionManager != nil {
+		activeSessions = s.sessionManager.GetActiveSessionCount()
+	}
+
+	resp := map[string]any{
+		"status":           status,
+		"ready":            s.ready.Load(),
+		"uptime_seconds":   int(time.Since(s.startTime).Seconds()),
+		"active_sessions":  activeSessions,
+		"db_status":        dbStatus,
+		"embedding_status": embeddingStatus,
+		"version":          s.version,
+	}
+
+	w.Header().Set("Content-Type", "application/json")
+	if status != "ready" {
+		w.WriteHeader(http.StatusServiceUnavailable)
+	}
+	json.NewEncoder(w).Encode(resp)
 }
 
 // handleVersion returns the worker version for version checking.

internal/worker/handlers_data.go

@@ -593,3 +593,118 @@ func (s *Service) handleGetObservationByID(w http.ResponseWriter, r *http.Reques
 
 	writeJSON(w, obs)
 }
+
+// handleGraphStats returns graph statistics for the dashboard.
+// Uses relation data to compute knowledge graph metrics.
+func (s *Service) handleGraphStats(w http.ResponseWriter, r *http.Request) {
+	// Get relation count (edges) - this represents the knowledge graph
+	edgeCount, err := s.relationStore.GetTotalRelationCount(r.Context())
+	if err != nil {
+		edgeCount = 0
+	}
+
+	// Count by relation type
+	edgeTypes := make(map[string]int)
+	for _, t := range models.AllRelationTypes {
+		relations, err := s.relationStore.GetRelationsByType(r.Context(), t, 10000)
+		if err == nil {
+			edgeTypes[string(t)] = len(relations)
+		}
+	}
+
+	// Get unique observation IDs involved in relations (approximate node count)
+	// For now, use edge count as a proxy - each edge has 2 nodes
+	nodeCount := 0
+	if edgeCount > 0 {
+		// Rough estimate: unique nodes ≈ edges * 1.5 (since nodes can have multiple edges)
+		nodeCount = int(float64(edgeCount) * 1.5)
+	}
+
+	// Calculate average degree
+	var avgDegree float64
+	if nodeCount > 0 {
+		avgDegree = float64(edgeCount*2) / float64(nodeCount)
+	}
+
+	// Graph is enabled if we have any edges (relations)
+	enabled := edgeCount > 0
+
+	writeJSON(w, map[string]any{
+		"enabled":      enabled,
+		"nodeCount":    nodeCount,
+		"edgeCount":    edgeCount,
+		"avgDegree":    avgDegree,
+		"maxDegree":    0,
+		"minDegree":    0,
+		"medianDegree": 0.0,
+		"edgeTypes":    edgeTypes,
+		"config": map[string]any{
+			"maxHops":            2,
+			"branchFactor":       10,
+			"edgeWeight":         0.3,
+			"rebuildIntervalMin": 30,
+		},
+	})
+}
+
+// handleVectorMetrics returns vector database metrics for the dashboard.
+// Returns enabled: false if vector features are not available.
+func (s *Service) handleVectorMetrics(w http.ResponseWriter, r *http.Request) {
+	if s.vectorClient == nil {
+		writeJSON(w, map[string]any{
+			"enabled": false,
+			"message": "Vector database not initialized",
+		})
+		return
+	}
+
+	// Get cache stats from vector client
+	cacheSize, cacheMax := s.vectorClient.CacheStats()
+	cacheStats := s.vectorClient.GetCacheStats()
+	count, _ := s.vectorClient.Count(r.Context())
+
+	uptime := time.Since(s.startTime).Round(time.Second).String()
+
+	// Calculate total queries from cache hits/misses
+	totalQueries := cacheStats.EmbeddingHits + cacheStats.EmbeddingMisses + cacheStats.ResultHits + cacheStats.ResultMisses
+	totalHits := cacheStats.EmbeddingHits + cacheStats.ResultHits
+	totalMisses := cacheStats.EmbeddingMisses + cacheStats.ResultMisses
+
+	writeJSON(w, map[string]any{
+		"enabled": true,
+		"queries": map[string]any{
+			"total":    totalQueries,
+			"hubOnly":  0,
+			"hybrid":   0,
+			"onDemand": 0,
+			"graph":    0,
+		},
+		"latency": map[string]any{
+			"avg":          "0ms",
+			"p50":          "0ms",
+			"p95":          "0ms",
+			"p99":          "0ms",
+			"avgHub":       "0ms",
+			"avgRecompute": "0ms",
+		},
+		"storage": map[string]any{
+			"totalDocuments":   count,
+			"hubDocuments":     0,
+			"storedEmbeddings": count,
+			"savingsPercent":   0.0,
+			"recomputedTotal":  0,
+		},
+		"cache": map[string]any{
+			"hits":    totalHits,
+			"misses":  totalMisses,
+			"hitRate": cacheStats.HitRate(),
+			"size":    cacheSize,
+			"maxSize": cacheMax,
+		},
+		"graph": map[string]any{
+			"traversals": 0,
+			"avgDepth":   0.0,
+		},
+		"uptime": uptime,
+	})
+}

internal/worker/handlers_relations.go

@@ -46,7 +46,7 @@ func (s *Service) handleGetRelationGraph(w http.ResponseWriter, r *http.Request)
 	// Get depth parameter (default 2)
 	depth := 2
 	if depthStr := r.URL.Query().Get("depth"); depthStr != "" {
-		if d, err := strconv.Atoi(depthStr); err == nil && d > 0 && d <= 5 {
+		if d, parseErr := strconv.Atoi(depthStr); parseErr == nil && d > 0 && d <= 5 {
 			depth = d
 		}
 	}
@@ -72,7 +72,7 @@ func (s *Service) handleGetRelatedObservations(w http.ResponseWriter, r *http.Re
 	// Get minimum confidence parameter (default 0.4)
 	minConfidence := 0.4
 	if confStr := r.URL.Query().Get("min_confidence"); confStr != "" {
-		if c, err := strconv.ParseFloat(confStr, 64); err == nil && c >= 0 && c <= 1 {
+		if c, parseErr := strconv.ParseFloat(confStr, 64); parseErr == nil && c >= 0 && c <= 1 {
 			minConfidence = c
 		}
 	}

internal/worker/handlers_scoring.go

@@ -42,11 +42,9 @@ func (s *Service) handleObservationFeedback(w http.ResponseWriter, r *http.Reque
 		return
 	}
 
-	// Get required components
-	s.initMu.RLock()
+	// Get required components (initMu.RLock held by requireReady middleware)
 	observationStore := s.observationStore
 	scoreCalculator := s.scoreCalculator
-	s.initMu.RUnlock()
 
 	if observationStore == nil {
 		http.Error(w, "service not ready", http.StatusServiceUnavailable)
@@ -95,10 +93,9 @@ func (s *Service) handleObservationFeedback(w http.ResponseWriter, r *http.Reque
 func (s *Service) handleGetScoringStats(w http.ResponseWriter, r *http.Request) {
 	project := r.URL.Query().Get("project")
 
-	s.initMu.RLock()
+	// initMu.RLock held by requireReady middleware
 	observationStore := s.observationStore
 	recalculator := s.recalculator
-	s.initMu.RUnlock()
 
 	if observationStore == nil {
 		http.Error(w, "service not ready", http.StatusServiceUnavailable)
@@ -130,9 +127,8 @@ func (s *Service) handleGetTopObservations(w http.ResponseWriter, r *http.Reques
 	limit := parseIntParam(r, "limit", 10)
 	project := r.URL.Query().Get("project")
 
-	s.initMu.RLock()
+	// initMu.RLock held by requireReady middleware
 	observationStore := s.observationStore
-	s.initMu.RUnlock()
 
 	if observationStore == nil {
 		http.Error(w, "service not ready", http.StatusServiceUnavailable)
@@ -158,9 +154,8 @@ func (s *Service) handleGetMostRetrieved(w http.ResponseWriter, r *http.Request)
 	limit := parseIntParam(r, "limit", 10)
 	project := r.URL.Query().Get("project")
 
-	s.initMu.RLock()
+	// initMu.RLock held by requireReady middleware
 	observationStore := s.observationStore
-	s.initMu.RUnlock()
 
 	if observationStore == nil {
 		http.Error(w, "service not ready", http.StatusServiceUnavailable)
@@ -191,10 +186,9 @@ func (s *Service) handleExplainScore(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 
-	s.initMu.RLock()
+	// initMu.RLock held by requireReady middleware
 	observationStore := s.observationStore
 	scoreCalculator := s.scoreCalculator
-	s.initMu.RUnlock()
 
 	if observationStore == nil || scoreCalculator == nil {
 		http.Error(w, "service not ready", http.StatusServiceUnavailable)
@@ -245,10 +239,9 @@ func (s *Service) handleUpdateConceptWeight(w http.ResponseWriter, r *http.Reque
 		return
 	}
 
-	s.initMu.RLock()
+	// initMu.RLock held by requireReady middleware
 	observationStore := s.observationStore
 	recalculator := s.recalculator
-	s.initMu.RUnlock()
 
 	if observationStore == nil {
 		http.Error(w, "service not ready", http.StatusServiceUnavailable)
@@ -279,9 +272,8 @@ func (s *Service) handleUpdateConceptWeight(w http.ResponseWriter, r *http.Reque
 // handleGetConceptWeights returns all concept weights.
 // GET /api/scoring/concepts
 func (s *Service) handleGetConceptWeights(w http.ResponseWriter, r *http.Request) {
-	s.initMu.RLock()
+	// initMu.RLock held by requireReady middleware
 	observationStore := s.observationStore
-	s.initMu.RUnlock()
 
 	if observationStore == nil {
 		http.Error(w, "service not ready", http.StatusServiceUnavailable)
@@ -300,19 +292,22 @@ func (s *Service) handleGetConceptWeights(w http.ResponseWriter, r *http.Request
 // handleTriggerRecalculation triggers an immediate score recalculation.
 // POST /api/scoring/recalculate
 func (s *Service) handleTriggerRecalculation(w http.ResponseWriter, r *http.Request) {
-	s.initMu.RLock()
+	// initMu.RLock held by requireReady middleware
 	recalculator := s.recalculator
-	s.initMu.RUnlock()
 
 	if recalculator == nil {
 		http.Error(w, "recalculator not available", http.StatusServiceUnavailable)
 		return
 	}
 
-	// Run recalculation in background
+	// Run recalculation in background with independent context
+	s.wg.Add(1)
 	go func() {
-		if err := recalculator.RecalculateNow(r.Context()); err != nil {
-			log.Warn().Err(err).Msg("Background score recalculation failed")
+		defer s.wg.Done()
+		ctx, cancel := context.WithTimeout(context.Background(), 60*time.Second)
+		defer cancel()
+		if err := recalculator.RecalculateNow(ctx); err != nil {
+			log.Error().Err(err).Msg("Background recalculation failed")
 		}
 	}()
 
@@ -336,27 +331,24 @@ func (s *Service) incrementRetrievalCounts(ids []int64) {
 		return
 	}
 
-	s.initMu.RLock()
+	// initMu.RLock held by requireReady middleware (caller is always behind requireReady)
 	store := s.observationStore
-	s.initMu.RUnlock()
 
 	if store == nil {
 		return
 	}
 
 	// Increment in background to not block response
-	// Use service context to respect shutdown signals
 	s.wg.Add(1)
 	go func() {
 		defer s.wg.Done()
-		ctx, cancel := context.WithTimeout(s.ctx, 3*time.Second)
+		// Create a new context with timeout for the background operation
+		ctx, cancel := context.WithTimeout(context.Background(), 3*time.Second)
 		defer cancel()
 
 		if err := store.IncrementRetrievalCount(ctx, ids); err != nil {
 			// Log but don't fail - this is a background operation
-			if s.ctx.Err() == nil { // Don't log during shutdown
-				log.Debug().Err(err).Msg("Failed to increment retrieval counts")
-			}
+			_ = err // Explicitly ignore - background operation
 		}
 	}()
 }

internal/worker/handlers_sessions.go

@@ -6,6 +6,7 @@ import (
 	"encoding/json"
 	"net/http"
 	"strconv"
+	"strings"
 	"time"
 
 	"github.com/go-chi/chi/v5"
@@ -44,7 +45,30 @@ func (s *Service) handleSessionInit(w http.ResponseWriter, r *http.Request) {
 		http.Error(w, err.Error(), http.StatusBadRequest)
 		return
 	}
+	// Reject requests from internal CLI calls (callClaudeCLI runs from /tmp with no session ID).
+	// These are the memory extraction agent's own prompts, not real user prompts.
+	if req.ClaudeSessionID == "" {
+		log.Debug().Str("project", req.Project).Msg("Rejecting session init with empty claude session ID (internal call)")
+		writeJSON(w, SessionInitResponse{
+			Skipped: true,
+			Reason:  "internal",
+		})
+		return
+	}
 
+	// Reject prompts that contain the internal system prompt (from SDK processor).
+	// The SDK processor sets CLAUDE_MNEMONIC_INTERNAL=1 on the subprocess env,
+	// but Claude Code does NOT propagate custom env vars to hook subprocesses,
+	// so that guard fails and the system prompt leaks into the database as a
+	// regular user prompt. This server-side check catches those cases reliably.
+	if strings.Contains(req.Prompt, "You are a memory extraction agent for Claude Code sessions") {
+		log.Debug().Str("project", req.Project).Msg("Rejecting session init with internal system prompt")
+		writeJSON(w, SessionInitResponse{
+			Skipped: true,
+			Reason:  "internal",
+		})
+		return
+	}
 	// Privacy check
 	if privacy.IsEntirelyPrivate(req.Prompt) {
 		// Create session but skip processing
@@ -84,6 +108,24 @@ func (s *Service) handleSessionInit(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 
+	// CROSS-SESSION DUPLICATE DETECTION: Same prompt text from ANY session within the window.
+	// This catches duplicates when the SDK processor's callClaudeCLI subprocess fires
+	// UserPromptSubmit with a different Claude session ID than the original hook.
+	if _, existingNum, found := s.promptStore.FindRecentPromptByTextGlobal(r.Context(), cleanedPrompt, DuplicatePromptWindowSeconds); found {
+		sessionID, _ := s.sessionStore.CreateSDKSession(r.Context(), req.ClaudeSessionID, req.Project, cleanedPrompt)
+
+		log.Debug().
+			Int64("sessionId", sessionID).
+			Int("promptNumber", existingNum).
+			Msg("Cross-session duplicate prompt detected")
+
+		writeJSON(w, SessionInitResponse{
+			SessionDBID:  sessionID,
+			PromptNumber: existingNum,
+		})
+		return
+	}
+
 	// Create session (idempotent)
 	sessionID, err := s.sessionStore.CreateSDKSession(r.Context(), req.ClaudeSessionID, req.Project, cleanedPrompt)
 	if err != nil {
@@ -210,6 +252,12 @@ func (s *Service) handleObservation(w http.ResponseWriter, r *http.Request) {
 		http.Error(w, err.Error(), http.StatusBadRequest)
 		return
 	}
+	// Reject requests from internal CLI calls (no session ID = internal callClaudeCLI)
+	if req.ClaudeSessionID == "" {
+		w.WriteHeader(http.StatusOK)
+		return
+	}
+
 
 	// Find session
 	sess, err := s.sessionStore.FindAnySDKSession(r.Context(), req.ClaudeSessionID)

internal/worker/handlers_test.go

@@ -459,14 +459,13 @@ func TestHandleHealth_ReturnsVersion(t *testing.T) {
 
 	svc.handleHealth(rec, req)
 
-	assert.Equal(t, http.StatusOK, rec.Code)
-
 	var response map[string]interface{}
 	err := json.Unmarshal(rec.Body.Bytes(), &response)
 	require.NoError(t, err)
 
-	assert.Equal(t, "ready", response["status"])
 	assert.Equal(t, "test-version-1.2.3", response["version"])
+	// Status may be "degraded" if embedSvc is nil in test, but version is always present
+	assert.Contains(t, []string{"ready", "degraded"}, response["status"])
 }
 
 func TestHandleVersion(t *testing.T) {
@@ -2028,13 +2027,14 @@ func TestHandleHealth_NotReady(t *testing.T) {
 
 	svc.handleHealth(rec, req)
 
-	assert.Equal(t, http.StatusOK, rec.Code)
+	assert.Equal(t, http.StatusServiceUnavailable, rec.Code)
 
 	var response map[string]interface{}
 	err := json.Unmarshal(rec.Body.Bytes(), &response)
 	require.NoError(t, err)
 
-	assert.Equal(t, "starting", response["status"])
+	assert.Equal(t, "initializing", response["status"])
+	assert.Equal(t, false, response["ready"])
 }
 
 // TestHandleContextInject_EmptyProject tests context inject with empty project.
@@ -2399,7 +2399,12 @@ func TestHandleHealthEndpoint(t *testing.T) {
 
 	svc.router.ServeHTTP(rec, req)
 
-	assert.Equal(t, http.StatusOK, rec.Code)
+	// Response is valid JSON with health details
+	var response map[string]interface{}
+	err := json.Unmarshal(rec.Body.Bytes(), &response)
+	require.NoError(t, err)
+	assert.NotNil(t, response["status"])
+	assert.NotNil(t, response["version"])
 }
 
 // TestHandleSelfCheckEndpoint tests self-check endpoint via router.
@@ -2894,12 +2899,18 @@ func TestHandleHealth(t *testing.T) {
 
 	svc.router.ServeHTTP(rec, req)
 
-	assert.Equal(t, http.StatusOK, rec.Code)
-
 	var response map[string]interface{}
 	err := json.Unmarshal(rec.Body.Bytes(), &response)
 	require.NoError(t, err)
-	assert.Equal(t, "ready", response["status"])
+
+	// Test service has store set but no embedSvc, so status is "degraded"
+	assert.Contains(t, []string{"ready", "degraded"}, response["status"])
+	assert.NotNil(t, response["version"])
+	assert.NotNil(t, response["uptime_seconds"])
+	assert.NotNil(t, response["active_sessions"])
+	assert.NotNil(t, response["db_status"])
+	assert.NotNil(t, response["embedding_status"])
+	assert.NotNil(t, response["ready"])
 }
 
 // TestHandleSessionInit_ValidRequest tests session init with valid request.

internal/worker/handlers_update.go

@@ -4,6 +4,7 @@ package worker
 import (
 	"fmt"
 	"net/http"
+	"strings"
 	"time"
 
 	"github.com/rs/zerolog/log"
@@ -13,7 +14,14 @@ import (
 func (s *Service) handleUpdateCheck(w http.ResponseWriter, r *http.Request) {
 	info, err := s.updater.CheckForUpdate(r.Context())
 	if err != nil {
-		http.Error(w, err.Error(), http.StatusInternalServerError)
+		// Return a proper JSON response for errors instead of 500
+		// This allows the frontend to handle it gracefully
+		writeJSON(w, map[string]any{
+			"available":       false,
+			"current_version": s.version,
+			"error":           err.Error(),
+			"rate_limited":    strings.Contains(err.Error(), "403"),
+		})
 		return
 	}
 	writeJSON(w, info)

internal/worker/sdk/dedup.go

@@ -0,0 +1,161 @@
+// Package sdk provides write-time observation deduplication via vector similarity.
+package sdk
+
+import (
+	"context"
+	"fmt"
+	"strings"
+
+	"github.com/lukaszraczylo/claude-mnemonic/internal/config"
+	"github.com/lukaszraczylo/claude-mnemonic/internal/db/gorm"
+	"github.com/lukaszraczylo/claude-mnemonic/internal/vector/sqlitevec"
+	"github.com/lukaszraczylo/claude-mnemonic/pkg/models"
+	"github.com/rs/zerolog/log"
+)
+
+// DeduplicationResult represents the outcome of a vector similarity dedup check.
+type DeduplicationResult struct {
+	ExistingID int64
+	Similarity float64
+	Action     string // "insert", "merge"
+}
+
+// checkVectorDeduplication checks if a similar observation already exists using vector similarity.
+// Returns a result indicating whether to insert or merge, or an error.
+// On any failure, returns Action="insert" so the caller always proceeds with storage.
+func (p *Processor) checkVectorDeduplication(ctx context.Context, obs *models.ParsedObservation, project string) *DeduplicationResult {
+	cfg := config.Get()
+	if !cfg.DeduplicationEnabled {
+		return &DeduplicationResult{Action: "insert"}
+	}
+
+	if p.vectorClient == nil {
+		return &DeduplicationResult{Action: "insert"}
+	}
+
+	// Build search text from observation fields
+	searchText := buildObservationSearchText(obs)
+	if searchText == "" {
+		return &DeduplicationResult{Action: "insert"}
+	}
+
+	// Query vector DB for similar observations in the same project
+	where := sqlitevec.BuildWhereFilter(sqlitevec.DocTypeObservation, project)
+	results, err := p.vectorClient.Query(ctx, searchText, 3, where)
+	if err != nil {
+		log.Debug().Err(err).Msg("Vector search failed during dedup check")
+		return &DeduplicationResult{Action: "insert"}
+	}
+
+	// Check results for high similarity
+	for _, r := range results {
+		if r.Similarity >= cfg.DeduplicationThreshold {
+			obsID := extractObservationIDFromVectorDoc(r)
+			if obsID > 0 {
+				return &DeduplicationResult{
+					ExistingID: obsID,
+					Similarity: r.Similarity,
+					Action:     "merge",
+				}
+			}
+		}
+	}
+
+	return &DeduplicationResult{Action: "insert"}
+}
+
+// buildObservationSearchText creates searchable text from a parsed observation.
+func buildObservationSearchText(obs *models.ParsedObservation) string {
+	var parts []string
+	if obs.Title != "" {
+		parts = append(parts, obs.Title)
+	}
+	if obs.Subtitle != "" {
+		parts = append(parts, obs.Subtitle)
+	}
+	if obs.Narrative != "" {
+		parts = append(parts, obs.Narrative)
+	}
+	text := strings.Join(parts, " ")
+	if len(text) > 2000 {
+		text = text[:2000]
+	}
+	return text
+}
+
+// extractObservationIDFromVectorDoc extracts the SQLite observation ID from a vector query result.
+func extractObservationIDFromVectorDoc(r sqlitevec.QueryResult) int64 {
+	// Prefer the sqlite_id metadata field (set during vector sync)
+	if sqliteID, ok := r.Metadata["sqlite_id"].(float64); ok && sqliteID > 0 {
+		return int64(sqliteID)
+	}
+	if sqliteID, ok := r.Metadata["sqlite_id"].(int64); ok && sqliteID > 0 {
+		return sqliteID
+	}
+
+	// Fallback: parse from doc_id format "obs_{id}_composite" or "obs_{id}_narrative"
+	if !strings.HasPrefix(r.ID, "obs_") {
+		return 0
+	}
+	parts := strings.SplitN(r.ID[4:], "_", 2)
+	if len(parts) == 0 {
+		return 0
+	}
+	var id int64
+	fmt.Sscanf(parts[0], "%d", &id)
+	return id
+}
+
+// mergeObservation updates an existing observation with new information from a duplicate.
+// It appends new facts, updates the narrative if the new one is longer,
+// and bumps the importance score to reflect reconfirmation.
+func (p *Processor) mergeObservation(ctx context.Context, existingID int64, newObs *models.ParsedObservation) error {
+	existing, err := p.observationStore.GetObservationByID(ctx, existingID)
+	if err != nil {
+		return fmt.Errorf("fetch existing observation %d: %w", existingID, err)
+	}
+	if existing == nil {
+		return fmt.Errorf("observation %d not found", existingID)
+	}
+
+	update := &gorm.ObservationUpdate{}
+	changed := false
+
+	// Merge facts: append new facts not already present
+	if len(newObs.Facts) > 0 {
+		existingFactSet := make(map[string]struct{}, len(existing.Facts))
+		for _, f := range existing.Facts {
+			existingFactSet[f] = struct{}{}
+		}
+		mergedFacts := make([]string, len(existing.Facts))
+		copy(mergedFacts, existing.Facts)
+		for _, f := range newObs.Facts {
+			if _, exists := existingFactSet[f]; !exists {
+				mergedFacts = append(mergedFacts, f)
+				changed = true
+			}
+		}
+		if changed {
+			update.Facts = &mergedFacts
+		}
+	}
+
+	// Update narrative if the new one is longer/more detailed
+	if len(newObs.Narrative) > len(existing.Narrative.String) {
+		update.Narrative = &newObs.Narrative
+		changed = true
+	}
+
+	if !changed {
+		// Nothing new to merge, but still count it as a confirmed observation
+		log.Debug().Int64("id", existingID).Msg("Dedup merge: no new content, skipping update")
+		return nil
+	}
+
+	_, err = p.observationStore.UpdateObservation(ctx, existingID, update)
+	if err != nil {
+		return fmt.Errorf("update observation %d: %w", existingID, err)
+	}
+
+	return nil
+}

internal/worker/sdk/dedup_test.go

@@ -0,0 +1,143 @@
+package sdk
+
+import (
+	"testing"
+
+	"github.com/lukaszraczylo/claude-mnemonic/internal/vector/sqlitevec"
+	"github.com/lukaszraczylo/claude-mnemonic/pkg/models"
+)
+
+func TestBuildObservationSearchText(t *testing.T) {
+	tests := []struct {
+		name     string
+		obs      *models.ParsedObservation
+		expected string
+	}{
+		{
+			name:     "empty observation",
+			obs:      &models.ParsedObservation{},
+			expected: "",
+		},
+		{
+			name: "title only",
+			obs: &models.ParsedObservation{
+				Title: "Fix database connection",
+			},
+			expected: "Fix database connection",
+		},
+		{
+			name: "all fields",
+			obs: &models.ParsedObservation{
+				Title:     "Fix database connection",
+				Subtitle:  "Connection pooling issue",
+				Narrative: "The database connection pool was exhausted due to leaked connections.",
+			},
+			expected: "Fix database connection Connection pooling issue The database connection pool was exhausted due to leaked connections.",
+		},
+		{
+			name: "truncates long text",
+			obs: &models.ParsedObservation{
+				Narrative: string(make([]byte, 3000)),
+			},
+			expected: string(make([]byte, 2000)),
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			result := buildObservationSearchText(tt.obs)
+			if result != tt.expected {
+				t.Errorf("got %q, want %q", result, tt.expected)
+			}
+		})
+	}
+}
+
+func TestExtractObservationIDFromVectorDoc(t *testing.T) {
+	tests := []struct {
+		name     string
+		result   sqlitevec.QueryResult
+		expected int64
+	}{
+		{
+			name: "from sqlite_id metadata (float64)",
+			result: sqlitevec.QueryResult{
+				ID:       "obs_42_narrative",
+				Metadata: map[string]any{"sqlite_id": float64(42)},
+			},
+			expected: 42,
+		},
+		{
+			name: "from sqlite_id metadata (int64)",
+			result: sqlitevec.QueryResult{
+				ID:       "obs_42_narrative",
+				Metadata: map[string]any{"sqlite_id": int64(42)},
+			},
+			expected: 42,
+		},
+		{
+			name: "fallback to doc_id parsing",
+			result: sqlitevec.QueryResult{
+				ID:       "obs_99_composite",
+				Metadata: map[string]any{},
+			},
+			expected: 99,
+		},
+		{
+			name: "non-observation doc_id",
+			result: sqlitevec.QueryResult{
+				ID:       "summary_5_text",
+				Metadata: map[string]any{},
+			},
+			expected: 0,
+		},
+		{
+			name: "zero sqlite_id falls back to doc_id",
+			result: sqlitevec.QueryResult{
+				ID:       "obs_123_narrative",
+				Metadata: map[string]any{"sqlite_id": float64(0)},
+			},
+			expected: 123,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			result := extractObservationIDFromVectorDoc(tt.result)
+			if result != tt.expected {
+				t.Errorf("got %d, want %d", result, tt.expected)
+			}
+		})
+	}
+}
+
+func TestCheckVectorDeduplication_NilClient(t *testing.T) {
+	p := &Processor{
+		// No vectorClient set
+	}
+
+	obs := &models.ParsedObservation{
+		Title:     "Test observation",
+		Narrative: "Some narrative text",
+	}
+
+	result := p.checkVectorDeduplication(nil, obs, "test-project")
+	if result.Action != "insert" {
+		t.Errorf("expected Action='insert' when vectorClient is nil, got %q", result.Action)
+	}
+}
+
+func TestCheckVectorDeduplication_EmptySearchText(t *testing.T) {
+	p := &Processor{
+		// vectorClient would be set but obs is empty
+	}
+
+	obs := &models.ParsedObservation{
+		// All empty fields
+	}
+
+	result := p.checkVectorDeduplication(nil, obs, "test-project")
+	if result.Action != "insert" {
+		t.Errorf("expected Action='insert' for empty observation, got %q", result.Action)
+	}
+}

internal/worker/sdk/processor.go

@@ -19,6 +19,7 @@ import (
 
 	"github.com/lukaszraczylo/claude-mnemonic/internal/config"
 	"github.com/lukaszraczylo/claude-mnemonic/internal/db/gorm"
+	"github.com/lukaszraczylo/claude-mnemonic/internal/vector/sqlitevec"
 	"github.com/lukaszraczylo/claude-mnemonic/pkg/models"
 	"github.com/lukaszraczylo/claude-mnemonic/pkg/similarity"
 	"github.com/rs/zerolog/log"
@@ -194,6 +195,36 @@ func hashRequest(toolName, input, output string) string {
 	return hex.EncodeToString(h.Sum(nil))[:16]       // Short hash is sufficient
 }
 
+// maxStdoutBytes is the maximum number of bytes to capture from CLI stdout.
+const maxStdoutBytes = 1 * 1024 * 1024 // 1 MiB
+
+// maxStderrBytes is the maximum number of bytes to capture from CLI stderr.
+const maxStderrBytes = 64 * 1024 // 64 KiB
+
+// limitedWriter wraps a bytes.Buffer and silently discards writes beyond a maximum size.
+type limitedWriter struct {
+	buf bytes.Buffer
+	max int
+}
+
+// Write implements io.Writer. It writes up to the remaining capacity and silently discards the rest.
+func (lw *limitedWriter) Write(p []byte) (int, error) {
+	remaining := lw.max - lw.buf.Len()
+	if remaining <= 0 {
+		return len(p), nil // Silently discard
+	}
+	if len(p) > remaining {
+		p = p[:remaining]
+	}
+	lw.buf.Write(p)
+	return len(p), nil
+}
+
+// String returns the buffered content as a string.
+func (lw *limitedWriter) String() string {
+	return lw.buf.String()
+}
+
 // BroadcastFunc is a callback for broadcasting events to SSE clients.
 type BroadcastFunc func(event map[string]any)
 
@@ -212,6 +243,7 @@ const MaxVectorSyncWorkers = 8
 type Processor struct {
 	observationStore    *gorm.ObservationStore
 	summaryStore        *gorm.SummaryStore
+	vectorClient        *sqlitevec.Client
 	broadcastFunc       BroadcastFunc
 	syncObservationFunc SyncObservationFunc
 	syncSummaryFunc     SyncSummaryFunc
@@ -240,6 +272,11 @@ func (p *Processor) SetSyncSummaryFunc(fn SyncSummaryFunc) {
 	p.syncSummaryFunc = fn
 }
 
+// SetVectorClient sets the vector client for write-time deduplication.
+func (p *Processor) SetVectorClient(client *sqlitevec.Client) {
+	p.vectorClient = client
+}
+
 // broadcast sends an event via the broadcast callback if set.
 func (p *Processor) broadcast(event map[string]any) {
 	if p.broadcastFunc != nil {
@@ -429,16 +466,34 @@ func (p *Processor) ProcessObservation(ctx context.Context, sdkSessionID, projec
 		// Convert to stored observation for similarity check
 		storedObs := obs.ToStoredObservation()
 
-		// Check if this observation is too similar to existing ones
+		// Check if this observation is too similar to existing ones (text-based Jaccard)
 		if existingObs != nil && similarity.IsSimilarToAny(storedObs, existingObs, similarityThreshold) {
 			log.Debug().
 				Str("type", string(obs.Type)).
 				Str("title", obs.Title).
-				Msg("Skipping observation - too similar to existing")
+				Msg("Skipping observation - too similar to existing (text)")
 			skippedCount++
 			continue
 		}
 
+		// Check vector similarity for high-confidence dedup with merge
+		dedupResult := p.checkVectorDeduplication(ctx, obs, project)
+		if dedupResult.Action == "merge" {
+			log.Info().
+				Int64("existing_id", dedupResult.ExistingID).
+				Float64("similarity", dedupResult.Similarity).
+				Str("title", obs.Title).
+				Msg("Merging duplicate observation (vector dedup)")
+			if err := p.mergeObservation(ctx, dedupResult.ExistingID, obs); err != nil {
+				log.Warn().Err(err).Int64("existing_id", dedupResult.ExistingID).
+					Msg("Merge failed, inserting as new observation")
+				// Fall through to normal insert
+			} else {
+				skippedCount++
+				continue
+			}
+		}
+
 		id, createdAtEpoch, err := p.observationStore.StoreObservation(ctx, sdkSessionID, project, obs, promptNumber, 0)
 		if err != nil {
 			log.Error().Err(err).Msg("Failed to store observation")
@@ -641,13 +696,19 @@ func (p *Processor) callClaudeCLI(ctx context.Context, prompt string) (string, e
 	// (hooks are triggered based on working directory)
 	cmd.Dir = "/tmp"
 
-	// Disable any plugin hooks by setting an env var that our hooks can check
+	// Disable any plugin hooks by setting an env var that our hooks can check.
+	// NOTE: This env var is NOT reliably propagated to hook subprocesses because
+	// Claude Code constructs its own environment for hooks. The server-side check
+	// in handleSessionInit (handlers_sessions.go) provides the reliable guard by
+	// detecting the system prompt signature in the prompt text. This env var is
+	// kept as a best-effort defense-in-depth measure for cases where it does work.
 	cmd.Env = append(os.Environ(), "CLAUDE_MNEMONIC_INTERNAL=1")
 
-	// Capture output
-	var stdout, stderr bytes.Buffer
-	cmd.Stdout = &stdout
-	cmd.Stderr = &stderr
+	// Capture output with size limits to prevent unbounded memory usage
+	stdout := &limitedWriter{max: maxStdoutBytes}
+	stderr := &limitedWriter{max: maxStderrBytes}
+	cmd.Stdout = stdout
+	cmd.Stderr = stderr
 
 	// Run command
 	err := cmd.Run()

internal/worker/sdk/processor_test.go

@@ -1,9 +1,12 @@
 package sdk
 
 import (
+	"context"
 	"os"
 	"path/filepath"
+	"sync"
 	"testing"
+	"time"
 
 	"github.com/lukaszraczylo/claude-mnemonic/pkg/models"
 	"github.com/stretchr/testify/assert"
@@ -1178,3 +1181,663 @@ func TestSafeResolvePath(t *testing.T) {
 		})
 	}
 }
+
+// =============================================================================
+// TESTS FOR CircuitBreaker
+// =============================================================================
+
+func TestNewCircuitBreaker(t *testing.T) {
+	cb := NewCircuitBreaker(5, 60)
+
+	assert.NotNil(t, cb)
+	assert.Equal(t, int64(5), cb.threshold)
+	assert.Equal(t, int64(60), cb.resetTimeout)
+	assert.Equal(t, "closed", cb.State())
+}
+
+func TestCircuitBreaker_Allow_Closed(t *testing.T) {
+	cb := NewCircuitBreaker(5, 60)
+
+	// Closed state should allow requests
+	assert.True(t, cb.Allow())
+	assert.True(t, cb.Allow())
+}
+
+func TestCircuitBreaker_Allow_Open(t *testing.T) {
+	cb := NewCircuitBreaker(2, 60) // Low threshold for testing
+
+	// Record enough failures to open the circuit
+	cb.RecordFailure()
+	cb.RecordFailure()
+
+	// Open state should block requests
+	assert.False(t, cb.Allow())
+	assert.Equal(t, "open", cb.State())
+}
+
+func TestCircuitBreaker_RecordSuccess(t *testing.T) {
+	cb := NewCircuitBreaker(2, 60)
+
+	// Record a failure
+	cb.RecordFailure()
+	assert.Equal(t, int64(1), cb.Metrics().Failures)
+
+	// Record success resets failures
+	cb.RecordSuccess()
+	assert.Equal(t, int64(0), cb.Metrics().Failures)
+	assert.Equal(t, "closed", cb.State())
+}
+
+func TestCircuitBreaker_RecordFailure_OpensCircuit(t *testing.T) {
+	cb := NewCircuitBreaker(3, 60)
+
+	// Record failures below threshold
+	cb.RecordFailure()
+	assert.Equal(t, "closed", cb.State())
+
+	cb.RecordFailure()
+	assert.Equal(t, "closed", cb.State())
+
+	// Third failure should open circuit
+	cb.RecordFailure()
+	assert.Equal(t, "open", cb.State())
+}
+
+func TestCircuitBreaker_State(t *testing.T) {
+	cb := NewCircuitBreaker(1, 60)
+
+	// Initially closed
+	assert.Equal(t, "closed", cb.State())
+
+	// After failure, open
+	cb.RecordFailure()
+	assert.Equal(t, "open", cb.State())
+
+	// After success, closed
+	cb.RecordSuccess()
+	assert.Equal(t, "closed", cb.State())
+}
+
+func TestCircuitBreaker_Metrics(t *testing.T) {
+	cb := NewCircuitBreaker(5, 120)
+
+	metrics := cb.Metrics()
+	assert.Equal(t, "closed", metrics.State)
+	assert.Equal(t, int64(0), metrics.Failures)
+	assert.Equal(t, int64(5), metrics.Threshold)
+	assert.Equal(t, int64(120), metrics.ResetTimeoutSecs)
+	assert.Equal(t, int64(0), metrics.LastFailureUnix)
+
+	// After failure
+	cb.RecordFailure()
+	metrics = cb.Metrics()
+	assert.Equal(t, int64(1), metrics.Failures)
+	assert.Greater(t, metrics.LastFailureUnix, int64(0))
+}
+
+func TestCircuitBreaker_Metrics_OpenWithReset(t *testing.T) {
+	cb := NewCircuitBreaker(1, 60)
+
+	cb.RecordFailure()
+	assert.Equal(t, "open", cb.State())
+
+	metrics := cb.Metrics()
+	assert.Equal(t, "open", metrics.State)
+	assert.Greater(t, metrics.SecondsUntilReset, int64(0))
+	assert.LessOrEqual(t, metrics.SecondsUntilReset, int64(60))
+}
+
+// =============================================================================
+// TESTS FOR RequestDeduplicator
+// =============================================================================
+
+func TestNewRequestDeduplicator(t *testing.T) {
+	d := NewRequestDeduplicator(300, 1000)
+
+	assert.NotNil(t, d)
+	assert.NotNil(t, d.seen)
+	assert.Equal(t, int64(300), d.ttlSecs)
+	assert.Equal(t, 1000, d.maxSize)
+}
+
+func TestRequestDeduplicator_IsDuplicate_NotSeen(t *testing.T) {
+	d := NewRequestDeduplicator(300, 1000)
+
+	// New hash is not a duplicate
+	assert.False(t, d.IsDuplicate("newhash"))
+}
+
+func TestRequestDeduplicator_IsDuplicate_AfterRecord(t *testing.T) {
+	d := NewRequestDeduplicator(300, 1000)
+
+	hash := "testhash"
+
+	// Record the hash
+	d.Record(hash)
+
+	// Now it should be a duplicate
+	assert.True(t, d.IsDuplicate(hash))
+}
+
+func TestRequestDeduplicator_Record(t *testing.T) {
+	d := NewRequestDeduplicator(300, 1000)
+
+	hash := "recordtest"
+	d.Record(hash)
+
+	// Check it was recorded
+	d.mu.RLock()
+	_, exists := d.seen[hash]
+	d.mu.RUnlock()
+
+	assert.True(t, exists)
+}
+
+func TestRequestDeduplicator_Record_Eviction(t *testing.T) {
+	// Small maxSize for testing eviction
+	d := NewRequestDeduplicator(0, 2) // TTL of 0 means everything is "old"
+
+	// Record until capacity
+	d.Record("hash1")
+	d.Record("hash2")
+
+	// Recording a third should trigger eviction (since TTL is 0)
+	d.Record("hash3")
+
+	// Should have cleaned up old entries
+	d.mu.RLock()
+	size := len(d.seen)
+	d.mu.RUnlock()
+
+	// Size should be limited (eviction occurred)
+	assert.LessOrEqual(t, size, 3)
+}
+
+func TestHashRequest(t *testing.T) {
+	tests := []struct {
+		name        string
+		toolName    string
+		input       string
+		output      string
+		compareWith []string
+		wantLen     int
+		wantSame    bool
+	}{
+		{
+			name:     "basic hash",
+			toolName: "Read",
+			input:    "file.txt",
+			output:   "content",
+			wantLen:  16,
+		},
+		{
+			name:     "consistent hashing",
+			toolName: "Edit",
+			input:    "same input",
+			output:   "same output",
+			wantLen:  16,
+		},
+		{
+			name:     "long output truncation",
+			toolName: "Bash",
+			input:    "command",
+			output:   string(make([]byte, 5000)), // Very long output
+			wantLen:  16,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			hash := hashRequest(tt.toolName, tt.input, tt.output)
+			assert.Len(t, hash, tt.wantLen)
+
+			// Same inputs should produce same hash
+			hash2 := hashRequest(tt.toolName, tt.input, tt.output)
+			assert.Equal(t, hash, hash2)
+		})
+	}
+}
+
+func TestHashRequest_DifferentInputs(t *testing.T) {
+	// Different inputs should produce different hashes
+	hash1 := hashRequest("Read", "file1.txt", "content1")
+	hash2 := hashRequest("Read", "file2.txt", "content2")
+
+	assert.NotEqual(t, hash1, hash2)
+}
+
+func TestHashRequest_OutputTruncation(t *testing.T) {
+	// Hash should be the same for outputs that differ only after 1000 chars
+	longOutput1 := string(make([]byte, 1500))
+	longOutput2 := longOutput1[:1000] + "different suffix here"
+
+	hash1 := hashRequest("Read", "input", longOutput1)
+	hash2 := hashRequest("Read", "input", longOutput2)
+
+	// Since we only hash first 1000 chars, these should be the same
+	assert.Equal(t, hash1, hash2)
+}
+
+// =============================================================================
+// TESTS FOR Processor methods
+// =============================================================================
+
+func TestProcessor_CircuitBreakerState(t *testing.T) {
+	p := &Processor{
+		circuitBreaker: NewCircuitBreaker(2, 60),
+	}
+
+	// Initially closed
+	assert.Equal(t, "closed", p.CircuitBreakerState())
+
+	// After enough failures, open
+	p.circuitBreaker.RecordFailure()
+	p.circuitBreaker.RecordFailure()
+	assert.Equal(t, "open", p.CircuitBreakerState())
+
+	// After success, closed
+	p.circuitBreaker.RecordSuccess()
+	assert.Equal(t, "closed", p.CircuitBreakerState())
+}
+
+func TestProcessor_CircuitBreakerMetrics(t *testing.T) {
+	p := &Processor{
+		circuitBreaker: NewCircuitBreaker(5, 120),
+	}
+
+	metrics := p.CircuitBreakerMetrics()
+	assert.Equal(t, "closed", metrics.State)
+	assert.Equal(t, int64(0), metrics.Failures)
+	assert.Equal(t, int64(5), metrics.Threshold)
+	assert.Equal(t, int64(120), metrics.ResetTimeoutSecs)
+
+	// Record a failure and check metrics update
+	p.circuitBreaker.RecordFailure()
+	metrics = p.CircuitBreakerMetrics()
+	assert.Equal(t, int64(1), metrics.Failures)
+	assert.Greater(t, metrics.LastFailureUnix, int64(0))
+}
+
+// =============================================================================
+// TESTS FOR Vector Sync Workers
+// =============================================================================
+
+func TestProcessor_StartAndStopVectorSyncWorkers(t *testing.T) {
+	var syncedObservations []*models.Observation
+	var mu sync.Mutex
+
+	p := &Processor{
+		vectorSyncChan: make(chan *models.Observation, MaxVectorSyncWorkers*2),
+		vectorSyncDone: make(chan struct{}),
+		syncObservationFunc: func(obs *models.Observation) {
+			mu.Lock()
+			syncedObservations = append(syncedObservations, obs)
+			mu.Unlock()
+		},
+	}
+
+	// Start workers
+	p.StartVectorSyncWorkers()
+
+	// Send some observations
+	obs1 := &models.Observation{SDKSessionID: "test1"}
+	obs2 := &models.Observation{SDKSessionID: "test2"}
+	p.vectorSyncChan <- obs1
+	p.vectorSyncChan <- obs2
+
+	// Give workers time to process
+	time.Sleep(50 * time.Millisecond)
+
+	// Stop workers
+	p.StopVectorSyncWorkers()
+
+	// Verify observations were synced
+	mu.Lock()
+	assert.Len(t, syncedObservations, 2)
+	mu.Unlock()
+}
+
+func TestProcessor_VectorSyncWorker_DrainOnShutdown(t *testing.T) {
+	var syncedCount int
+	var mu sync.Mutex
+
+	p := &Processor{
+		vectorSyncChan: make(chan *models.Observation, 10),
+		vectorSyncDone: make(chan struct{}),
+		syncObservationFunc: func(obs *models.Observation) {
+			mu.Lock()
+			syncedCount++
+			mu.Unlock()
+		},
+	}
+
+	// Queue observations before starting workers
+	for i := 0; i < 5; i++ {
+		p.vectorSyncChan <- &models.Observation{SDKSessionID: "pre-queued"}
+	}
+
+	// Start workers
+	p.StartVectorSyncWorkers()
+
+	// Stop immediately - workers should drain the queue
+	p.StopVectorSyncWorkers()
+
+	// All pre-queued items should have been processed
+	mu.Lock()
+	assert.Equal(t, 5, syncedCount)
+	mu.Unlock()
+}
+
+func TestProcessor_VectorSyncWorker_NilSyncFunc(t *testing.T) {
+	p := &Processor{
+		vectorSyncChan:      make(chan *models.Observation, 10),
+		vectorSyncDone:      make(chan struct{}),
+		syncObservationFunc: nil, // No sync function set
+	}
+
+	// Start workers
+	p.StartVectorSyncWorkers()
+
+	// Send observation - should not panic even with nil sync func
+	p.vectorSyncChan <- &models.Observation{SDKSessionID: "test"}
+
+	// Give it time to process
+	time.Sleep(50 * time.Millisecond)
+
+	// Stop workers - should not panic
+	p.StopVectorSyncWorkers()
+}
+
+// =============================================================================
+// TESTS FOR CircuitBreaker Additional Behaviors
+// =============================================================================
+
+func TestCircuitBreaker_Allow_OpenBlocksRequests(t *testing.T) {
+	cb := NewCircuitBreaker(1, 60)
+
+	// Open the circuit
+	cb.RecordFailure()
+	assert.Equal(t, "open", cb.State())
+
+	// All requests should be blocked
+	assert.False(t, cb.Allow())
+	assert.False(t, cb.Allow())
+	assert.False(t, cb.Allow())
+}
+
+func TestCircuitBreaker_MultipleFailures(t *testing.T) {
+	cb := NewCircuitBreaker(3, 60) // Higher threshold
+
+	// Record failures below threshold
+	cb.RecordFailure()
+	assert.Equal(t, "closed", cb.State())
+	assert.Equal(t, int64(1), cb.Metrics().Failures)
+
+	cb.RecordFailure()
+	assert.Equal(t, "closed", cb.State())
+	assert.Equal(t, int64(2), cb.Metrics().Failures)
+
+	// Third failure opens circuit
+	cb.RecordFailure()
+	assert.Equal(t, "open", cb.State())
+	assert.Equal(t, int64(3), cb.Metrics().Failures)
+}
+
+func TestCircuitBreaker_SuccessResetsFailures(t *testing.T) {
+	cb := NewCircuitBreaker(5, 60)
+
+	// Record some failures
+	cb.RecordFailure()
+	cb.RecordFailure()
+	assert.Equal(t, int64(2), cb.Metrics().Failures)
+
+	// Success resets failures
+	cb.RecordSuccess()
+	assert.Equal(t, int64(0), cb.Metrics().Failures)
+	assert.Equal(t, "closed", cb.State())
+}
+
+func TestCircuitBreaker_Metrics_Comprehensive(t *testing.T) {
+	cb := NewCircuitBreaker(5, 120)
+
+	// Initial state
+	metrics := cb.Metrics()
+	assert.Equal(t, "closed", metrics.State)
+	assert.Equal(t, int64(0), metrics.Failures)
+	assert.Equal(t, int64(5), metrics.Threshold)
+	assert.Equal(t, int64(120), metrics.ResetTimeoutSecs)
+	assert.Equal(t, int64(0), metrics.LastFailureUnix)
+	assert.Equal(t, int64(0), metrics.SecondsUntilReset)
+
+	// After failures that open circuit
+	for i := 0; i < 5; i++ {
+		cb.RecordFailure()
+	}
+	metrics = cb.Metrics()
+	assert.Equal(t, "open", metrics.State)
+	assert.Equal(t, int64(5), metrics.Failures)
+	assert.Greater(t, metrics.LastFailureUnix, int64(0))
+	assert.Greater(t, metrics.SecondsUntilReset, int64(0))
+}
+
+// =============================================================================
+// TESTS FOR MaxVectorSyncWorkers constant
+// =============================================================================
+
+func TestMaxVectorSyncWorkers(t *testing.T) {
+	assert.Equal(t, 8, MaxVectorSyncWorkers)
+}
+
+// =============================================================================
+// ADDITIONAL EDGE CASE TESTS
+// =============================================================================
+
+func TestRequestDeduplicator_IsDuplicate_ExpiredEntry(t *testing.T) {
+	if testing.Short() {
+		t.Skip("Skipping time-dependent test in short mode")
+	}
+	// Use a 1-second TTL with enough margin
+	d := NewRequestDeduplicator(1, 100)
+
+	hash := "expiretest"
+	d.Record(hash)
+
+	// Initially duplicate
+	assert.True(t, d.IsDuplicate(hash))
+
+	// Wait for TTL to expire (2.5 seconds to ensure crossing second boundaries)
+	time.Sleep(2500 * time.Millisecond)
+
+	// Should no longer be considered duplicate
+	assert.False(t, d.IsDuplicate(hash))
+}
+
+// =============================================================================
+// TESTS FOR ProcessObservation Early Returns
+// =============================================================================
+
+func TestProcessObservation_SkipTool(t *testing.T) {
+	p := &Processor{
+		circuitBreaker: NewCircuitBreaker(5, 60),
+		deduplicator:   NewRequestDeduplicator(300, 1000),
+	}
+
+	ctx := context.Background()
+
+	// TodoWrite should be skipped
+	err := p.ProcessObservation(ctx, "session-1", "project-1", "TodoWrite",
+		map[string]string{"content": "test"}, "success", 1, "/test/cwd")
+	assert.NoError(t, err)
+
+	// Glob should be skipped
+	err = p.ProcessObservation(ctx, "session-1", "project-1", "Glob",
+		map[string]string{"pattern": "*.go"}, []string{"main.go", "test.go"}, 1, "/test/cwd")
+	assert.NoError(t, err)
+
+	// AskUserQuestion should be skipped
+	err = p.ProcessObservation(ctx, "session-1", "project-1", "AskUserQuestion",
+		"question", "answer", 1, "/test/cwd")
+	assert.NoError(t, err)
+}
+
+func TestProcessObservation_SkipTrivial(t *testing.T) {
+	p := &Processor{
+		circuitBreaker: NewCircuitBreaker(5, 60),
+		deduplicator:   NewRequestDeduplicator(300, 1000),
+	}
+
+	ctx := context.Background()
+
+	// Short output should be skipped
+	err := p.ProcessObservation(ctx, "session-1", "project-1", "Read",
+		map[string]string{"file_path": "/test.go"}, "short", 1, "/test/cwd")
+	assert.NoError(t, err)
+
+	// "No matches found" should be skipped
+	err = p.ProcessObservation(ctx, "session-1", "project-1", "Grep",
+		map[string]string{"pattern": "test"}, "No matches found in the repository", 1, "/test/cwd")
+	assert.NoError(t, err)
+}
+
+func TestProcessObservation_SkipDuplicate(t *testing.T) {
+	p := &Processor{
+		circuitBreaker: NewCircuitBreaker(5, 60),
+		deduplicator:   NewRequestDeduplicator(300, 1000),
+		sem:            make(chan struct{}, 4),
+		claudePath:     "/nonexistent/path", // Will fail at CLI call stage
+	}
+
+	ctx := context.Background()
+
+	// Valid input that would be processed
+	input := map[string]string{"file_path": "/project/main.go"}
+	output := "package main\n\nimport \"fmt\"\n\nfunc main() {\n\tfmt.Println(\"Hello World\")\n}"
+
+	// First call should try to process (will fail because claudePath doesn't exist)
+	err := p.ProcessObservation(ctx, "session-1", "project-1", "Read", input, output, 1, "/test/cwd")
+	// Expect error because claudePath doesn't exist
+	assert.Error(t, err)
+
+	// Second call with same input should be skipped as duplicate
+	err = p.ProcessObservation(ctx, "session-1", "project-1", "Read", input, output, 1, "/test/cwd")
+	assert.NoError(t, err) // No error because it was skipped as duplicate
+}
+
+func TestProcessObservation_CircuitBreakerOpen(t *testing.T) {
+	cb := NewCircuitBreaker(1, 60) // Threshold of 1
+	cb.RecordFailure()             // Open the circuit breaker
+
+	p := &Processor{
+		circuitBreaker: cb,
+		deduplicator:   NewRequestDeduplicator(300, 1000),
+	}
+
+	ctx := context.Background()
+
+	// Valid input that would be processed
+	input := map[string]string{"file_path": "/project/main.go"}
+	output := "package main\n\nimport \"fmt\"\n\nfunc main() {\n\tfmt.Println(\"Hello World\")\n}"
+
+	err := p.ProcessObservation(ctx, "session-1", "project-1", "Read", input, output, 1, "/test/cwd")
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "circuit breaker open")
+}
+
+func TestProcessObservation_ContextCancel(t *testing.T) {
+	p := &Processor{
+		circuitBreaker: NewCircuitBreaker(5, 60),
+		deduplicator:   NewRequestDeduplicator(300, 1000),
+		sem:            make(chan struct{}, 1), // Small semaphore
+		claudePath:     "/fake/claude",
+	}
+
+	// Fill the semaphore
+	p.sem <- struct{}{}
+
+	// Create a cancelled context
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel() // Cancel immediately
+
+	// Valid input that would be processed
+	input := map[string]string{"file_path": "/project/main.go"}
+	output := "package main\n\nimport \"fmt\"\n\nfunc main() {\n\tfmt.Println(\"Hello World\")\n}"
+
+	err := p.ProcessObservation(ctx, "session-1", "project-1", "Read", input, output, 1, "/test/cwd")
+	assert.Error(t, err)
+	assert.ErrorIs(t, err, context.Canceled)
+}
+
+// =============================================================================
+// TESTS FOR ProcessSummary Early Returns
+// =============================================================================
+
+func TestProcessSummary_SkipEmptyRequest(t *testing.T) {
+	p := &Processor{
+		circuitBreaker: NewCircuitBreaker(5, 60),
+		deduplicator:   NewRequestDeduplicator(300, 1000),
+	}
+
+	ctx := context.Background()
+
+	// Empty request should be skipped (sessionDBID, sdkSessionID, project, userPrompt, lastUserMsg, lastAssistantMsg)
+	err := p.ProcessSummary(ctx, 1, "session-1", "project-1", "", "", "")
+	assert.NoError(t, err)
+}
+
+func TestProcessSummary_CircuitBreakerOpen(t *testing.T) {
+	cb := NewCircuitBreaker(1, 60)
+	cb.RecordFailure() // Open the circuit breaker
+
+	p := &Processor{
+		circuitBreaker: cb,
+		deduplicator:   NewRequestDeduplicator(300, 1000),
+		sem:            make(chan struct{}, 4),
+		claudePath:     "/nonexistent/path",
+	}
+
+	ctx := context.Background()
+
+	// Meaningful assistant message (> 200 chars, contains code discussion)
+	assistantMsg := `I've updated the handler.go file to fix the authentication bug.
+The function validateToken() was not checking token expiry correctly.
+I've added a check for the exp claim and implemented proper error handling.
+The changes have been tested and the build passes successfully.
+Here's the implementation details and code review.`
+
+	// Valid request but circuit breaker is open
+	err := p.ProcessSummary(ctx, 1, "session-1", "project-1",
+		"Implement authentication", "User message", assistantMsg)
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "claude CLI failed")
+}
+
+// =============================================================================
+// TESTS FOR callClaudeCLI Error Paths
+// =============================================================================
+
+func TestCallClaudeCLI_PromptTooLarge(t *testing.T) {
+	p := &Processor{
+		claudePath: "/fake/claude",
+	}
+
+	ctx := context.Background()
+
+	// Create a prompt that exceeds MaxPromptSize
+	largePrompt := string(make([]byte, MaxPromptSize+1))
+
+	_, err := p.callClaudeCLI(ctx, largePrompt)
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "prompt exceeds maximum size")
+}
+
+func TestCallClaudeCLI_BinaryNotFound(t *testing.T) {
+	p := &Processor{
+		claudePath: "/nonexistent/path/to/claude",
+	}
+
+	ctx := context.Background()
+
+	_, err := p.callClaudeCLI(ctx, "test prompt")
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "claude CLI failed")
+}

internal/worker/sdk/prompts.go

@@ -49,6 +49,10 @@ func BuildObservationPrompt(exec ToolExecution) string {
 	inputJSON, _ := json.MarshalIndent(toolInput, "  ", "  ")
 	outputJSON, _ := json.MarshalIndent(toolOutput, "  ", "  ")
 
+	// Strip system XML artifacts from tool data before building prompt
+	cleanInput := StripSystemXML(string(inputJSON))
+	cleanOutput := StripSystemXML(string(outputJSON))
+
 	timestamp := time.UnixMilli(exec.CreatedAtEpoch).Format(time.RFC3339)
 
 	var sb strings.Builder
@@ -58,8 +62,8 @@ func BuildObservationPrompt(exec ToolExecution) string {
 	if exec.CWD != "" {
 		sb.WriteString(fmt.Sprintf("  <working_directory>%s</working_directory>\n", exec.CWD))
 	}
-	sb.WriteString(fmt.Sprintf("  <parameters>%s</parameters>\n", truncate(string(inputJSON), 3000)))
-	sb.WriteString(fmt.Sprintf("  <outcome>%s</outcome>\n", truncate(string(outputJSON), 5000)))
+	sb.WriteString(fmt.Sprintf("  <parameters>%s</parameters>\n", truncate(cleanInput, 3000)))
+	sb.WriteString(fmt.Sprintf("  <outcome>%s</outcome>\n", truncate(cleanOutput, 5000)))
 	sb.WriteString("</observed_from_primary_session>")
 
 	return sb.String()
@@ -84,8 +88,10 @@ func BuildSummaryPrompt(req SummaryRequest) string {
 	sb.WriteString("Write progress notes of what was done, what was learned, and what's next. This is a checkpoint to capture progress so far. The session is ongoing - you may receive more requests and tool executions after this summary. Write \"next_steps\" as the current trajectory of work (what's actively being worked on or coming up next), not as post-session future work. Always write at least a minimal summary explaining current progress, even if work is still in early stages, so that users see a summary output tied to each request.\n\n")
 
 	if req.LastAssistantMessage != "" {
+		// Strip system XML artifacts from captured transcript content
+		cleanMsg := StripSystemXML(req.LastAssistantMessage)
 		sb.WriteString("Claude's Full Response to User:\n")
-		sb.WriteString(truncate(req.LastAssistantMessage, 4000))
+		sb.WriteString(truncate(cleanMsg, 4000))
 		sb.WriteString("\n\n")
 	}
 

internal/worker/sdk/sanitize.go

@@ -0,0 +1,8 @@
+// Package sdk provides SDK agent integration for claude-mnemonic.
+package sdk
+
+import "github.com/lukaszraczylo/claude-mnemonic/pkg/sanitize"
+
+// StripSystemXML removes known Claude Code internal XML blocks from text.
+// Delegates to the shared sanitize package.
+var StripSystemXML = sanitize.StripSystemXML

internal/worker/service.go

@@ -43,6 +43,13 @@ const (
 	// QueueProcessInterval is how often the background queue processor runs.
 	QueueProcessInterval = 2 * time.Second
 
+	// reinitializationDrainDelay is the delay after marking the service as not ready
+	// to allow in-flight requests to complete before reinitializing.
+	reinitializationDrainDelay = 200 * time.Millisecond
+
+	// MaxConcurrentProcessing limits the number of concurrent session processing goroutines.
+	MaxConcurrentProcessing = 4
+
 	// VectorSyncMaxRetries is the maximum number of retries for vector sync operations.
 	VectorSyncMaxRetries = 3
 
@@ -138,6 +145,7 @@ type Service struct {
 	updater            *update.Updater
 	rateLimiter        *PerClientRateLimiter
 	expensiveOpLimiter *ExpensiveOperationLimiter
+	contextCache       sync.Map
 	version            string
 	recentQueriesBuf   [maxRecentQueries]RecentSearchQuery
 	wg                 sync.WaitGroup
@@ -178,6 +186,13 @@ type staleVerifyRequest struct {
 	observationID int64
 }
 
+// contextCacheEntry caches clustering results for context injection.
+type contextCacheEntry struct {
+	timestamp    time.Time
+	observations []*models.Observation
+	obsCount     int
+}
+
 // RecentSearchQuery tracks a search query for analytics.
 type RecentSearchQuery struct {
 	Timestamp  time.Time `json:"timestamp"`
@@ -288,6 +303,11 @@ func (s *Service) setupVectorSyncCallbacks(
 		})
 	}
 
+	// Set vector client on processor for write-time deduplication
+	if processor != nil && s.vectorClient != nil {
+		processor.SetVectorClient(s.vectorClient)
+	}
+
 	// Set cleanup callback on observation store to sync deletes to vector store
 	if observationStore != nil && vectorSync != nil {
 		observationStore.SetCleanupFunc(func(ctx context.Context, deletedIDs []int64) {
@@ -614,6 +634,7 @@ func (s *Service) startWatchers() {
 func (s *Service) reinitializeDatabase() {
 	// Block new requests
 	s.ready.Store(false)
+	time.Sleep(reinitializationDrainDelay) // Allow in-flight requests to complete
 	log.Info().Msg("Database reinitialization starting...")
 
 	// Get old store references
@@ -1198,6 +1219,8 @@ func (s *Service) setupRoutes() {
 	// Vector management endpoints
 	s.router.Post("/api/vectors/rebuild", s.handleTriggerVectorRebuild)
 	s.router.Get("/api/vectors/health", s.handleVectorHealth)
+	s.router.Get("/api/vector/metrics", s.handleVectorMetrics)
+	s.router.Get("/api/graph/stats", s.handleGraphStats)
 
 	// Readiness check - returns 200 only when fully initialized
 	s.router.Get("/api/ready", s.handleReady)
@@ -1585,12 +1608,13 @@ func (s *Service) processQueue() {
 
 // processAllSessions processes pending messages for all active sessions.
 // Messages are processed in parallel using goroutines, with concurrency
-// limited by the processor's semaphore.
+// limited by a channel-based semaphore.
 func (s *Service) processAllSessions() {
 	// Get all sessions with pending messages
 	sessions := s.sessionManager.GetAllSessions()
 
 	var wg sync.WaitGroup
+	sem := make(chan struct{}, MaxConcurrentProcessing)
 
 	for _, sess := range sessions {
 		// Get pending messages
@@ -1599,11 +1623,13 @@ func (s *Service) processAllSessions() {
 			continue
 		}
 
-		// Process each message in a goroutine
+		// Process each message in a goroutine with semaphore
 		for _, msg := range messages {
 			wg.Add(1)
+			sem <- struct{}{} // Acquire semaphore slot
 			go func(sess *session.ActiveSession, msg session.PendingMessage) {
 				defer wg.Done()
+				defer func() { <-sem }() // Release semaphore slot
 
 				switch msg.Type {
 				case session.MessageTypeObservation:

internal/worker/session/manager.go

@@ -75,6 +75,7 @@ type Manager struct {
 	onDeleted     func(int64)
 	cancel        context.CancelFunc
 	ProcessNotify chan struct{}
+	wg            sync.WaitGroup
 	mu            sync.RWMutex
 }
 
@@ -89,12 +90,14 @@ func NewManager(sessionStore *gorm.SessionStore) *Manager {
 		ProcessNotify: make(chan struct{}, 1),
 	}
 	// Start background cleanup goroutine
+	m.wg.Add(1)
 	go m.cleanupLoop()
 	return m
 }
 
 // cleanupLoop periodically removes stale sessions.
 func (m *Manager) cleanupLoop() {
+	defer m.wg.Done()
 	ticker := time.NewTicker(CleanupInterval)
 	defer ticker.Stop()
 
@@ -350,6 +353,7 @@ func (m *Manager) DeleteSession(sessionDBID int64) {
 func (m *Manager) ShutdownAll(ctx context.Context) {
 	// Stop cleanup goroutine
 	m.cancel()
+	m.wg.Wait()
 
 	m.mu.Lock()
 	sessionIDs := make([]int64, 0, len(m.sessions))

internal/worker/session/manager_test.go

@@ -693,3 +693,735 @@ func TestToolInputResponse(t *testing.T) {
 		})
 	}
 }
+
+// =============================================================================
+// TESTS FOR NewManager AND CLEANUP
+// =============================================================================
+
+// TestNewManager tests the NewManager function.
+func TestNewManager(t *testing.T) {
+	t.Parallel()
+
+	// Test with nil session store (valid for testing)
+	manager := NewManager(nil)
+
+	assert.NotNil(t, manager)
+	assert.NotNil(t, manager.sessions)
+	assert.NotNil(t, manager.ProcessNotify)
+	assert.NotNil(t, manager.ctx)
+	assert.NotNil(t, manager.cancel)
+	assert.Equal(t, 0, manager.GetActiveSessionCount())
+
+	// Clean up - cancel context to stop cleanup goroutine
+	manager.cancel()
+}
+
+// TestNewManager_CleanupGoroutineStops tests that cleanup goroutine stops on cancel.
+func TestNewManager_CleanupGoroutineStops(t *testing.T) {
+	t.Parallel()
+
+	manager := NewManager(nil)
+
+	// Give goroutine time to start
+	time.Sleep(10 * time.Millisecond)
+
+	// Cancel should stop the cleanup goroutine
+	manager.cancel()
+
+	// Context should be done
+	select {
+	case <-manager.ctx.Done():
+		// Expected
+	case <-time.After(100 * time.Millisecond):
+		t.Error("Context should be done after cancel")
+	}
+}
+
+// TestCleanupStaleSessions_NoSessions tests cleanup with no sessions.
+func TestCleanupStaleSessions_NoSessions(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Should not panic with empty sessions
+	manager.cleanupStaleSessions()
+	assert.Equal(t, 0, manager.GetActiveSessionCount())
+}
+
+// TestCleanupStaleSessions_FreshSession tests that fresh sessions are not cleaned.
+func TestCleanupStaleSessions_FreshSession(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Add a fresh session
+	sessionCtx, sessionCancel := context.WithCancel(context.Background())
+	manager.sessions[1] = &ActiveSession{
+		SessionDBID:     1,
+		StartTime:       time.Now(), // Fresh
+		pendingMessages: []PendingMessage{},
+		ctx:             sessionCtx,
+		cancel:          sessionCancel,
+	}
+
+	manager.cleanupStaleSessions()
+
+	// Session should still exist (not stale)
+	assert.Equal(t, 1, manager.GetActiveSessionCount())
+	sessionCancel()
+}
+
+// TestCleanupStaleSessions_StaleSession tests that stale sessions are cleaned.
+func TestCleanupStaleSessions_StaleSession(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Add a stale session
+	sessionCtx, sessionCancel := context.WithCancel(context.Background())
+	manager.sessions[1] = &ActiveSession{
+		SessionDBID:     1,
+		StartTime:       time.Now().Add(-SessionTimeout - time.Minute), // Stale
+		pendingMessages: []PendingMessage{},
+		ctx:             sessionCtx,
+		cancel:          sessionCancel,
+	}
+
+	manager.cleanupStaleSessions()
+
+	// Session should be deleted
+	assert.Equal(t, 0, manager.GetActiveSessionCount())
+}
+
+// TestCleanupStaleSessions_StaleWithPending tests stale sessions with pending messages are not cleaned.
+func TestCleanupStaleSessions_StaleWithPending(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Add a stale session with pending messages
+	sessionCtx, sessionCancel := context.WithCancel(context.Background())
+	defer sessionCancel()
+	manager.sessions[1] = &ActiveSession{
+		SessionDBID:     1,
+		StartTime:       time.Now().Add(-SessionTimeout - time.Minute), // Stale
+		pendingMessages: []PendingMessage{{Type: MessageTypeObservation}},
+		ctx:             sessionCtx,
+		cancel:          sessionCancel,
+	}
+
+	manager.cleanupStaleSessions()
+
+	// Session should NOT be deleted (has pending messages)
+	assert.Equal(t, 1, manager.GetActiveSessionCount())
+}
+
+// TestCleanupStaleSessions_StaleWithActiveGenerator tests stale sessions with active generator are not cleaned.
+func TestCleanupStaleSessions_StaleWithActiveGenerator(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Add a stale session with active generator
+	sessionCtx, sessionCancel := context.WithCancel(context.Background())
+	defer sessionCancel()
+	session := &ActiveSession{
+		SessionDBID:     1,
+		StartTime:       time.Now().Add(-SessionTimeout - time.Minute), // Stale
+		pendingMessages: []PendingMessage{},
+		ctx:             sessionCtx,
+		cancel:          sessionCancel,
+	}
+	session.generatorActive.Store(true)
+	manager.sessions[1] = session
+
+	manager.cleanupStaleSessions()
+
+	// Session should NOT be deleted (generator is active)
+	assert.Equal(t, 1, manager.GetActiveSessionCount())
+}
+
+// TestCleanupStaleSessions_MixedSessions tests cleanup with mixed fresh and stale sessions.
+func TestCleanupStaleSessions_MixedSessions(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Fresh session
+	ctx1, cancel1 := context.WithCancel(context.Background())
+	defer cancel1()
+	manager.sessions[1] = &ActiveSession{
+		SessionDBID:     1,
+		StartTime:       time.Now(),
+		pendingMessages: []PendingMessage{},
+		ctx:             ctx1,
+		cancel:          cancel1,
+	}
+
+	// Stale session (should be deleted)
+	ctx2, cancel2 := context.WithCancel(context.Background())
+	manager.sessions[2] = &ActiveSession{
+		SessionDBID:     2,
+		StartTime:       time.Now().Add(-SessionTimeout - time.Minute),
+		pendingMessages: []PendingMessage{},
+		ctx:             ctx2,
+		cancel:          cancel2,
+	}
+
+	// Stale session with pending (should NOT be deleted)
+	ctx3, cancel3 := context.WithCancel(context.Background())
+	defer cancel3()
+	manager.sessions[3] = &ActiveSession{
+		SessionDBID:     3,
+		StartTime:       time.Now().Add(-SessionTimeout - time.Minute),
+		pendingMessages: []PendingMessage{{Type: MessageTypeObservation}},
+		ctx:             ctx3,
+		cancel:          cancel3,
+	}
+
+	manager.cleanupStaleSessions()
+
+	// Should have 2 sessions left (1 fresh, 1 stale with pending)
+	assert.Equal(t, 2, manager.GetActiveSessionCount())
+
+	// Verify which sessions remain
+	manager.mu.RLock()
+	_, has1 := manager.sessions[1]
+	_, has2 := manager.sessions[2]
+	_, has3 := manager.sessions[3]
+	manager.mu.RUnlock()
+
+	assert.True(t, has1, "Fresh session should remain")
+	assert.False(t, has2, "Stale session should be deleted")
+	assert.True(t, has3, "Stale session with pending should remain")
+}
+
+// TestCleanupLoop_ExitsOnCancel tests that cleanup loop exits when context is cancelled.
+func TestCleanupLoop_ExitsOnCancel(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Start cleanup loop in goroutine
+	done := make(chan struct{})
+	manager.wg.Add(1)
+	go func() {
+		manager.cleanupLoop()
+		close(done)
+	}()
+
+	// Cancel immediately
+	cancel()
+
+	// Should exit quickly
+	select {
+	case <-done:
+		// Success - loop exited
+	case <-time.After(100 * time.Millisecond):
+		t.Error("Cleanup loop should exit when context is cancelled")
+	}
+}
+
+// =============================================================================
+// TESTS FOR InitializeSession (without DB)
+// =============================================================================
+
+// TestInitializeSession_AlreadyActive tests reusing an already active session.
+func TestInitializeSession_AlreadyActive(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Pre-add an active session
+	existingSession := &ActiveSession{
+		SessionDBID:      42,
+		ClaudeSessionID:  "claude-existing",
+		Project:          "test-project",
+		UserPrompt:       "original prompt",
+		LastPromptNumber: 1,
+		StartTime:        time.Now(),
+		pendingMessages:  make([]PendingMessage, 0),
+	}
+	manager.sessions[42] = existingSession
+
+	// Initialize same session - should reuse
+	session, err := manager.InitializeSession(context.Background(), 42, "new prompt", 5)
+
+	assert.NoError(t, err)
+	assert.NotNil(t, session)
+	assert.Same(t, existingSession, session)
+	assert.Equal(t, "new prompt", session.UserPrompt)
+	assert.Equal(t, 5, session.LastPromptNumber)
+}
+
+// TestInitializeSession_AlreadyActive_EmptyPrompt tests reusing session with empty prompt.
+func TestInitializeSession_AlreadyActive_EmptyPrompt(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Pre-add an active session
+	existingSession := &ActiveSession{
+		SessionDBID:      42,
+		UserPrompt:       "original prompt",
+		LastPromptNumber: 1,
+	}
+	manager.sessions[42] = existingSession
+
+	// Initialize with empty prompt - should NOT update
+	session, err := manager.InitializeSession(context.Background(), 42, "", 0)
+
+	assert.NoError(t, err)
+	assert.NotNil(t, session)
+	assert.Equal(t, "original prompt", session.UserPrompt) // Unchanged
+	assert.Equal(t, 1, session.LastPromptNumber)           // Unchanged
+}
+
+// TestInitializeSession_NoStore tests initialization without session store.
+func TestInitializeSession_NoStore(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessionStore:  nil, // No store
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Should fail gracefully with nil store (panic recovery not expected)
+	// This tests the guard against nil sessionStore
+	defer func() {
+		if r := recover(); r != nil {
+			_ = r // Expected panic when calling nil store - intentionally ignored
+		}
+	}()
+
+	_, _ = manager.InitializeSession(context.Background(), 999, "prompt", 1)
+}
+
+// TestInitializeSession_CallbackTriggered tests that created callback is triggered.
+func TestInitializeSession_CallbackTriggered(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	var calledWithID int64
+	manager.SetOnSessionCreated(func(id int64) {
+		calledWithID = id
+	})
+
+	// Add session directly (simulating what would happen after DB fetch)
+	sessionCtx, sessionCancel := context.WithCancel(context.Background())
+	defer sessionCancel()
+	session := &ActiveSession{
+		SessionDBID:     100,
+		ClaudeSessionID: "test",
+		Project:         "project",
+		StartTime:       time.Now(),
+		pendingMessages: make([]PendingMessage, 0),
+		notify:          make(chan struct{}, 1),
+		ctx:             sessionCtx,
+		cancel:          sessionCancel,
+	}
+
+	manager.mu.Lock()
+	manager.sessions[100] = session
+	onCreated := manager.onCreated
+	manager.mu.Unlock()
+
+	// Trigger callback
+	if onCreated != nil {
+		onCreated(100)
+	}
+
+	assert.Equal(t, int64(100), calledWithID)
+}
+
+// =============================================================================
+// TESTS FOR QueueObservation AND QueueSummarize (without DB)
+// =============================================================================
+
+// TestQueueObservation_ToExistingSession tests queuing to an existing session.
+func TestQueueObservation_ToExistingSession(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Pre-add session
+	session := &ActiveSession{
+		SessionDBID:     1,
+		pendingMessages: make([]PendingMessage, 0),
+		notify:          make(chan struct{}, 1),
+	}
+	manager.sessions[1] = session
+
+	// Queue observation
+	err := manager.QueueObservation(context.Background(), 1, ObservationData{
+		ToolName:     "Read",
+		ToolInput:    map[string]string{"path": "/test"},
+		ToolResponse: "content",
+		PromptNumber: 1,
+		CWD:          "/project",
+	})
+
+	assert.NoError(t, err)
+	assert.Equal(t, 1, manager.GetTotalQueueDepth())
+
+	// Verify message
+	messages := manager.DrainMessages(1)
+	assert.Len(t, messages, 1)
+	assert.Equal(t, MessageTypeObservation, messages[0].Type)
+	assert.Equal(t, "Read", messages[0].Observation.ToolName)
+	assert.Equal(t, "/project", messages[0].Observation.CWD)
+}
+
+// TestQueueObservation_NotifiesSession tests that notification is sent to session.
+func TestQueueObservation_NotifiesSession(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Pre-add session with notify channel
+	session := &ActiveSession{
+		SessionDBID:     1,
+		pendingMessages: make([]PendingMessage, 0),
+		notify:          make(chan struct{}, 1),
+	}
+	manager.sessions[1] = session
+
+	// Queue observation
+	err := manager.QueueObservation(context.Background(), 1, ObservationData{ToolName: "Test"})
+	assert.NoError(t, err)
+
+	// Should receive notification on session channel
+	select {
+	case <-session.notify:
+		// Success
+	default:
+		t.Error("Session should receive notification")
+	}
+
+	// Should receive notification on process channel
+	select {
+	case <-manager.ProcessNotify:
+		// Success
+	default:
+		t.Error("Manager ProcessNotify should receive notification")
+	}
+}
+
+// TestQueueSummarize_ToExistingSession tests queuing summarize to an existing session.
+func TestQueueSummarize_ToExistingSession(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Pre-add session
+	session := &ActiveSession{
+		SessionDBID:     1,
+		pendingMessages: make([]PendingMessage, 0),
+		notify:          make(chan struct{}, 1),
+	}
+	manager.sessions[1] = session
+
+	// Queue summarize
+	err := manager.QueueSummarize(context.Background(), 1, "User asked question", "Assistant answered")
+	assert.NoError(t, err)
+	assert.Equal(t, 1, manager.GetTotalQueueDepth())
+
+	// Verify message
+	messages := manager.DrainMessages(1)
+	assert.Len(t, messages, 1)
+	assert.Equal(t, MessageTypeSummarize, messages[0].Type)
+	assert.Equal(t, "User asked question", messages[0].Summarize.LastUserMessage)
+	assert.Equal(t, "Assistant answered", messages[0].Summarize.LastAssistantMessage)
+}
+
+// TestQueueSummarize_NotifiesSession tests that notification is sent to session.
+func TestQueueSummarize_NotifiesSession(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Pre-add session with notify channel
+	session := &ActiveSession{
+		SessionDBID:     1,
+		pendingMessages: make([]PendingMessage, 0),
+		notify:          make(chan struct{}, 1),
+	}
+	manager.sessions[1] = session
+
+	// Queue summarize
+	err := manager.QueueSummarize(context.Background(), 1, "user", "assistant")
+	assert.NoError(t, err)
+
+	// Should receive notification on session channel
+	select {
+	case <-session.notify:
+		// Success
+	default:
+		t.Error("Session should receive notification")
+	}
+
+	// Should receive notification on process channel
+	select {
+	case <-manager.ProcessNotify:
+		// Success
+	default:
+		t.Error("Manager ProcessNotify should receive notification")
+	}
+}
+
+// TestQueueOperations_MultipleMessages tests queuing multiple messages.
+func TestQueueOperations_MultipleMessages(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Pre-add session
+	session := &ActiveSession{
+		SessionDBID:     1,
+		pendingMessages: make([]PendingMessage, 0),
+		notify:          make(chan struct{}, 1),
+	}
+	manager.sessions[1] = session
+
+	// Queue multiple messages
+	for i := 0; i < 10; i++ {
+		if i%2 == 0 {
+			err := manager.QueueObservation(context.Background(), 1, ObservationData{
+				ToolName: "Tool" + string(rune('A'+i)),
+			})
+			assert.NoError(t, err)
+		} else {
+			err := manager.QueueSummarize(context.Background(), 1, "user", "assistant")
+			assert.NoError(t, err)
+		}
+	}
+
+	assert.Equal(t, 10, manager.GetTotalQueueDepth())
+
+	// Drain and verify
+	messages := manager.DrainMessages(1)
+	assert.Len(t, messages, 10)
+}
+
+// TestQueueOperations_NonBlockingNotification tests non-blocking notification behavior.
+func TestQueueOperations_NonBlockingNotification(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Pre-add session with full notify channel
+	session := &ActiveSession{
+		SessionDBID:     1,
+		pendingMessages: make([]PendingMessage, 0),
+		notify:          make(chan struct{}, 1),
+	}
+	// Fill the notify channel
+	session.notify <- struct{}{}
+	manager.sessions[1] = session
+
+	// Fill ProcessNotify channel
+	manager.ProcessNotify <- struct{}{}
+
+	// Queue should NOT block even with full channels
+	done := make(chan bool)
+	go func() {
+		err := manager.QueueObservation(context.Background(), 1, ObservationData{ToolName: "Test"})
+		assert.NoError(t, err)
+		done <- true
+	}()
+
+	select {
+	case <-done:
+		// Success - didn't block
+	case <-time.After(100 * time.Millisecond):
+		t.Error("Queue operation should not block even with full notification channels")
+	}
+}
+
+// TestConcurrentQueueAndCleanup tests concurrent queue operations and cleanup.
+func TestConcurrentQueueAndCleanup(t *testing.T) {
+	t.Parallel()
+
+	manager := &Manager{
+		sessions:      make(map[int64]*ActiveSession),
+		ProcessNotify: make(chan struct{}, 1),
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	manager.ctx = ctx
+	manager.cancel = cancel
+
+	// Pre-add multiple sessions
+	for i := int64(1); i <= 5; i++ {
+		sessionCtx, sessionCancel := context.WithCancel(context.Background())
+		manager.sessions[i] = &ActiveSession{
+			SessionDBID:     i,
+			StartTime:       time.Now(),
+			pendingMessages: make([]PendingMessage, 0),
+			notify:          make(chan struct{}, 1),
+			ctx:             sessionCtx,
+			cancel:          sessionCancel,
+		}
+	}
+
+	var wg sync.WaitGroup
+
+	// Concurrent queue operations
+	for i := 0; i < 50; i++ {
+		wg.Add(1)
+		go func(idx int) {
+			defer wg.Done()
+			sessionID := int64((idx % 5) + 1)
+			if idx%2 == 0 {
+				_ = manager.QueueObservation(context.Background(), sessionID, ObservationData{ToolName: "Test"})
+			} else {
+				_ = manager.QueueSummarize(context.Background(), sessionID, "user", "assistant")
+			}
+		}(i)
+	}
+
+	// Concurrent cleanup
+	for i := 0; i < 10; i++ {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			manager.cleanupStaleSessions()
+		}()
+	}
+
+	// Concurrent reads
+	for i := 0; i < 20; i++ {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			_ = manager.GetActiveSessionCount()
+			_ = manager.GetTotalQueueDepth()
+			_ = manager.IsAnySessionProcessing()
+			_ = manager.GetAllSessions()
+		}()
+	}
+
+	wg.Wait()
+
+	// Should have all sessions (none are stale)
+	assert.Equal(t, 5, manager.GetActiveSessionCount())
+	// Should have 50 messages total
+	assert.Equal(t, 50, manager.GetTotalQueueDepth())
+}

internal/worker/sse/broadcaster.go

@@ -212,7 +212,7 @@ func (b *Broadcaster) HandleSSE(w http.ResponseWriter, r *http.Request) {
 	defer b.RemoveClient(client)
 
 	// Send initial connection message
-	fmt.Fprintf(w, "data: {\"type\":\"connected\",\"clientId\":\"%s\"}\n\n", client.ID)
+	_, _ = fmt.Fprintf(w, "data: {\"type\":\"connected\",\"clientId\":\"%s\"}\n\n", client.ID)
 	client.Flusher.Flush()
 
 	// Wait for client disconnect

marketplace.json

@@ -13,7 +13,7 @@
     {
       "name": "claude-mnemonic",
       "description": "Persistent memory system for Claude Code - Go implementation with SQLite and ChromaDB vector search",
-      "version": "0.11.9",
+      "version": "0.11.65",
       "author": {
         "name": "lukaszraczylo",
         "url": "https://github.com/lukaszraczylo"
@@ -23,25 +23,25 @@
       "license": "MIT",
       "homepage": "https://github.com/lukaszraczylo/claude-mnemonic",
       "releases": {
-        "latest": "0.11.9",
+        "latest": "0.11.65",
         "versions": {
-          "0.11.9": {
-            "releaseDate": "2026-01-10",
+          "0.11.65": {
+            "releaseDate": "2026-03-09",
             "downloads": {
               "darwin-arm64": {
-                "url": "https://github.com/lukaszraczylo/claude-mnemonic/releases/download/v0.11.9/claude-mnemonic_0.11.9_darwin_arm64.tar.gz",
+                "url": "https://github.com/lukaszraczylo/claude-mnemonic/releases/download/v0.11.65/claude-mnemonic_0.11.65_darwin_arm64.tar.gz",
                 "format": "tar.gz",
-                "sha256": "f93ace966b70fccbe6dd7ed8c9e4babbb968ad92e51e71a19c7dd4d15b765237"
+                "sha256": "0d2f821d2a31a36d2d7b0db8b84b23910a882052a00a4e8a08b826d8b7998e73"
               },
               "linux-amd64": {
-                "url": "https://github.com/lukaszraczylo/claude-mnemonic/releases/download/v0.11.9/claude-mnemonic_0.11.9_linux_amd64.tar.gz",
+                "url": "https://github.com/lukaszraczylo/claude-mnemonic/releases/download/v0.11.65/claude-mnemonic_0.11.65_linux_amd64.tar.gz",
                 "format": "tar.gz",
-                "sha256": "be2e952fb58c305c0a9258e755898988fb6a7c47445bb77374812999154c0209"
+                "sha256": "056267bc915f4126af525acf34776bfc7c713059af171370f9c071dad41f90eb"
               },
               "windows-amd64": {
-                "url": "https://github.com/lukaszraczylo/claude-mnemonic/releases/download/v0.11.9/claude-mnemonic_0.11.9_windows_amd64.zip",
+                "url": "https://github.com/lukaszraczylo/claude-mnemonic/releases/download/v0.11.65/claude-mnemonic_0.11.65_windows_amd64.zip",
                 "format": "zip",
-                "sha256": "89cc7eca2c1a5fd354ec0b2de18bfecf1f5d3e52e6062070a20394c83cc675a5"
+                "sha256": "e5242d71ae34b85e4d2b48281e03f897d85f4d5e5c7700e9ca5779061f790135"
               }
             }
           }

pkg/hooks/response.go

@@ -2,6 +2,7 @@
 package hooks
 
 import (
+	"context"
 	"crypto/sha256"
 	"encoding/hex"
 	"encoding/json"
@@ -9,6 +10,7 @@ import (
 	"io"
 	"os"
 	"path/filepath"
+	"time"
 )
 
 // HookResponse is the response sent back to Claude Code.
@@ -31,6 +33,14 @@ func ProjectIDWithName(cwd string) string {
 	return fmt.Sprintf("%s_%s", dirName, shortHash)
 }
 
+// HookDeadline returns a context with the hook's timeout budget minus a safety margin.
+// This ensures hooks return gracefully before Claude kills them.
+func HookDeadline(timeout time.Duration) (context.Context, context.CancelFunc) {
+	// Use 80% of the timeout to leave margin for response serialization
+	safeTimeout := time.Duration(float64(timeout) * 0.8)
+	return context.WithTimeout(context.Background(), safeTimeout)
+}
+
 // Exit codes for Claude Code hooks
 const (
 	ExitSuccess         = 0
@@ -92,7 +102,7 @@ func RunHook[T any](hookName string, handler HookHandler[T]) {
 
 	// Parse input
 	var input T
-	if err := json.Unmarshal(inputData, &input); err != nil {
+	if err = json.Unmarshal(inputData, &input); err != nil {
 		WriteError(hookName, err)
 		os.Exit(1)
 	}

pkg/hooks/worker.go

@@ -3,6 +3,7 @@ package hooks
 
 import (
 	"bytes"
+	"context"
 	"encoding/json"
 	"fmt"
 	"net"
@@ -12,6 +13,8 @@ import (
 	"path/filepath"
 	"strconv"
 	"strings"
+	"sync"
+	"syscall"
 	"time"
 )
 
@@ -22,13 +25,53 @@ const (
 	// DefaultWorkerPort is the default worker port.
 	DefaultWorkerPort = 37777
 
-	// HealthCheckTimeout is the timeout for health checks (reduced from 5s for faster startup).
-	HealthCheckTimeout = 1 * time.Second
+	// HealthCheckTimeout is the timeout for health checks.
+	HealthCheckTimeout = 2 * time.Second
 
 	// StartupTimeout is the timeout for worker startup.
 	StartupTimeout = 30 * time.Second
+
+	// workerCacheMaxAge is how long the worker cache is considered fresh.
+	workerCacheMaxAge = 10 * time.Second
+
+	// circuitBreakerCooldown is how long to wait after a startup failure before retrying.
+	circuitBreakerCooldown = 30 * time.Second
+
+	// healthCheckRetries is the number of health check attempts before declaring dead.
+	healthCheckRetries = 3
+
+	// healthCheckRetryDelay is the delay between health check retries.
+	healthCheckRetryDelay = 200 * time.Millisecond
 )
 
+var (
+	// circuitBreakerMu protects lastStartupFailure.
+	circuitBreakerMu   sync.Mutex
+	lastStartupFailure time.Time
+)
+
+// IsWorkerAvailable performs a fast check without network calls.
+// Returns true if the worker is likely available, false if definitely down.
+func IsWorkerAvailable() bool {
+	// Check circuit breaker first
+	circuitBreakerMu.Lock()
+	if !lastStartupFailure.IsZero() && time.Since(lastStartupFailure) < circuitBreakerCooldown {
+		circuitBreakerMu.Unlock()
+		return false
+	}
+	circuitBreakerMu.Unlock()
+
+	// Check PID cache
+	entry := readWorkerCache()
+	if entry == nil {
+		return true // No cache = unknown, don't block
+	}
+
+	// Cache exists and is fresh (readWorkerCache already checks staleness)
+	// Check if cached process is alive
+	return isProcessAlive(entry.PID)
+}
+
 // GetWorkerPort returns the worker port from environment or default.
 func GetWorkerPort() int {
 	if port := os.Getenv("CLAUDE_MNEMONIC_WORKER_PORT"); port != "" {
@@ -40,29 +83,149 @@ func GetWorkerPort() int {
 }
 
 // IsWorkerRunning checks if the worker is running and healthy.
+// Parses the JSON health response to check the "ready" field when available.
+// Falls back to HTTP status code check for backwards compatibility.
 func IsWorkerRunning(port int) bool {
 	client := &http.Client{Timeout: HealthCheckTimeout}
 	resp, err := client.Get(fmt.Sprintf("http://127.0.0.1:%d/api/health", port))
 	if err != nil {
 		return false
 	}
-	defer resp.Body.Close()
+	defer func() { _ = resp.Body.Close() }()
+
+	// Try to parse JSON response for structured health check
+	var health struct {
+		Ready bool `json:"ready"`
+	}
+	if err := json.NewDecoder(resp.Body).Decode(&health); err == nil {
+		return health.Ready
+	}
+
+	// Fallback: treat HTTP 200 as healthy (backwards compatibility)
 	return resp.StatusCode == http.StatusOK
 }
 
+// workerCachePath returns the path to the worker cache file.
+func workerCachePath() string {
+	home := os.Getenv("HOME")
+	if home == "" {
+		return ""
+	}
+	return filepath.Join(home, ".claude-mnemonic", ".worker-cache")
+}
+
+// workerCacheEntry holds cached worker state: "port:pid:timestamp".
+type workerCacheEntry struct {
+	Timestamp time.Time
+	Port      int
+	PID       int
+}
+
+// readWorkerCache reads the worker cache file and returns the entry if fresh.
+func readWorkerCache() *workerCacheEntry {
+	path := workerCachePath()
+	if path == "" {
+		return nil
+	}
+	data, err := os.ReadFile(path)
+	if err != nil {
+		return nil
+	}
+	parts := strings.SplitN(strings.TrimSpace(string(data)), ":", 3)
+	if len(parts) != 3 {
+		return nil
+	}
+	port, err := strconv.Atoi(parts[0])
+	if err != nil || port <= 0 {
+		return nil
+	}
+	pid, err := strconv.Atoi(parts[1])
+	if err != nil || pid <= 0 {
+		return nil
+	}
+	ts, err := strconv.ParseInt(parts[2], 10, 64)
+	if err != nil {
+		return nil
+	}
+	entry := &workerCacheEntry{
+		Port:      port,
+		PID:       pid,
+		Timestamp: time.Unix(ts, 0),
+	}
+	// Check freshness
+	if time.Since(entry.Timestamp) > workerCacheMaxAge {
+		return nil
+	}
+	return entry
+}
+
+// writeWorkerCache writes the worker cache file.
+func writeWorkerCache(port, pid int) {
+	path := workerCachePath()
+	if path == "" {
+		return
+	}
+	// Ensure directory exists
+	dir := filepath.Dir(path)
+	_ = os.MkdirAll(dir, 0o700)
+	data := fmt.Sprintf("%d:%d:%d", port, pid, time.Now().Unix())
+	_ = os.WriteFile(path, []byte(data), 0o600)
+}
+
+// isProcessAlive checks if a process with the given PID exists and is alive.
+func isProcessAlive(pid int) bool {
+	proc, err := os.FindProcess(pid)
+	if err != nil {
+		return false
+	}
+	// Signal 0 checks if process exists without actually sending a signal.
+	err = proc.Signal(syscall.Signal(0))
+	return err == nil
+}
+
+// isWorkerRunningWithRetries checks if the worker is running, retrying on timeout.
+// Returns true only if health check succeeds. Returns false if all retries fail.
+func isWorkerRunningWithRetries(port int) bool {
+	for i := 0; i < healthCheckRetries; i++ {
+		if IsWorkerRunning(port) {
+			return true
+		}
+		if i < healthCheckRetries-1 {
+			time.Sleep(healthCheckRetryDelay)
+		}
+	}
+	return false
+}
+
 // EnsureWorkerRunning ensures the worker is running, starting it if necessary.
 // If a worker is already running and healthy with matching version, it reuses it.
 // If version mismatch or unhealthy, it kills the old worker and starts fresh.
 func EnsureWorkerRunning() (int, error) {
 	port := GetWorkerPort()
 
-	// Check if already running and healthy
-	if IsWorkerRunning(port) {
+	// Fast path: check PID cache before making any HTTP calls.
+	if entry := readWorkerCache(); entry != nil && entry.Port == port {
+		if isProcessAlive(entry.PID) {
+			return port, nil
+		}
+	}
+
+	// Circuit breaker: if we failed to start recently, don't retry immediately.
+	circuitBreakerMu.Lock()
+	if !lastStartupFailure.IsZero() && time.Since(lastStartupFailure) < circuitBreakerCooldown {
+		circuitBreakerMu.Unlock()
+		return 0, fmt.Errorf("worker startup failed recently (circuit breaker open, retry after %v)", circuitBreakerCooldown-time.Since(lastStartupFailure))
+	}
+	circuitBreakerMu.Unlock()
+
+	// Check if already running and healthy (with retries to avoid false negatives under load)
+	if isWorkerRunningWithRetries(port) {
 		// Check version - if mismatch, restart (unless both are dev builds)
 		if runningVersion := GetWorkerVersion(port); runningVersion != "" {
 			if runningVersion != Version {
 				// For dev/dirty builds, don't restart if base versions match
 				if versionsCompatible(runningVersion, Version) {
+					updateCacheFromPort(port)
 					return port, nil
 				}
 				fmt.Fprintf(os.Stderr, "[claude-mnemonic] Worker version mismatch (running: %s, expected: %s), restarting...\n", runningVersion, Version)
@@ -72,23 +235,34 @@ func EnsureWorkerRunning() (int, error) {
 				time.Sleep(500 * time.Millisecond)
 			} else {
 				// Version matches, reuse existing worker
+				updateCacheFromPort(port)
 				return port, nil
 			}
 		} else {
 			// Couldn't get version, assume it's fine
+			updateCacheFromPort(port)
 			return port, nil
 		}
 	}
 
-	// Check if port is in use but worker is unhealthy
+	// Port is in use but health check failed -- worker may be slow, not dead.
 	if IsPortInUse(port) {
-		// Something is using the port but not responding to health checks
-		// Try to kill it
+		// The port is responding to TCP but health check timed out.
+		// Don't kill it -- it's likely just under load. Give it more time.
+		fmt.Fprintf(os.Stderr, "[claude-mnemonic] Worker on port %d is slow to respond, waiting...\n", port)
+		// Try a few more times with longer delays before giving up
+		for i := 0; i < 3; i++ {
+			time.Sleep(500 * time.Millisecond)
+			if IsWorkerRunning(port) {
+				updateCacheFromPort(port)
+				return port, nil
+			}
+		}
+		// Still not healthy after extended wait -- kill and restart
+		fmt.Fprintf(os.Stderr, "[claude-mnemonic] Worker unresponsive after extended wait, restarting...\n")
 		if err := KillProcessOnPort(port); err != nil {
-			// Log but continue - maybe it will die on its own
 			fmt.Fprintf(os.Stderr, "[claude-mnemonic] Warning: failed to kill unhealthy process on port %d: %v\n", port, err)
 		}
-		// Wait a moment for port to be released
 		time.Sleep(500 * time.Millisecond)
 	}
 
@@ -103,9 +277,14 @@ func EnsureWorkerRunning() (int, error) {
 	cmd.Stdout = os.Stderr
 	cmd.Stderr = os.Stderr
 	if err := cmd.Start(); err != nil {
+		circuitBreakerMu.Lock()
+		lastStartupFailure = time.Now()
+		circuitBreakerMu.Unlock()
 		return 0, fmt.Errorf("failed to start worker: %w", err)
 	}
 
+	pid := cmd.Process.Pid
+
 	// Wait for worker to be ready with exponential backoff
 	deadline := time.Now().Add(StartupTimeout)
 	backoff := 50 * time.Millisecond
@@ -113,6 +292,7 @@ func EnsureWorkerRunning() (int, error) {
 
 	for time.Now().Before(deadline) {
 		if IsWorkerRunning(port) {
+			writeWorkerCache(port, pid)
 			return port, nil
 		}
 		time.Sleep(backoff)
@@ -123,9 +303,31 @@ func EnsureWorkerRunning() (int, error) {
 		}
 	}
 
+	circuitBreakerMu.Lock()
+	lastStartupFailure = time.Now()
+	circuitBreakerMu.Unlock()
 	return 0, fmt.Errorf("worker failed to start within timeout")
 }
 
+// updateCacheFromPort finds the PID of the process on the port and updates the cache.
+func updateCacheFromPort(port int) {
+	cmd := exec.Command("lsof", "-t", "-i", fmt.Sprintf(":%d", port)) // #nosec G204 -- port is from internal config
+	output, err := cmd.Output()
+	if err != nil {
+		return
+	}
+	pidStr := strings.TrimSpace(string(output))
+	// Take first PID if multiple
+	if idx := strings.Index(pidStr, "\n"); idx > 0 {
+		pidStr = pidStr[:idx]
+	}
+	pid, err := strconv.Atoi(pidStr)
+	if err != nil || pid <= 0 {
+		return
+	}
+	writeWorkerCache(port, pid)
+}
+
 // GetWorkerVersion gets the version of the running worker.
 func GetWorkerVersion(port int) string {
 	client := &http.Client{Timeout: HealthCheckTimeout}
@@ -133,7 +335,7 @@ func GetWorkerVersion(port int) string {
 	if err != nil {
 		return ""
 	}
-	defer resp.Body.Close()
+	defer func() { _ = resp.Body.Close() }()
 
 	if resp.StatusCode != http.StatusOK {
 		return ""
@@ -208,8 +410,6 @@ func findWorkerBinary() string {
 	locations := []string{
 		"./worker",
 		"./bin/worker",
-		filepath.Join(home, ".claude/plugins/cache/claude-mnemonic/claude-mnemonic/1.0.0/worker"),
-		filepath.Join(home, ".claude/plugins/marketplaces/claude-mnemonic/worker"),
 	}
 
 	for _, loc := range locations {
@@ -218,6 +418,19 @@ func findWorkerBinary() string {
 		}
 	}
 
+	// Try cache directory with any version (glob returns lexically sorted matches)
+	matches, _ := filepath.Glob(filepath.Join(home, ".claude/plugins/cache/claude-mnemonic/claude-mnemonic/*/worker"))
+	if len(matches) > 0 {
+		// Use the last match (latest version due to lexical sorting)
+		return matches[len(matches)-1]
+	}
+
+	// Try marketplaces directory
+	marketplacePath := filepath.Join(home, ".claude/plugins/marketplaces/claude-mnemonic/worker")
+	if _, err := os.Stat(marketplacePath); err == nil {
+		return marketplacePath
+	}
+
 	// Try PATH
 	if path, err := exec.LookPath("claude-mnemonic-worker"); err == nil {
 		return path
@@ -243,7 +456,7 @@ func POST(port int, path string, body interface{}) (map[string]interface{}, erro
 	if err != nil {
 		return nil, err
 	}
-	defer resp.Body.Close()
+	defer func() { _ = resp.Body.Close() }()
 
 	if resp.StatusCode >= 400 {
 		return nil, fmt.Errorf("request failed: %s", resp.Status)
@@ -251,13 +464,38 @@ func POST(port int, path string, body interface{}) (map[string]interface{}, erro
 
 	var result map[string]interface{}
 	if err := json.NewDecoder(resp.Body).Decode(&result); err != nil {
-		// Not all endpoints return JSON body - return empty map for success with no body
-		return map[string]interface{}{}, nil
+		// Not all endpoints return JSON
+		return nil, nil
 	}
 
 	return result, nil
 }
 
+// POSTWithContext sends a POST request using the provided context.
+// Used for fire-and-forget calls where we want to control the timeout externally.
+func POSTWithContext(ctx context.Context, port int, path string, body interface{}) error {
+	jsonBody, err := json.Marshal(body)
+	if err != nil {
+		return err
+	}
+
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost,
+		fmt.Sprintf("http://127.0.0.1:%d%s", port, path),
+		bytes.NewReader(jsonBody))
+	if err != nil {
+		return err
+	}
+	req.Header.Set("Content-Type", "application/json")
+
+	client := &http.Client{Timeout: 10 * time.Second}
+	resp, err := client.Do(req)
+	if err != nil {
+		return err
+	}
+	defer func() { _ = resp.Body.Close() }()
+	return nil
+}
+
 // GET sends a GET request to the worker.
 func GET(port int, path string) (map[string]interface{}, error) {
 	client := &http.Client{Timeout: 10 * time.Second}
@@ -266,7 +504,7 @@ func GET(port int, path string) (map[string]interface{}, error) {
 	if err != nil {
 		return nil, err
 	}
-	defer resp.Body.Close()
+	defer func() { _ = resp.Body.Close() }()
 
 	if resp.StatusCode >= 400 {
 		return nil, fmt.Errorf("request failed: %s", resp.Status)

pkg/hooks/worker_test.go

@@ -517,7 +517,7 @@ func TestProjectIDWithName_Uniqueness(t *testing.T) {
 // TestHookConstants tests hook-related constants.
 func TestHookConstants(t *testing.T) {
 	assert.Equal(t, 37777, DefaultWorkerPort)
-	assert.Equal(t, 1*time.Second, HealthCheckTimeout)
+	assert.Equal(t, 2*time.Second, HealthCheckTimeout)
 	assert.Equal(t, 30*time.Second, StartupTimeout)
 }
 

pkg/sanitize/sanitize.go

@@ -0,0 +1,63 @@
+// Package sanitize provides content cleaning utilities for stripping
+// Claude Code internal XML artifacts from captured text.
+package sanitize
+
+import (
+	"regexp"
+	"strings"
+)
+
+// systemXMLTags lists Claude Code internal XML tags that should be stripped
+// from captured content before processing. These are system-level artifacts
+// that pollute observations and summaries when stored.
+var systemXMLTags = []string{
+	// Claude Code task/agent system
+	"task-notification",
+	// System reminders injected by Claude Code
+	"system-reminder",
+	// Claude-mnemonic's own context injection
+	"relevant-memory",
+	// Hook output wrappers
+	"user-prompt-submit-hook",
+	// Large output persistence
+	"persisted-output",
+	// Tool loading system
+	"available-deferred-tools",
+	// Fast mode info
+	"fast_mode_info",
+	// Anthropic internal
+	"antml_thinking",
+	"antml_function_calls",
+}
+
+// systemXMLRegexps are compiled regexps for each tag, built once at init.
+var systemXMLRegexps []*regexp.Regexp
+
+func init() {
+	systemXMLRegexps = make([]*regexp.Regexp, len(systemXMLTags))
+	for i, tag := range systemXMLTags {
+		// Match opening tag (with optional attributes), content (including newlines), and closing tag
+		systemXMLRegexps[i] = regexp.MustCompile(`(?s)<` + regexp.QuoteMeta(tag) + `[^>]*>.*?</` + regexp.QuoteMeta(tag) + `>`)
+	}
+}
+
+// StripSystemXML removes known Claude Code internal XML blocks from text.
+// This prevents system artifacts like <task-notification>, <system-reminder>,
+// and <relevant-memory> from being stored in observations and summaries.
+func StripSystemXML(s string) string {
+	// Quick check: if no angle brackets, nothing to strip
+	if !strings.Contains(s, "<") {
+		return s
+	}
+
+	for _, re := range systemXMLRegexps {
+		s = re.ReplaceAllString(s, "")
+	}
+
+	// Clean up resulting double-blank-lines from removed blocks
+	for strings.Contains(s, "\n\n\n") {
+		s = strings.ReplaceAll(s, "\n\n\n", "\n\n")
+	}
+
+	return strings.TrimSpace(s)
+}

pkg/sanitize/sanitize_test.go

@@ -0,0 +1,113 @@
+package sanitize
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestStripSystemXML(t *testing.T) {
+	tests := []struct {
+		name     string
+		input    string
+		expected string
+	}{
+		{
+			name:     "no XML tags",
+			input:    "Hello, this is plain text with no XML.",
+			expected: "Hello, this is plain text with no XML.",
+		},
+		{
+			name:     "empty string",
+			input:    "",
+			expected: "",
+		},
+		{
+			name:  "task-notification block",
+			input: "Before\n<task-notification>\n<task-id>abc123</task-id>\n<status>completed</status>\n<summary>Agent completed</summary>\n<result>Some result text</result>\n</task-notification>\nAfter",
+			expected: "Before\n\nAfter",
+		},
+		{
+			name:     "system-reminder block",
+			input:    "Content before\n<system-reminder>\nSome system reminder text\n</system-reminder>\nContent after",
+			expected: "Content before\n\nContent after",
+		},
+		{
+			name:     "relevant-memory block",
+			input:    "<relevant-memory>\n# Relevant Knowledge\n## 1. Something\n</relevant-memory>\nActual content",
+			expected: "Actual content",
+		},
+		{
+			name:     "system-reminder with attributes",
+			input:    "Before <system-reminder type=\"hook\">reminder text</system-reminder> After",
+			expected: "Before  After",
+		},
+		{
+			name:     "multiple different tags",
+			input:    "Start\n<task-notification><status>done</status></task-notification>\nMiddle\n<system-reminder>reminder</system-reminder>\nEnd",
+			expected: "Start\n\nMiddle\n\nEnd",
+		},
+		{
+			name:     "persisted-output block",
+			input:    "Result: <persisted-output>\nLarge output stored at: /tmp/foo\n</persisted-output> done",
+			expected: "Result:  done",
+		},
+		{
+			name:     "available-deferred-tools block",
+			input:    "<available-deferred-tools>\nTool1\nTool2\n</available-deferred-tools>\nUser message",
+			expected: "User message",
+		},
+		{
+			name:     "fast_mode_info block",
+			input:    "Text <fast_mode_info>\nFast mode uses same model\n</fast_mode_info> more text",
+			expected: "Text  more text",
+		},
+		{
+			name:     "preserves non-system XML",
+			input:    "<observation><type>bugfix</type><title>Fixed thing</title></observation>",
+			expected: "<observation><type>bugfix</type><title>Fixed thing</title></observation>",
+		},
+		{
+			name:     "no angle brackets fast path",
+			input:    "Just plain text without any special characters",
+			expected: "Just plain text without any special characters",
+		},
+		{
+			name:     "collapses triple newlines",
+			input:    "Before\n<system-reminder>x</system-reminder>\n\n\nAfter",
+			expected: "Before\n\nAfter",
+		},
+		{
+			name:  "real-world task notification",
+			input: "Here is the analysis:\n<task-notification>\n<task-id>a077fd04aed547ce1</task-id>\n<tool-use-id>toolu_01Kw2GwsArQQR1F9aV3VfzhP</tool-use-id>\n<status>completed</status>\n<summary>Agent \"Analyze agency-agents repo structure\" completed</summary>\n<result>Now I have all the information I need. Let me compile a comprehensive report.\n\n## Agency-Agents Directory Exploration Report\n\n### 1. Agent File Count and Organization\n\n**Total Agent Files (excluding strategy docs):** 61 agents across 9 categories</result>\n</task-notification>\nThe repo has 61 agents.",
+			expected: "Here is the analysis:\n\nThe repo has 61 agents.",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			result := StripSystemXML(tt.input)
+			assert.Equal(t, tt.expected, result)
+		})
+	}
+}
+
+func BenchmarkStripSystemXML(b *testing.B) {
+	// Text with no XML (fast path)
+	b.Run("no_xml", func(b *testing.B) {
+		text := "This is plain text without any XML tags at all."
+		b.ResetTimer()
+		for i := 0; i < b.N; i++ {
+			StripSystemXML(text)
+		}
+	})
+
+	// Text with system XML to strip
+	b.Run("with_xml", func(b *testing.B) {
+		text := "Before\n<task-notification>\n<task-id>abc</task-id>\n<status>done</status>\n</task-notification>\n<system-reminder>reminder text here</system-reminder>\nAfter"
+		b.ResetTimer()
+		for i := 0; i < b.N; i++ {
+			StripSystemXML(text)
+		}
+	})
+}

pkg/similarity/clustering_test.go

@@ -290,3 +290,334 @@ func TestClusterObservations_PreservesOrder(t *testing.T) {
 	require.NotEmpty(t, clustered)
 	assert.Equal(t, int64(1), clustered[0].ID, "First observation should be kept as first result")
 }
+
+// =============================================================================
+// TESTS FOR OPTIMIZED CLUSTERING (triggered when len(observations) > 50)
+// =============================================================================
+
+func TestClusterObservationsOptimized_LargeSet(t *testing.T) {
+	t.Parallel()
+
+	// Create 60 observations to trigger optimized path (threshold is 50)
+	observations := make([]*models.Observation, 60)
+
+	// Create 30 pairs of similar observations
+	topics := []string{
+		"authentication", "authorization", "database", "caching", "logging",
+		"monitoring", "testing", "deployment", "scaling", "security",
+		"networking", "storage", "messaging", "scheduling", "configuration",
+		"validation", "serialization", "encryption", "compression", "indexing",
+		"backup", "recovery", "migration", "versioning", "documentation",
+		"profiling", "debugging", "tracing", "alerting", "reporting",
+	}
+
+	for i := 0; i < 30; i++ {
+		// First observation of pair
+		observations[i*2] = &models.Observation{
+			ID:        int64(i*2 + 1),
+			Title:     sql.NullString{String: topics[i] + " implementation", Valid: true},
+			Narrative: sql.NullString{String: "Detailed " + topics[i] + " system design", Valid: true},
+		}
+		// Second observation of pair (similar to first)
+		observations[i*2+1] = &models.Observation{
+			ID:        int64(i*2 + 2),
+			Title:     sql.NullString{String: topics[i] + " update", Valid: true},
+			Narrative: sql.NullString{String: "Updated " + topics[i] + " logic", Valid: true},
+		}
+	}
+
+	clustered := ClusterObservations(observations, 0.4)
+
+	// With similar pairs, we should get roughly 30 clusters (one per topic)
+	t.Logf("Clustered %d observations down to %d", len(observations), len(clustered))
+	assert.Less(t, len(clustered), 60, "Similar observations should be clustered together")
+	assert.GreaterOrEqual(t, len(clustered), 1, "Should have at least one cluster")
+}
+
+func TestClusterObservationsOptimized_AllUnique(t *testing.T) {
+	t.Parallel()
+
+	// Create 55 completely unique observations with NO shared terms
+	// Each observation has only its unique term (no common words like "topic" or "content")
+	uniqueTerms := []string{
+		"aardvark", "butterfly", "caterpillar", "dragonfly", "elephant",
+		"flamingo", "giraffe", "hippopotamus", "iguana", "jellyfish",
+		"kangaroo", "leopard", "mongoose", "nightingale", "octopus",
+		"penguin", "quail", "rhinoceros", "salamander", "toucan",
+		"umbrella", "vulture", "walrus", "xylophone", "yakking",
+		"zebra123", "astronomy99", "biology88", "chemistry77", "dynamics66",
+		"economics55", "forensics44", "genetics33", "hydraulics22", "immunology11",
+		"jurisprudence", "kinetics", "linguistics", "metallurgy", "neurology",
+		"oceanography", "pharmacology", "quantumphysics", "robotics", "sociology",
+		"thermodynamics", "ultrasound", "virology", "wavelength", "xenobiology",
+		"yeastculture", "zoology123", "algebra456", "botany789", "calculus012",
+	}
+
+	observations := make([]*models.Observation, 55)
+	for i := 0; i < 55; i++ {
+		// Each observation has ONLY its unique term - no shared words
+		observations[i] = &models.Observation{
+			ID:        int64(i + 1),
+			Title:     sql.NullString{String: uniqueTerms[i], Valid: true},
+			Narrative: sql.NullString{String: uniqueTerms[i], Valid: true},
+		}
+	}
+
+	clustered := ClusterObservations(observations, 0.4)
+
+	// All unique content should remain unclustered
+	assert.Len(t, clustered, 55, "All unique observations should be kept")
+}
+
+func TestClusterObservationsOptimized_SignaturePrefiltering(t *testing.T) {
+	t.Parallel()
+
+	// Test that signature prefiltering works correctly
+	// Create observations where some have very different signatures
+	observations := make([]*models.Observation, 60)
+
+	// First half: all identical (about "authentication") - should cluster to 1
+	for i := 0; i < 30; i++ {
+		observations[i] = &models.Observation{
+			ID:        int64(i + 1),
+			Title:     sql.NullString{String: "authentication security login", Valid: true},
+			Narrative: sql.NullString{String: "JWT tokens OAuth authentication", Valid: true},
+		}
+	}
+
+	// Second half: each completely unique with NO shared terms
+	diffTerms := []string{
+		"quantumphysics", "photosynthesis", "archaeologydig", "linguisticstudy", "astronomystar",
+		"paleontologyfossil", "oceanographywave", "entomologybug", "mycologyfungi", "herpetologysnake",
+		"ornithologybird", "ichthyologyfish", "seismologyquake", "volcanologylava", "meteorologyrain",
+		"cartographymap", "ethnographyculture", "philologyword", "numismaticscoin", "heraldryshield",
+		"genealogytree", "chronologytime", "typographyfont", "calligraphyink", "epigraphystone",
+		"papyrologytext", "codicologybook", "diplomaticseal", "sigillographywax", "sphragisticsring",
+	}
+	for i := 30; i < 60; i++ {
+		term := diffTerms[i-30]
+		// Each has ONLY its unique term - no shared words
+		observations[i] = &models.Observation{
+			ID:        int64(i + 1),
+			Title:     sql.NullString{String: term, Valid: true},
+			Narrative: sql.NullString{String: term, Valid: true},
+		}
+	}
+
+	clustered := ClusterObservations(observations, 0.5)
+
+	// Should have 31 clusters: 1 for all auth topics + 30 unique topics
+	t.Logf("Clustered %d observations down to %d", len(observations), len(clustered))
+	assert.Equal(t, 31, len(clustered), "Should have 31 clusters (1 auth + 30 unique)")
+}
+
+// =============================================================================
+// TESTS FOR HELPER FUNCTIONS
+// =============================================================================
+
+func TestComputeTermSignature(t *testing.T) {
+	tests := []struct {
+		terms      map[string]bool
+		compareTo  map[string]bool
+		name       string
+		expectZero bool
+		expectSame bool
+	}{
+		// ===== GOOD CASES =====
+		{
+			name:       "single term",
+			terms:      map[string]bool{"hello": true},
+			expectZero: false,
+		},
+		{
+			name:       "multiple terms",
+			terms:      map[string]bool{"hello": true, "world": true},
+			expectZero: false,
+		},
+		{
+			name:       "identical terms produce same signature",
+			terms:      map[string]bool{"alpha": true, "beta": true},
+			expectSame: true,
+			compareTo:  map[string]bool{"alpha": true, "beta": true},
+		},
+
+		// ===== EDGE CASES =====
+		{
+			name:       "empty set",
+			terms:      map[string]bool{},
+			expectZero: true,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			sig := computeTermSignature(tt.terms)
+
+			if tt.expectZero {
+				assert.Equal(t, uint64(0), sig, "Empty set should produce zero signature")
+			} else {
+				assert.NotEqual(t, uint64(0), sig, "Non-empty set should produce non-zero signature")
+			}
+
+			if tt.expectSame && tt.compareTo != nil {
+				sig2 := computeTermSignature(tt.compareTo)
+				assert.Equal(t, sig, sig2, "Identical term sets should produce identical signatures")
+			}
+		})
+	}
+}
+
+func TestComputeTermSignature_DifferentSets(t *testing.T) {
+	t.Parallel()
+
+	// Different term sets should usually produce different signatures
+	set1 := map[string]bool{"authentication": true, "security": true}
+	set2 := map[string]bool{"database": true, "migration": true}
+
+	sig1 := computeTermSignature(set1)
+	sig2 := computeTermSignature(set2)
+
+	// While hash collisions are possible, they should be rare
+	assert.NotEqual(t, sig1, sig2, "Different term sets should usually produce different signatures")
+}
+
+func TestPopCount64(t *testing.T) {
+	tests := []struct {
+		name     string
+		input    uint64
+		expected int
+	}{
+		// ===== GOOD CASES =====
+		{name: "zero", input: 0, expected: 0},
+		{name: "one", input: 1, expected: 1},
+		{name: "powers of two", input: 8, expected: 1},
+		{name: "all ones in byte", input: 0xFF, expected: 8},
+		{name: "alternating bits", input: 0xAAAAAAAAAAAAAAAA, expected: 32},
+		{name: "max uint64", input: 0xFFFFFFFFFFFFFFFF, expected: 64},
+
+		// ===== EDGE CASES =====
+		{name: "single high bit", input: 1 << 63, expected: 1},
+		{name: "sparse bits", input: 0x8000000000000001, expected: 2},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			result := popCount64(tt.input)
+			assert.Equal(t, tt.expected, result)
+		})
+	}
+}
+
+func TestIsSimilarToAny_EmptyTerms(t *testing.T) {
+	t.Parallel()
+
+	// Observation with no extractable terms
+	emptyObs := &models.Observation{
+		ID:        1,
+		Title:     sql.NullString{String: "", Valid: false},
+		Narrative: sql.NullString{String: "", Valid: false},
+	}
+
+	existing := []*models.Observation{
+		{
+			ID:        2,
+			Title:     sql.NullString{String: "Some content here", Valid: true},
+			Narrative: sql.NullString{String: "More content", Valid: true},
+		},
+	}
+
+	// Should return false when new observation has no terms
+	assert.False(t, IsSimilarToAny(emptyObs, existing, 0.3))
+}
+
+func TestExtractObservationTerms_FilesModified(t *testing.T) {
+	t.Parallel()
+
+	obs := &models.Observation{
+		ID:            1,
+		Title:         sql.NullString{String: "Code changes", Valid: true},
+		FilesModified: models.JSONStringArray{"/src/handler.go", "/pkg/models/user.go"},
+	}
+
+	terms := ExtractObservationTerms(obs)
+
+	// Should contain filenames from FilesModified
+	assert.Contains(t, terms, "handler.go")
+	assert.Contains(t, terms, "user.go")
+}
+
+func TestAddTerms_ShortWords(t *testing.T) {
+	t.Parallel()
+
+	terms := make(map[string]bool)
+
+	addTerms(terms, "I am a go developer")
+
+	// Short words (< 3 chars) should be excluded
+	assert.NotContains(t, terms, "i")
+	assert.NotContains(t, terms, "am")
+	assert.NotContains(t, terms, "a")
+	assert.NotContains(t, terms, "go") // Only 2 chars
+
+	// "developer" should be included
+	assert.Contains(t, terms, "developer")
+}
+
+func TestAddTerms_SpecialCharacters(t *testing.T) {
+	t.Parallel()
+
+	terms := make(map[string]bool)
+
+	addTerms(terms, "user_id authentication-flow JWT_token")
+
+	// Hyphens split words, but underscores are kept as part of the word
+	// (underscore is included in the tokenization regex)
+	assert.Contains(t, terms, "user_id")
+	assert.Contains(t, terms, "authentication")
+	assert.Contains(t, terms, "flow")
+	assert.Contains(t, terms, "jwt_token")
+}
+
+func TestJaccardSimilarity_SubsetSuperset(t *testing.T) {
+	t.Parallel()
+
+	subset := map[string]bool{"a": true, "b": true}
+	superset := map[string]bool{"a": true, "b": true, "c": true, "d": true}
+
+	// Subset similarity should be intersection/union = 2/4 = 0.5
+	result := JaccardSimilarity(subset, superset)
+	assert.InDelta(t, 0.5, result, 0.001)
+}
+
+func TestClusterObservations_HighThreshold(t *testing.T) {
+	t.Parallel()
+
+	// With a very high threshold, almost nothing should be clustered
+	observations := []*models.Observation{
+		{ID: 1, Title: sql.NullString{String: "authentication implementation", Valid: true}},
+		{ID: 2, Title: sql.NullString{String: "authentication update", Valid: true}},
+		{ID: 3, Title: sql.NullString{String: "authentication refactor", Valid: true}},
+	}
+
+	// With threshold of 0.9, even similar observations shouldn't cluster
+	clustered := ClusterObservations(observations, 0.9)
+
+	assert.Len(t, clustered, 3, "High threshold should prevent clustering")
+}
+
+func TestClusterObservations_LowThreshold(t *testing.T) {
+	t.Parallel()
+
+	// With a very low threshold, more things should be clustered
+	observations := []*models.Observation{
+		{ID: 1, Title: sql.NullString{String: "authentication implementation details", Valid: true}},
+		{ID: 2, Title: sql.NullString{String: "authentication security update", Valid: true}},
+		{ID: 3, Title: sql.NullString{String: "something completely different topic", Valid: true}},
+	}
+
+	// With threshold of 0.1, partial overlap should cluster
+	clustered := ClusterObservations(observations, 0.1)
+
+	// First two share "authentication", should likely cluster
+	assert.LessOrEqual(t, len(clustered), 3)
+}

plugin/.claude-plugin/plugin.json.tpl

@@ -5,5 +5,14 @@
   "author": {
     "name": "lukaszraczylo",
     "email": "[email protected]"
-  }
+  },
+  "hooks": "./hooks/hooks.json",
+  "mcpServers": {
+    "claude-mnemonic": {
+      "command": "${CLAUDE_PLUGIN_ROOT}/mcp-server",
+      "args": ["--project", "${CLAUDE_PROJECT}"],
+      "env": {}
+    }
+  },
+  "commands": ["./commands/restart.md"]
 }

scripts/download-onnx-libs.sh

@@ -6,7 +6,7 @@
 
 set -e
 
-ONNX_VERSION="1.23.2"
+ONNX_VERSION="1.24.3"
 ASSETS_DIR="internal/embedding/assets/lib"
 PLATFORM="${1:-all}"
 FORCE_DOWNLOAD=false

scripts/install.sh

@@ -7,6 +7,8 @@
 
 set -e
 
+INSTALLER_VERSION="1.1.0"
+
 # Configuration
 GITHUB_REPO="lukaszraczylo/claude-mnemonic"
 INSTALL_DIR="$HOME/.claude/plugins/marketplaces/claude-mnemonic"
@@ -40,6 +42,50 @@ error() {
     exit 1
 }
 
+# Gracefully stop worker processes (SIGTERM first, then SIGKILL after timeout)
+graceful_stop_worker() {
+    # Send SIGTERM first
+    pkill -TERM -f 'claude-mnemonic.*worker' 2>/dev/null || true
+    pkill -TERM -f '\.claude/plugins/.*/worker' 2>/dev/null || true
+    if command -v lsof &> /dev/null; then
+        lsof -ti :37777 2>/dev/null | xargs kill -TERM 2>/dev/null || true
+    elif command -v ss &> /dev/null; then
+        ss -tlnp 'sport = :37777' 2>/dev/null | awk 'NR>1 {print $6}' | grep -oP 'pid=\K[0-9]+' | xargs -r kill -TERM 2>/dev/null || true
+    elif command -v fuser &> /dev/null; then
+        fuser -k -TERM 37777/tcp 2>/dev/null || true
+    fi
+
+    # Wait up to 5 seconds for graceful shutdown
+    local waited=0
+    while [[ $waited -lt 5 ]]; do
+        if ! pgrep -f 'claude-mnemonic.*worker' &>/dev/null && ! pgrep -f '\.claude/plugins/.*/worker' &>/dev/null; then
+            return 0
+        fi
+        sleep 1
+        waited=$((waited + 1))
+    done
+
+    # Force kill if still running
+    pkill -9 -f 'claude-mnemonic.*worker' 2>/dev/null || true
+    pkill -9 -f '\.claude/plugins/.*/worker' 2>/dev/null || true
+    if command -v lsof &> /dev/null; then
+        lsof -ti :37777 2>/dev/null | xargs kill -9 2>/dev/null || true
+    elif command -v ss &> /dev/null; then
+        ss -tlnp 'sport = :37777' 2>/dev/null | awk 'NR>1 {print $6}' | grep -oP 'pid=\K[0-9]+' | xargs -r kill -9 2>/dev/null || true
+    elif command -v fuser &> /dev/null; then
+        fuser -k 37777/tcp 2>/dev/null || true
+    fi
+    sleep 1
+
+    # Remove stale PID cache to prevent hooks from using old worker info
+    rm -f "$HOME/.claude-mnemonic/.worker-cache" 2>/dev/null || true
+
+    # Verify process is gone
+    if pgrep -f 'claude-mnemonic.*worker' &>/dev/null; then
+        warn "Could not stop existing worker process"
+    fi
+}
+
 # Detect OS and architecture
 detect_platform() {
     local os arch
@@ -131,7 +177,7 @@ download_release() {
     local tmp_dir
 
     tmp_dir=$(mktemp -d)
-    trap "rm -rf $tmp_dir" EXIT
+    trap 'rm -rf "$tmp_dir"' EXIT
 
     # Construct download URL (use .zip for Windows, .tar.gz for others)
     local archive_ext="tar.gz"
@@ -147,8 +193,38 @@ download_release() {
         error "Failed to download release from: $download_url"
     fi
 
+    # Verify download integrity via checksum
+    local checksums_url="https://github.com/${GITHUB_REPO}/releases/download/${version}/checksums.txt"
+    info "Verifying download integrity..."
+    if curl -fsSL -o "$tmp_dir/checksums.txt" "$checksums_url" 2>/dev/null; then
+        local expected_hash actual_hash
+        expected_hash=$(grep "${archive_name}" "$tmp_dir/checksums.txt" | awk '{print $1}')
+        if [[ -z "$expected_hash" ]]; then
+            warn "Archive ${archive_name} not found in checksums.txt, skipping verification"
+        else
+            if command -v shasum &> /dev/null; then
+                actual_hash=$(shasum -a 256 "$tmp_dir/release.${archive_ext}" | awk '{print $1}')
+            elif command -v sha256sum &> /dev/null; then
+                actual_hash=$(sha256sum "$tmp_dir/release.${archive_ext}" | awk '{print $1}')
+            else
+                warn "No SHA256 tool found (shasum or sha256sum), skipping checksum verification"
+                actual_hash=""
+            fi
+            if [[ -n "$actual_hash" ]]; then
+                if [[ "$expected_hash" != "$actual_hash" ]]; then
+                    error "Checksum verification failed! Expected: $expected_hash Got: $actual_hash"
+                fi
+                success "Checksum verified"
+            fi
+        fi
+    else
+        warn "No checksum file available, skipping verification"
+    fi
     info "Extracting archive..."
     if [[ "$archive_ext" == "zip" ]]; then
+        if ! command -v unzip &> /dev/null; then
+            error "unzip is required for Windows archives but not installed"
+        fi
         if ! unzip -q "$tmp_dir/release.zip" -d "$tmp_dir"; then
             error "Failed to extract archive"
         fi
@@ -160,17 +236,7 @@ download_release() {
 
     # Stop existing worker if running
     info "Stopping existing worker (if running)..."
-    pkill -9 -f 'claude-mnemonic.*worker' 2>/dev/null || true
-    pkill -9 -f '\.claude/plugins/.*/worker' 2>/dev/null || true
-    # Kill process on port 37777 (use lsof on macOS, ss/fuser on Linux)
-    if command -v lsof &> /dev/null; then
-        lsof -ti :37777 | xargs kill -9 2>/dev/null || true
-    elif command -v ss &> /dev/null; then
-        ss -tlnp 'sport = :37777' 2>/dev/null | awk 'NR>1 {print $6}' | grep -oP 'pid=\K[0-9]+' | xargs -r kill -9 2>/dev/null || true
-    elif command -v fuser &> /dev/null; then
-        fuser -k 37777/tcp 2>/dev/null || true
-    fi
-    sleep 1
+    graceful_stop_worker
 
     # Create installation directories
     info "Installing to ${INSTALL_DIR}..."
@@ -178,13 +244,21 @@ download_release() {
     mkdir -p "$INSTALL_DIR/.claude-plugin"
     mkdir -p "$INSTALL_DIR/commands"
 
-    # Copy binaries
-    cp "$tmp_dir/worker" "$INSTALL_DIR/"
-    cp "$tmp_dir/mcp-server" "$INSTALL_DIR/"
-    cp "$tmp_dir/hooks/"* "$INSTALL_DIR/hooks/"
+    # Copy binaries (abort on failure — could indicate disk full or permissions issue)
+    if ! cp "$tmp_dir/worker" "$INSTALL_DIR/"; then
+        error "Failed to copy worker binary to $INSTALL_DIR/"
+    fi
+    if ! cp "$tmp_dir/mcp-server" "$INSTALL_DIR/"; then
+        error "Failed to copy mcp-server binary to $INSTALL_DIR/"
+    fi
+    if ! cp "$tmp_dir/hooks/"* "$INSTALL_DIR/hooks/"; then
+        error "Failed to copy hook binaries to $INSTALL_DIR/hooks/"
+    fi
 
     # Copy plugin configuration
-    cp "$tmp_dir/.claude-plugin/"* "$INSTALL_DIR/.claude-plugin/"
+    if ! cp "$tmp_dir/.claude-plugin/"* "$INSTALL_DIR/.claude-plugin/"; then
+        error "Failed to copy plugin configuration to $INSTALL_DIR/.claude-plugin/"
+    fi
 
     # Copy slash commands if they exist in the release
     if [[ -d "$tmp_dir/commands" ]]; then
@@ -338,72 +412,51 @@ start_worker() {
         error "Worker binary not found at $worker_path"
     fi
 
+    # Check for port conflict with a non-mnemonic process
+    if command -v lsof &> /dev/null; then
+        local port_pid
+        port_pid=$(lsof -ti :37777 2>/dev/null || true)
+        if [[ -n "$port_pid" ]]; then
+            local port_cmd
+            port_cmd=$(ps -p "$port_pid" -o comm= 2>/dev/null || true)
+            if [[ -n "$port_cmd" ]] && ! echo "$port_cmd" | grep -q "worker"; then
+                warn "Port 37777 is in use by another process: $port_cmd (PID $port_pid)"
+                warn "The worker may fail to start. Consider stopping the conflicting process."
+            fi
+        fi
+    fi
+
     info "Starting worker service..."
     nohup "$worker_path" > /tmp/claude-mnemonic-worker.log 2>&1 &
 
-    sleep 2
+    # Retry health check up to 5 times with 1s interval
+    local retries=0
+    local max_retries=5
+    while [[ $retries -lt $max_retries ]]; do
+        sleep 1
+        if curl -sS http://localhost:37777/health > /dev/null 2>&1; then
+            success "Worker started successfully at http://localhost:37777"
+            return 0
+        fi
+        retries=$((retries + 1))
+    done
 
-    if curl -sS http://localhost:37777/health > /dev/null 2>&1; then
-        success "Worker started successfully at http://localhost:37777"
-    else
-        warn "Worker may not have started properly. Check /tmp/claude-mnemonic-worker.log"
-    fi
+    warn "Worker may not have started properly after ${max_retries} attempts. Check /tmp/claude-mnemonic-worker.log"
 }
 
 # Check optional dependencies for semantic search
 check_optional_deps() {
-    local missing_deps=()
-    local install_hints=""
+    # Semantic search uses embedded ONNX runtime - no external Python/uvx dependencies needed
+    success "Semantic search enabled (embedded ONNX runtime)"
+}
 
-    # Check for Python 3.13+
-    if command -v python3 &> /dev/null; then
-        local py_version=$(python3 -c 'import sys; print(f"{sys.version_info.major}.{sys.version_info.minor}")' 2>/dev/null)
-        if [[ "$py_version" < "3.13" ]]; then
-            missing_deps+=("Python 3.13+ (found $py_version)")
-        fi
-    else
-        missing_deps+=("Python 3.13+")
-    fi
-
-    # Check for uvx
-    if ! command -v uvx &> /dev/null; then
-        missing_deps+=("uvx")
-    fi
-
-    if [[ ${#missing_deps[@]} -gt 0 ]]; then
-        echo ""
-        warn "Optional dependencies missing (needed for semantic search):"
-        for dep in "${missing_deps[@]}"; do
-            echo "  - $dep"
-        done
-        echo ""
-
-        # Detect OS and show appropriate install command
-        case "$(uname -s)" in
-            Darwin)
-                info "Install on macOS:"
-                echo "  brew install [email protected]"
-                echo "  pip3 install uv"
-                ;;
-            Linux)
-                info "Install on Linux:"
-                echo "  sudo apt install python3 python3-pip"
-                echo "  pip3 install uv"
-                ;;
-            MINGW*|MSYS*|CYGWIN*)
-                info "Install on Windows:"
-                echo "  winget install Python.Python.3.13"
-                echo "  pip install uv"
-                ;;
-        esac
-        echo ""
-        info "Note: Requires Python 3.13+. Most package managers install the latest version."
-        echo ""
-        info "Semantic search will be disabled until these are installed."
-        info "Core functionality (SQLite storage, full-text search) will work."
-        echo ""
-    else
-        success "Optional dependencies found (semantic search enabled)"
+# Rollback partially installed files on failure
+INSTALL_COMPLETE=false
+cleanup_on_failure() {
+    if [[ "$INSTALL_COMPLETE" != "true" ]]; then
+        warn "Installation did not complete — cleaning up partial install..."
+        rm -rf "$INSTALL_DIR" 2>/dev/null || true
+        rm -rf "$CACHE_DIR" 2>/dev/null || true
     fi
 }
 
@@ -411,6 +464,8 @@ check_optional_deps() {
 main() {
     local version="${1:-}"
 
+    trap cleanup_on_failure EXIT
+
     echo ""
     echo "╔═══════════════════════════════════════════════════════════╗"
     echo "║           Claude Mnemonic - Installation Script           ║"
@@ -455,6 +510,8 @@ main() {
     # Check optional dependencies
     check_optional_deps
 
+    INSTALL_COMPLETE=true
+
     echo ""
     echo "╔═══════════════════════════════════════════════════════════╗"
     echo "║                  Installation Complete!                   ║"
@@ -467,6 +524,12 @@ main() {
     echo ""
 }
 
+# Handle --version flag
+if [[ "${1:-}" == "--version" ]]; then
+    echo "claude-mnemonic installer v${INSTALLER_VERSION}"
+    exit 0
+fi
+
 # Handle --register-only flag
 if [[ "${1:-}" == "--register-only" ]]; then
     version=$(cat "$INSTALL_DIR/.claude-plugin/plugin.json" 2>/dev/null | grep '"version"' | sed -E 's/.*"([^"]+)".*/\1/' || echo "1.0.0")
@@ -486,17 +549,7 @@ if [[ "${1:-}" == "--uninstall" ]]; then
     echo ""
 
     info "Stopping worker processes..."
-    pkill -9 -f 'claude-mnemonic.*worker' 2>/dev/null || true
-    pkill -9 -f '\.claude/plugins/.*/worker' 2>/dev/null || true
-    # Kill process on port 37777 (use lsof on macOS, ss/fuser on Linux)
-    if command -v lsof &> /dev/null; then
-        lsof -ti :37777 | xargs kill -9 2>/dev/null || true
-    elif command -v ss &> /dev/null; then
-        ss -tlnp 'sport = :37777' 2>/dev/null | awk 'NR>1 {print $6}' | grep -oP 'pid=\K[0-9]+' | xargs -r kill -9 2>/dev/null || true
-    elif command -v fuser &> /dev/null; then
-        fuser -k 37777/tcp 2>/dev/null || true
-    fi
-    sleep 1
+    graceful_stop_worker
 
     info "Removing plugin directories..."
     rm -rf "$INSTALL_DIR"

scripts/register-plugin.sh

@@ -16,6 +16,35 @@ CACHE_BASE="$HOME/.claude/plugins/cache/claude-mnemonic/claude-mnemonic"
 CACHE_PATH="$CACHE_BASE/$VERSION"
 TIMESTAMP=$(date -u +"%Y-%m-%dT%H:%M:%S.000Z")
 
+# Helper: safely write JSON via tmp file with validation
+# Usage: safe_jq_write <jq_args...> <input_file>
+# The last argument is treated as the input file, output goes to input_file.tmp
+safe_jq_write() {
+    local args=("$@")
+    local input_file="${args[-1]}"
+    local tmp_file="${input_file}.tmp"
+
+    if jq "${args[@]}" > "$tmp_file"; then
+        if jq . "$tmp_file" > /dev/null 2>&1; then
+            mv "$tmp_file" "$input_file"
+        else
+            echo "ERROR: jq produced invalid JSON for $input_file, aborting"
+            rm -f "$tmp_file"
+            return 1
+        fi
+    else
+        echo "ERROR: jq failed for $input_file, aborting"
+        rm -f "$tmp_file"
+        return 1
+    fi
+}
+
+# Check that Claude Code directory exists
+if [ ! -d "$HOME/.claude" ]; then
+    echo "Warning: $HOME/.claude directory not found. Claude Code may not be installed."
+    echo "Continuing anyway, but plugin may not function until Claude Code is installed."
+fi
+
 # Ensure plugins directory exists
 mkdir -p "$HOME/.claude/plugins"
 
@@ -40,15 +69,119 @@ if [ ! -f "$MARKETPLACES_FILE" ]; then
     echo '{}' > "$MARKETPLACES_FILE"
 fi
 
-# Check if jq is available
-if command -v jq &> /dev/null; then
-    # Ensure cache directory exists and copy plugin files
-    mkdir -p "$CACHE_PATH/.claude-plugin"
-    mkdir -p "$CACHE_PATH/hooks"
-    mkdir -p "$CACHE_PATH/commands"
+# Validate marketplace path exists and contains expected files
+if [ ! -d "$MARKETPLACE_PATH" ]; then
+    echo "Warning: Marketplace directory not found at $MARKETPLACE_PATH"
+    echo "Plugin files may not be copied to cache correctly."
+fi
 
-    # Copy files from marketplace to cache
-    cp -r "$MARKETPLACE_PATH/"* "$CACHE_PATH/" 2>/dev/null || true
+# Ensure cache directory exists and copy plugin files
+mkdir -p "$CACHE_PATH/.claude-plugin"
+mkdir -p "$CACHE_PATH/hooks"
+mkdir -p "$CACHE_PATH/commands"
+
+# Copy files from marketplace to cache
+if ! cp -r "$MARKETPLACE_PATH/"* "$CACHE_PATH/" 2>/dev/null; then
+    echo "ERROR: Failed to copy plugin files to cache directory"
+    exit 1
+fi
+
+# Verify critical files exist in cache
+for f in worker mcp-server hooks/hooks.json .claude-plugin/plugin.json; do
+    if [ ! -f "$CACHE_PATH/$f" ]; then
+        echo "WARNING: Expected file $f not found in cache after copy"
+    fi
+done
+
+# --- JSON registration ---
+# Uses jq if available, falls back to python3 for systems without jq.
+
+STATUSLINE_CMD="\${CLAUDE_PLUGIN_ROOT}/hooks/statusline"
+
+# Helper: register plugin using python3 (jq-free fallback)
+register_with_python() {
+    python3 - "$PLUGINS_FILE" "$SETTINGS_FILE" "$MARKETPLACES_FILE" \
+        "$PLUGIN_KEY" "$CACHE_PATH" "$VERSION" "$TIMESTAMP" \
+        "$STATUSLINE_CMD" "$MARKETPLACE_NAME" "$MARKETPLACE_PATH" <<'PYEOF'
+import json, sys, os
+
+plugins_file, settings_file, marketplaces_file = sys.argv[1], sys.argv[2], sys.argv[3]
+plugin_key, cache_path, version, timestamp = sys.argv[4], sys.argv[5], sys.argv[6], sys.argv[7]
+statusline_cmd, marketplace_name, marketplace_path = sys.argv[8], sys.argv[9], sys.argv[10]
+
+def load_json(path):
+    try:
+        with open(path) as f:
+            return json.load(f)
+    except (json.JSONDecodeError, FileNotFoundError):
+        return {}
+
+def save_json(path, data):
+    tmp = path + ".tmp"
+    with open(tmp, "w") as f:
+        json.dump(data, f, indent=2)
+        f.write("\n")
+    os.replace(tmp, path)
+
+# 1. installed_plugins.json
+plugins = load_json(plugins_file)
+plugins.setdefault("version", 2)
+plugins.setdefault("plugins", {})
+plugins["plugins"][plugin_key] = [{
+    "scope": "user",
+    "installPath": cache_path,
+    "version": version,
+    "installedAt": timestamp,
+    "lastUpdated": timestamp,
+    "isLocal": True
+}]
+save_json(plugins_file, plugins)
+print("Plugin registered in installed_plugins.json")
+
+# 2. settings.json
+settings = load_json(settings_file)
+settings.setdefault("enabledPlugins", {})
+settings["enabledPlugins"][plugin_key] = True
+settings["statusLine"] = {
+    "type": "command",
+    "command": statusline_cmd,
+    "padding": 0
+}
+save_json(settings_file, settings)
+print("Plugin enabled in settings.json")
+print("Statusline configured in settings.json")
+
+# 3. known_marketplaces.json
+marketplaces = load_json(marketplaces_file)
+marketplaces[marketplace_name] = {
+    "source": {"source": "directory", "path": marketplace_path},
+    "installLocation": marketplace_path,
+    "lastUpdated": timestamp
+}
+save_json(marketplaces_file, marketplaces)
+print("Marketplace registered in known_marketplaces.json")
+PYEOF
+}
+
+# Detect JSON tool and register
+if command -v jq &> /dev/null; then
+    # Validate jq version (1.6+ required for //= operator)
+    JQ_VERSION=$(jq --version 2>/dev/null | sed 's/jq-//')
+    JQ_MAJOR=$(echo "$JQ_VERSION" | cut -d. -f1)
+    JQ_MINOR=$(echo "$JQ_VERSION" | cut -d. -f2)
+    if [ -n "$JQ_MAJOR" ] && [ -n "$JQ_MINOR" ]; then
+        if [ "$JQ_MAJOR" -lt 1 ] || { [ "$JQ_MAJOR" -eq 1 ] && [ "$JQ_MINOR" -lt 6 ]; }; then
+            echo "jq $JQ_VERSION found but too old (need 1.6+), trying python3..."
+            if command -v python3 &> /dev/null; then
+                register_with_python
+            else
+                echo "ERROR: jq 1.6+ or python3 is required for plugin registration"
+                exit 1
+            fi
+            echo "Plugin registered successfully using python3"
+            exit 0
+        fi
+    fi
 
     # Use jq for proper JSON manipulation
     PLUGIN_ENTRY=$(cat <<EOF
@@ -64,15 +197,12 @@ EOF
 )
 
     # Add or update the plugin entry in installed_plugins.json
-    jq --arg key "$PLUGIN_KEY" --argjson entry "$PLUGIN_ENTRY" \
-        '.plugins[$key] = $entry' "$PLUGINS_FILE" > "${PLUGINS_FILE}.tmp" \
-        && mv "${PLUGINS_FILE}.tmp" "$PLUGINS_FILE"
+    safe_jq_write --arg key "$PLUGIN_KEY" --argjson entry "$PLUGIN_ENTRY" \
+        '.plugins[$key] = $entry' "$PLUGINS_FILE"
 
     echo "Plugin registered in installed_plugins.json"
 
     # Enable the plugin in settings.json and configure statusline
-    # First ensure enabledPlugins object exists, then add our plugin
-    STATUSLINE_CMD="$MARKETPLACE_PATH/hooks/statusline"
     STATUSLINE_ENTRY=$(cat <<EOF
 {
     "type": "command",
@@ -82,9 +212,8 @@ EOF
 EOF
 )
 
-    jq --arg key "$PLUGIN_KEY" --argjson statusline "$STATUSLINE_ENTRY" \
-        '.enabledPlugins //= {} | .enabledPlugins[$key] = true | .statusLine = $statusline' "$SETTINGS_FILE" > "${SETTINGS_FILE}.tmp" \
-        && mv "${SETTINGS_FILE}.tmp" "$SETTINGS_FILE"
+    safe_jq_write --arg key "$PLUGIN_KEY" --argjson statusline "$STATUSLINE_ENTRY" \
+        '.enabledPlugins //= {} | .enabledPlugins[$key] = true | .statusLine = $statusline' "$SETTINGS_FILE"
 
     echo "Plugin enabled in settings.json"
     echo "Statusline configured in settings.json"
@@ -102,45 +231,17 @@ EOF
 EOF
 )
 
-    jq --arg key "$MARKETPLACE_NAME" --argjson entry "$MARKETPLACE_ENTRY" \
-        '.[$key] = $entry' "$MARKETPLACES_FILE" > "${MARKETPLACES_FILE}.tmp" \
-        && mv "${MARKETPLACES_FILE}.tmp" "$MARKETPLACES_FILE"
+    safe_jq_write --arg key "$MARKETPLACE_NAME" --argjson entry "$MARKETPLACE_ENTRY" \
+        '.[$key] = $entry' "$MARKETPLACES_FILE"
 
     echo "Marketplace registered in known_marketplaces.json"
-
-    # Register MCP server in settings.json
-    MCP_BINARY="$MARKETPLACE_PATH/mcp-server"
-    if [ -f "$MCP_BINARY" ]; then
-        echo "Registering MCP server in settings.json..."
-
-        # MCP server entry - note the escaped ${CLAUDE_PROJECT}
-        MCP_ENTRY=$(cat <<'EOF'
-{
-    "command": "MCP_BINARY_PLACEHOLDER",
-    "args": ["--project", "${CLAUDE_PROJECT}"],
-    "env": {}
-}
-EOF
-)
-        # Replace placeholder with actual path
-        MCP_ENTRY=$(echo "$MCP_ENTRY" | sed "s|MCP_BINARY_PLACEHOLDER|$MCP_BINARY|g")
-
-        # Add or update mcpServers field
-        if jq --arg key "claude-mnemonic" --argjson entry "$MCP_ENTRY" \
-            '.mcpServers //= {} | .mcpServers[$key] = $entry' "$SETTINGS_FILE" > "${SETTINGS_FILE}.tmp"; then
-            mv "${SETTINGS_FILE}.tmp" "$SETTINGS_FILE"
-            echo "MCP server registered successfully"
-        else
-            echo "Warning: Failed to register MCP server (jq error)"
-            rm -f "${SETTINGS_FILE}.tmp"
-        fi
-    else
-        echo "MCP server binary not found at $MCP_BINARY, skipping MCP registration"
-    fi
-
     echo "Plugin registered successfully using jq"
+
+elif command -v python3 &> /dev/null; then
+    register_with_python
+    echo "Plugin registered successfully using python3"
 else
-    echo "ERROR: jq is required for plugin registration"
-    echo "Please install jq: brew install jq (macOS) or apt-get install jq (Linux)"
+    echo "ERROR: jq or python3 is required for plugin registration"
+    echo "Please install one: brew install jq (macOS) or apt-get install jq (Linux)"
     exit 1
 fi

scripts/unregister-plugin.sh

@@ -3,6 +3,18 @@
 
 set -e
 
+# Stop running worker processes before removing binaries
+echo "Stopping worker processes..."
+pkill -TERM -f 'claude-mnemonic.*worker' 2>/dev/null || true
+pkill -TERM -f '\.claude/plugins/.*/worker' 2>/dev/null || true
+sleep 2
+# Force kill if still running
+pkill -9 -f 'claude-mnemonic.*worker' 2>/dev/null || true
+pkill -9 -f '\.claude/plugins/.*/worker' 2>/dev/null || true
+# Clean up port
+lsof -ti :37777 | xargs kill -9 2>/dev/null || true
+sleep 1
+
 PLUGINS_FILE="$HOME/.claude/plugins/installed_plugins.json"
 SETTINGS_FILE="$HOME/.claude/settings.json"
 MARKETPLACES_FILE="$HOME/.claude/plugins/known_marketplaces.json"
@@ -30,16 +42,17 @@ else
     echo "No plugins file found, skipping"
 fi
 
-# Remove from settings.json (enabledPlugins and statusLine if it points to our plugin)
+# Remove from settings.json (enabledPlugins, statusLine, and mcpServers)
 if [ -f "$SETTINGS_FILE" ]; then
-    # Remove from enabledPlugins and clear statusLine if it references our plugin
+    # Remove from enabledPlugins, clear statusLine if it references our plugin, and remove MCP server
     jq --arg key "$PLUGIN_KEY" '
         del(.enabledPlugins[$key]) |
         if .statusLine.command and (.statusLine.command | contains("claude-mnemonic")) then
             del(.statusLine)
         else
             .
-        end
+        end |
+        del(.mcpServers["claude-mnemonic"])
     ' "$SETTINGS_FILE" > "${SETTINGS_FILE}.tmp" \
         && mv "${SETTINGS_FILE}.tmp" "$SETTINGS_FILE"
     echo "Plugin removed from settings.json"

ui/index.html

@@ -4,7 +4,6 @@
   <meta charset="UTF-8">
   <meta name="viewport" content="width=device-width, initial-scale=1.0">
   <title>Claude Mnemonic Dashboard</title>
-  <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.5.1/css/all.min.css">
 </head>
 <body class="bg-gradient-to-br from-slate-900 via-slate-800 to-slate-900 min-h-screen text-white">
   <div id="app"></div>

ui/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "claude-mnemonic-dashboard",
-  "version": "v0.10.5-1-g7ab4b07-dirty",
+  "version": "v0.11.47-3-gfcab3eb-dirty",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "claude-mnemonic-dashboard",
-      "version": "v0.10.5-1-g7ab4b07-dirty",
+      "version": "v0.11.47-3-gfcab3eb-dirty",
       "dependencies": {
         "vis-data": "^7.1.9",
         "vis-network": "^9.1.9",

ui/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-mnemonic-dashboard",
-  "version": "v0.10.5-1-g7ab4b07-dirty",
+  "version": "v0.11.47-3-gfcab3eb-dirty",
   "private": true,
   "type": "module",
   "scripts": {

ui/src/main.ts

@@ -1,5 +1,6 @@
 import { createApp } from 'vue'
 import App from './App.vue'
 import './assets/main.css'
+import '@fortawesome/fontawesome-free/css/all.min.css'
 
 createApp(App).mount('#app')

ui/src/utils/api.ts

@@ -89,10 +89,19 @@ async function fetchWithRetry<T>(url: string, options: FetchOptions = {}): Promi
   throw lastError!
 }
 
+interface ObservationsResponse {
+  observations: Observation[]
+  total: number
+  limit: number
+  offset: number
+  hasMore: boolean
+}
+
 export async function fetchObservations(limit: number = 100, project?: string, signal?: AbortSignal): Promise<Observation[]> {
   const params = new URLSearchParams({ limit: String(limit) })
   if (project) params.append('project', project)
-  return fetchWithRetry<Observation[]>(`${API_BASE}/observations?${params}`, { signal })
+  const response = await fetchWithRetry<ObservationsResponse>(`${API_BASE}/observations?${params}`, { signal })
+  return response.observations || []
 }
 
 export async function fetchPrompts(limit: number = 100, project?: string, signal?: AbortSignal): Promise<UserPrompt[]> {