Skip to content

History / Troubleshooting

Revisions

  • docs: refresh wiki + cortex/docs/ to match current code (Anthropic SDK, capability-token auth, full WS catalog) Brought every actively-used .md file in the repo into alignment with the post-Phase-J codebase. Two parallel agents performed the sweep after a precise code-fact reconnaissance pass. Root-level wiki (GitHub Wiki pages) ----------------------------------- - API-Reference.md: added Authentication subsection (Bearer token at ~/Library/Application Support/Cortex/auth.token, legacy X-Cortex-Auth-Token fallback, 401 + WWW-Authenticate, X-Cortex- Request-ID echo); added GET /health, GET /consent/level, POST /consent/reset, GET /api/projects, POST /api/launch/{name}; fixed /api/stress-integral and /api/helpfulness/summary response JSON shapes to match routes.py; added AUTH-first WebSocket handshake (close code 1011 on pre-AUTH frames); expanded WS message catalog to all 38 MessageType members (inbound, outbound, LeetCode cues). - Browser-Extension.md: WS section now reflects AUTH-before-IDENTIFY, close code 1011, LeetCode cue stream, and the generated-types boundary (cortex_schemas.d.ts). - Calibration.md: corrected UserBaselines example JSON to use the real schema field names from cortex/libs/schemas/state.py. - Home.md: replaced stale CORTEX_LLM__MODE with CORTEX_LLM__PROVIDER and named the three Anthropic SDK transports. - How-It-Works.md: L4 LLM Engine now names Anthropic SDK (Bedrock / Vertex / direct) + rule-based fallback explicitly. - Troubleshooting.md: removed obsolete Azure OpenAI and Ollama sections, added unified LLM provider errors section covering all three transports + fallback; fixed CORTEX_LLM__MODE crash bullet to CORTEX_LLM__FALLBACK_MODE. - README.md: corrected stale test-file count (55 → 120+). Internal cortex/ docs --------------------- - cortex/README.md: rewrote step 4 (LLM Engine) for Anthropic SDK + three transports + logical model tiers (Sonnet 4.6 / Haiku 4.5 / Opus 4.7) + rule-based fallback; updated ASCII architecture diagram (Azure/Qwen/Ollama → Anthropic SDK / Bedrock / Vertex / Direct / AMIP arm sel.); replaced LinUCB bandit description with AMIP (Adaptive Microrandomized Intervention Policy); rewrote llm_engine and eval entries in the project structure; added a new Schema Codegen section explaining the pre-commit/CI drift gate; added capability-token + AUTH-handshake notes near API Endpoints. - cortex/docs/architecture.md: L4 line now Anthropic SDK + rule fallback; added Capability Token subsection under L4/L5 Safety; expanded Repository Map to cover every current services/ subdir. - cortex/docs/apis.md: added Authentication subsection, X-Cortex- Request-ID note, /health optional-auth note, Consent section (GET /consent/level, POST /consent/reset), AUTH-handshake subsection, Generated TypeScript Types note, and the full inbound/outbound/LeetCode WS message catalog. - cortex/docs/setup.md: replaced Azure/Ollama/rule_based LLM Options with three Anthropic SDK transports + fallback; verify snippet now reads llm.provider; Azure LLM errors troubleshooting bullet replaced with LLM provider errors. - cortex/docs/deploy_anthropic.md: created (replaces deploy_azure .md); same 6-step experience flow but for Bedrock / Vertex / Direct. - cortex/docs/deploy_azure.md: deleted (Azure removed in v0.2.0). - cortex/docs/adapters.md: aligned CortexAdapter interface description and sample code with the real protocol in cortex/libs/adapters/base.py. - cortex/docs/calibration.md: corrected UserBaselines field names to match cortex/libs/schemas/state.py (hr_baseline, hrv_baseline, blink_rate_baseline, …, metric_distributions, circadian_hr_cosinor, rolling_rebaseline_seconds, ew_decay_half_life_days). Code is untouched. Audit ledger, .ralph/, .claude/, CHANGELOG, and CLAUDE.md were intentionally left alone (CHANGELOG follows the release-boundary convention; CLAUDE.md is already current).

    @StevenWang-CY StevenWang-CY committed May 19, 2026
  • docs: update all documentation to reflect DMG installer as primary setup method

    @StevenWang-CY StevenWang-CY committed Apr 4, 2026
  • Add full wiki: setup, architecture, how it works, API, troubleshooting, privacy

    @StevenWang-CY StevenWang-CY committed Mar 24, 2026