feat(brain): remote HTTP MCP server on Charlie :8501#138
Open
Mikecranesync wants to merge 9 commits into
Open
Conversation
- brain-feed.yml: continue-on-error, fast timeouts, jq payload, dead letter queue - ci-watchdog.yml: 30-min health check, severity classification, auto-issue management - replay-brain-dlq.sh: replay failed payloads once endpoint recovers - INC-2026-03-09-001: incident report for 14 consecutive failures - Ops trace documenting the change Fixes brain-feed blocking all pushes when brain-ingest endpoint is down. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- brain_server.py: env-var-driven host/port/transport (MCP_HOST, MCP_PORT, MCP_TRANSPORT) - brain_server.py: /health endpoint via custom_route (public, no auth) - brain_server.py: streamable-http mode with optional bearer auth (BRAIN_ACCESS_KEY) - deploy-brain-mcp-mac.sh: one-shot deploy for Charlie Mac Mini (port 8501, launchd) - requirements.txt: add mcp[cli] and httpx Enables multi-device access: claude mcp add --transport http open-brain http://<tailscale-ip>:8501/mcp Brain-ingest on :8500 untouched. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Telegram fault alerts now fire immediately without waiting for Cosmos analysis (which could block up to 30s). Cosmos enrichment runs as a background task and sends a follow-up message when ready. - Add send_raw_incident_alert() and send_enrichment_followup() to TelegramAlerter - Restructure _check_incidents() and _compare_twins() to alert-first - Add circuit breaker to CosmosClient (3-fail threshold, 300s cooldown) - Fix Python 3.9 syntax (str | None -> Optional[str]) in cosmos/client.py - Add 6 isolation tests verifying decoupling and circuit breaker Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace hardcoded GEMINI_API_KEY/GROQ_API_KEY/NEON_DATABASE_URL in LaunchAgent plists with doppler run -p factorylm -c dev -- wrapper. Fixes API key expired errors after key rotation in Doppler. - Update .mcp.json open-brain entry to use doppler run with full paths (doppler at ~/.local/bin/doppler, python3 at brain-venv — no bash escaping or --format flag compatibility issues) - Fix com.factorylm.brain-mcp.plist and brain-ingest.plist to use doppler run instead of hardcoded EnvironmentVariables block - Fix brain-ingest WorkingDirectory (was stale worktree path) - Add config/network.env — node identity and IP config for CHARLIE - Add scripts/health-check.sh — cluster health check script - Add telegram_bot.py timeout config (connect/read/write/pool) - Add work_orders JSON files from troubleshoot service Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
POST /api/query endpoint that accepts {question, machine?} and returns
a unified response from Qdrant KB, Nautobot device records, and the
LLM router (with Groq direct fallback).
- Searches Qdrant :8000/factorylm_brain for top-5 relevant KB chunks
using keyword-scored scroll (no embedding dependency for V1)
- Fetches Nautobot device context via HTTP when machine is provided
- Assembles context + calls LLM router :8100 (Groq fallback chain)
- Returns {answer, sources, machine_context, latency_ms}
- GET /health reports Qdrant + service connectivity
- 18 integration tests hitting real Qdrant, Nautobot, and query-api
(no mocks — per project policy)
Services: query-api port 8300
Tests: 18/18 passing
Co-Authored-By: Paperclip <noreply@paperclip.ing>
- tools/ingest_factorylm_kb.py: indexes FactoryLM equipment data into Qdrant factorylm_brain - kb/factorylm_modbus_map.md: Micro820 PLC Modbus TCP register reference Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
All nodes inherit this via git pull on session start: - Board: https://github.com/users/Mikecranesync/projects/4 - Session start: show In Progress + Todo items - On commit: add issues, move to Done - Field IDs hardcoded (Status, Todo, In Progress, Done) Closes Mikecranesync/MIRA#90 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- docs/gtm/YOUTUBE_SHORTS_GTM.md — full strategy doc (FAC-18) - docs/gtm/CONTENT_CALENDAR_4W.md — 4-week launch calendar (FAC-18) - docs/gtm/REVIEW_CHECKLIST.md — pre-publish quality gate (FAC-19) - tools/shorts_pipeline.py — LinkedIn-first vertical video producer (crop 9:16, Whisper captions, hook card, progress bar, end card, validation) - tools/thumbnail_generator.py — Pillow 1280×720 thumbnails, 6 series color map - tools/youtube_uploader.py — YouTube Data API v3 wrapper with quota guard - tools/cross_post.py — one LinkedIn master → 5 platform derivatives - tools/analytics_reporter.py — weekly analytics + self-improving calendar - workers/content_capture_tasks.py — auto-trigger Shorts pipeline on score ≥ 8 Platform priority: LinkedIn native video first (B2B ICP), Shorts as derivative. Closes #82, #83, #84, #85, #86, #87, #88. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
API_URL was hardcoded to http://165.245.138.91:8082. Now uses ${PUBLIC_API_URL:-http://localhost:8082} so it works on any host without file edits. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
MCP_HOST/MCP_PORT/MCP_TRANSPORTconfig,/healthendpoint, streamable-http mode with optional bearer auth viaBRAIN_ACCESS_KEYclaude mcp addcommand.mcp.jsonusage unaffectedWhy
Every cluster device needs brain tools (
brain_search,brain_capture, etc.). Previously required local Python, venv, and repo on each machine. Now: single persistent HTTP server on Charlie, oneclaude mcp addper device.Test plan
curl -sf http://localhost:8501/healthclaude mcp add --transport httpbrain_search("test")from remote Claude Code session🤖 Generated with Claude Code