Commit Graph
26 Commits
Author SHA1 Message Date
lukaszraczylo 87088e5065 feat(mcp): expose samples/upscale/annotate tool params
Wire Samples+Upscale onto SolveTextIn and Samples+Upscale+Annotate onto
SolveGridIn, pass through to TextOptions/GridOptions. Fields ordered
fieldalignment-clean. GridResult Out now carries no_targets.
2026-06-07 17:37:25 +01:00
lukaszraczylo edb655f590 feat(captcha): self-consistency voting, optimized prompts, no-targets signal
- Add Samples to TextOptions/GridOptions; N>1 runs N solves at temp 0.5
  and majority-votes (text: most frequent; grid: tiles in strict majority).
  N<=1 keeps the single deterministic call path unchanged.
- Add NoTargets to GridResult so callers know to click Verify instead of
  looping; prompt instructs empty tiles + no_targets:true when nothing matches.
- Grid prompt: numbered cells 0..R*C-1, select any partial overlap.
- Text prompt: disambiguate similar glyphs (0/O 1/l/I 5/S 2/Z 8/B).
- Rotation prompt: reason step-by-step then emit final JSON.
2026-06-07 17:37:18 +01:00
lukaszraczylo c8edd4aa06 feat(llm): fake supports per-call response sequences
Add VisionResponses to Fake. Vision returns the element matching the
call index (clamped to last beyond length), falling back to VisionResp
when empty. Enables self-consistency voting tests.
2026-06-07 17:37:10 +01:00
lukaszraczylo a49c6ab08d feat(imageutil): add upscale and numbered grid-overlay preprocessing 2026-06-07 17:32:35 +01:00
lukaszraczylo 4ffe555bfd fix(captcha): strip reasoning-model <think> blocks before json extraction
Reasoning models prefix responses with <think>...</think> (or
<thinking>...</thinking>) scratchpad that can itself contain a JSON-looking
object. The brace-counting extractor would grab that inner object instead of
the real answer. Strip think/thinking blocks via a package-level regexp before
the balanced-object scan. Adds table cases covering both tag spellings while
keeping the plain, prose-wrapped, and unbalanced cases intact.
2026-06-07 14:34:35 +01:00
lukaszraczylo f0e1264edf chore(stealth): remove orphaned static evasions.js superseded by template 2026-06-07 14:29:54 +01:00
lukaszraczylo 22775d446e feat(stealth): make evasion script coherent with generated fingerprint
- Convert assets/evasions.js to evasions.js.tmpl (Go text/template)
- Placeholders: WebGLVendor, WebGLRenderer, Platform, Languages
- Add navigator.platform spoof from fingerprint
- ScriptParams gains OS, Browser, Locale, Seed fields (fieldalignment ordered)
- ScriptResult gains Fingerprint field — returns the identity the script was rendered for
- StealthScript calls GenerateFingerprint and renders template; single call gives both script and coherent fingerprint
- languagesJS: locale to JS list (en-GB -> 'en-GB', 'en')
- Template parsed once at init via template.Must
- StealthScriptIn in tools.go gains OS, Browser, Locale, Seed pass-through
- Tests: added TestStealthScriptCoherence asserting rendered values match fingerprint
2026-06-07 14:29:13 +01:00
lukaszraczylo 5476aefb11 feat(llm): implement native gemini generateContent provider (vision + audio)
- Replace stub Vision/Transcribe with real REST implementation
- Endpoint: POST {base}/v1beta/models/{model}:generateContent
- Auth via x-goog-api-key header
- Vision: text part + inline_data per image (mime_type from Image.MediaType)
- Transcribe: transcription prompt + inline_data audio (default audio/wav)
- Model selection: opts.Model > audioModel (transcribe) > c.model
- maxTokens mapped to generationConfig.maxOutputTokens when non-zero
- Add httptest coverage: TestGeminiVision, TestGeminiTranscribe
- Re-enable gemini in tools.go audio capability check
2026-06-07 14:24:16 +01:00
lukaszraczylo 8bfd64f7ed fix: address code review (v3/v2 detect, gemini caps, fetch timeout, audio mime, json extract, dead temp config)
- detect: recaptcha_v3 now suppresses recaptcha_v2 (hidden g-recaptcha div on v3 pages no longer mis-detected as LLM-solvable); funcaptcha drops over-broad bare "arkose" substring
- llm/gemini: stub methods return ErrUnsupported{Provider:"gemini"} so audio handler errors.As guard catches it; drop unused fmt import
- mcp/tools: remove gemini from audio capability (stub); pick audio filename (.wav/.mp3) from sniffed media type; schema notes for browser/actions no longer advertise unimplemented options
- imageutil: fetchURL uses a package client with 30s Timeout instead of http.DefaultClient
- captcha/json: extractJSON brace-counts to return the first balanced object (handles prose, markdown fences, example-before-real)
- config: remove dead CAPTCHA_LLM_TEMPERATURE / Config.Temperature; solving is fixed at temperature 0
- stealth: Action.Type doc comment drops nonexistent scroll case
2026-06-07 14:15:43 +01:00
lukaszraczylo d96b7134f1 fix(llm): retry openai chat with max_completion_tokens when max_tokens is rejected 2026-06-07 13:38:55 +01:00
lukaszraczylo e97aa3b459 feat(mcp): tool registration and handlers for solve+stealth+detect 2026-06-07 13:26:00 +01:00
lukaszraczylo cd7f6c50db feat(stealth): consistent header set with TLS/JA3 advisories 2026-06-07 13:22:31 +01:00
lukaszraczylo d477b27e49 feat(stealth): humanized interaction plan (bezier + cadence) 2026-06-07 13:21:49 +01:00
lukaszraczylo d269c87b82 feat(stealth): embeddable evasion init script with apply hints 2026-06-07 13:20:55 +01:00
lukaszraczylo 6a70f6c9d5 feat(stealth): deterministic coherent fingerprint generator 2026-06-07 13:20:13 +01:00
lukaszraczylo 13ff6449a3 feat(captcha): best-effort rotation/funcaptcha solver 2026-06-07 13:17:26 +01:00
lukaszraczylo 6e496eeac8 feat(captcha): image-grid tile solver with centroid mapping 2026-06-07 13:16:54 +01:00
lukaszraczylo 0f1ad1dc44 feat(captcha): audio captcha transcription solver 2026-06-07 13:15:57 +01:00
lukaszraczylo c0da1121d9 feat(captcha): LLM text-image OCR solver with strict-retry 2026-06-07 13:15:32 +01:00
lukaszraczylo ca9add612e feat(captcha): deterministic HTML detection for known providers 2026-06-07 13:13:10 +01:00
lukaszraczylo d438673a1f feat(imageutil): input decoding and grid centroid geometry 2026-06-07 13:12:24 +01:00
lukaszraczylo d008ba1424 feat(llm): provider factory with per-provider default base URLs 2026-06-07 13:08:27 +01:00
lukaszraczylo f6def2f71d feat(llm): anthropic vision provider (audio unsupported) 2026-06-07 13:07:57 +01:00
lukaszraczylo 967b28769f feat(llm): openai-compatible vision and transcription provider 2026-06-07 13:07:25 +01:00
lukaszraczylo 43aa78abc5 feat(llm): client interface and fake test double 2026-06-07 13:05:59 +01:00
lukaszraczylo 9a4d329224 feat(config): env-driven config with validation 2026-06-07 13:02:25 +01:00