lukaszraczylo/claude-mnemonic
1
0
Fork 0
mirror of https://github.com/lukaszraczylo/claude-mnemonic.git synced 2026-10-12 04:55:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix: prevent MCP server hanging by adding concurrency, timeouts, and context propagation (#45)

Browse Source 
Root cause: synchronous MCP request processing combined with missing
context propagation to the embedding layer caused indefinite hangs when
ONNX inference was slow or the database was contended.

Changes:
- MCP server: dispatch each request in its own goroutine with semaphore
  (cap 10) and WaitGroup for clean shutdown drain
- Embedding: add context-aware mutex acquisition (acquireMutex) so
  callers can bail out instead of blocking forever on a stuck ONNX model
- Vector client: propagate context through getOrComputeEmbedding and
  replace bare RLock() calls with context-aware acquireRLockWithContext
- Worker handlers: add 15s request-scoped timeouts to all search/context
  handlers (handleSearchByPrompt, handleContextInject, handleFileContext,
  handleContextCount, handleGetObservations/Summaries/Prompts)
- Worker HTTP server: set WriteTimeout=60s (was 0); SSE endpoint extends
  deadline per-request via http.ResponseController

Fixes #45
This commit is contained in:
lukaszraczylo
2026-05-26 14:29:34 +01:00
parent 56616d0616
commit 29d57857ff
7 changed files with 244 additions and 40 deletions
Download Patch File Download Diff File Expand all files Collapse all files
internal/worker/handlers_data.go
+25 -12
View File
@@ -2,6 +2,7 @@
package worker
import (
"context"
"encoding/json"
"net/http"
"runtime"
@@ -20,6 +21,10 @@ import (
// Supports optional query parameter for semantic search via sqlite-vec.
// Supports pagination via limit and offset query parameters.
func (s *Service) handleGetObservations(w http.ResponseWriter, r *http.Request) {
// Add request-scoped timeout to prevent indefinite blocking on DB operations
ctx, cancel := context.WithTimeout(r.Context(), 15*time.Second)
defer cancel()
pagination := gorm.ParsePaginationParams(r, DefaultObservationsLimit)
project := r.URL.Query().Get("project")
query := r.URL.Query().Get("query")
@@ -38,11 +43,11 @@ func (s *Service) handleGetObservations(w http.ResponseWriter, r *http.Request)
// Use vector search if query is provided and vector client is available
if query != "" && s.vectorClient != nil && s.vectorClient.IsConnected() {
where := sqlitevec.BuildWhereFilter(sqlitevec.DocTypeObservation, "")
vectorResults, vecErr := s.vectorClient.Query(r.Context(), query, pagination.Limit*2, where)
vectorResults, vecErr := s.vectorClient.Query(ctx, query, pagination.Limit*2, where)
if vecErr == nil && len(vectorResults) > 0 {
obsIDs := sqlitevec.ExtractObservationIDs(vectorResults, project)
if len(obsIDs) > 0 {
observations, err = s.observationStore.GetObservationsByIDs(r.Context(), obsIDs, "date_desc", pagination.Limit)
observations, err = s.observationStore.GetObservationsByIDs(ctx, obsIDs, "date_desc", pagination.Limit)
if err == nil {
usedVector = true
total = int64(len(observations)) // Vector search doesn't have total, use returned count
@@ -55,10 +60,10 @@ func (s *Service) handleGetObservations(w http.ResponseWriter, r *http.Request)
if !usedVector {
if project != "" {
// Strict project filtering for dashboard - only observations from this project
observations, total, err = s.observationStore.GetObservationsByProjectStrictPaginated(r.Context(), project, pagination.Limit, pagination.Offset)
observations, total, err = s.observationStore.GetObservationsByProjectStrictPaginated(ctx, project, pagination.Limit, pagination.Offset)
} else {
// All projects
observations, total, err = s.observationStore.GetAllRecentObservationsPaginated(r.Context(), pagination.Limit, pagination.Offset)
observations, total, err = s.observationStore.GetAllRecentObservationsPaginated(ctx, pagination.Limit, pagination.Offset)
}
}
@@ -90,6 +95,10 @@ func (s *Service) handleGetObservations(w http.ResponseWriter, r *http.Request)
// handleGetSummaries returns recent summaries.
// Supports optional query parameter for semantic search via sqlite-vec.
func (s *Service) handleGetSummaries(w http.ResponseWriter, r *http.Request) {
// Add request-scoped timeout to prevent indefinite blocking on DB operations
ctx, cancel := context.WithTimeout(r.Context(), 15*time.Second)
defer cancel()
limit := gorm.ParseLimitParam(r, DefaultSummariesLimit)
project := r.URL.Query().Get("project")
query := r.URL.Query().Get("query")
@@ -107,11 +116,11 @@ func (s *Service) handleGetSummaries(w http.ResponseWriter, r *http.Request) {
// Use vector search if query is provided and vector client is available
if query != "" && s.vectorClient != nil && s.vectorClient.IsConnected() {
where := sqlitevec.BuildWhereFilter(sqlitevec.DocTypeSessionSummary, "")
vectorResults, vecErr := s.vectorClient.Query(r.Context(), query, limit*2, where)
vectorResults, vecErr := s.vectorClient.Query(ctx, query, limit*2, where)
if vecErr == nil && len(vectorResults) > 0 {
summaryIDs := sqlitevec.ExtractSummaryIDs(vectorResults, project)
if len(summaryIDs) > 0 {
summaries, err = s.summaryStore.GetSummariesByIDs(r.Context(), summaryIDs, "date_desc", limit)
summaries, err = s.summaryStore.GetSummariesByIDs(ctx, summaryIDs, "date_desc", limit)
if err == nil {
usedVector = true
}
@@ -122,9 +131,9 @@ func (s *Service) handleGetSummaries(w http.ResponseWriter, r *http.Request) {
// Fall back to SQLite if vector search not used
if !usedVector {
if project != "" {
summaries, err = s.summaryStore.GetRecentSummaries(r.Context(), project, limit)
summaries, err = s.summaryStore.GetRecentSummaries(ctx, project, limit)
} else {
summaries, err = s.summaryStore.GetAllRecentSummaries(r.Context(), limit)
summaries, err = s.summaryStore.GetAllRecentSummaries(ctx, limit)
}
}
@@ -143,6 +152,10 @@ func (s *Service) handleGetSummaries(w http.ResponseWriter, r *http.Request) {
// handleGetPrompts returns recent user prompts.
// Supports optional query parameter for semantic search via sqlite-vec.
func (s *Service) handleGetPrompts(w http.ResponseWriter, r *http.Request) {
// Add request-scoped timeout to prevent indefinite blocking on DB operations
ctx, cancel := context.WithTimeout(r.Context(), 15*time.Second)
defer cancel()
limit := gorm.ParseLimitParam(r, DefaultPromptsLimit)
project := r.URL.Query().Get("project")
query := r.URL.Query().Get("query")
@@ -160,11 +173,11 @@ func (s *Service) handleGetPrompts(w http.ResponseWriter, r *http.Request) {
// Use vector search if query is provided and vector client is available
if query != "" && s.vectorClient != nil && s.vectorClient.IsConnected() {
where := sqlitevec.BuildWhereFilter(sqlitevec.DocTypeUserPrompt, "")
vectorResults, vecErr := s.vectorClient.Query(r.Context(), query, limit*2, where)
vectorResults, vecErr := s.vectorClient.Query(ctx, query, limit*2, where)
if vecErr == nil && len(vectorResults) > 0 {
promptIDs := sqlitevec.ExtractPromptIDs(vectorResults, project)
if len(promptIDs) > 0 {
prompts, err = s.promptStore.GetPromptsByIDs(r.Context(), promptIDs, "date_desc", limit)
prompts, err = s.promptStore.GetPromptsByIDs(ctx, promptIDs, "date_desc", limit)
if err == nil {
usedVector = true
}
@@ -175,9 +188,9 @@ func (s *Service) handleGetPrompts(w http.ResponseWriter, r *http.Request) {
// Fall back to SQLite if vector search not used
if !usedVector {
if project != "" {
prompts, err = s.promptStore.GetRecentUserPromptsByProject(r.Context(), project, limit)
prompts, err = s.promptStore.GetRecentUserPromptsByProject(ctx, project, limit)
} else {
prompts, err = s.promptStore.GetAllRecentUserPrompts(r.Context(), limit)
prompts, err = s.promptStore.GetAllRecentUserPrompts(ctx, limit)
}
}