From qrec
Provides expert guidance on qrec session recall engine for Claude Code: status checks, debugging errors, re-indexing sessions, daemon management, search troubleshooting, and teardown.
How this skill is triggered — by the user, by Claude, or both
Slash command
/qrec:qrecThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
You are the qrec expert. Use the knowledge below to answer any question.
You are the qrec expert. Use the knowledge below to answer any question. For specific actions, load and follow the matching sub-file.
Route based on $ARGUMENTS or user intent:
status / health / "is qrec running?" → status.mddebug / diagnose / troubleshoot / any error message → debug.mdre-index / reindex / "sessions not showing up" → re-index.mdteardown / uninstall / remove → teardown.mdqrec is a persistent session recall engine for Claude Code. It indexes every Claude Code conversation into a local SQLite database and provides hybrid BM25 + vector search. Sessions are indexed automatically via a cron loop inside the daemon; the daemon starts automatically when Claude Code opens.
SessionStart hook → qrec serve --daemon --no-open (starts daemon if not running; exits immediately if already up)
Daemon (port 25927) → loads embedding model in background
→ cron auto-index every 60s (picks up new/changed sessions)
→ HTTP server always up; search returns 503 until model ready
HTTP server → http://localhost:25927 (search UI + REST API)
SQLite DB → ~/.qrec/qrec.db
Embedding model → ~/.qrec/models/embeddinggemma-300M-Q8_0.gguf (~300MB, local)
Server log → ~/.qrec/qrec.log
PID file → ~/.qrec/qrec.pid
Activity log → ~/.qrec/activity.jsonl
qrec status # health + session/chunk counts + log tail
qrec serve --daemon # first-time setup + all subsequent starts; auto-downloads model, indexes, opens browser
qrec serve --daemon --no-open # start without opening browser (used by SessionStart hook)
qrec stop # stop daemon
qrec index [path] # index path (default: ~/.claude/projects/)
qrec index --force # force re-index all sessions from scratch
qrec search "<query>" [--project P] [--tag T] [--from DATE] [--to DATE] [--k N] # search (with query: POST /search; filters only: GET /sessions browse)
qrec get <session_id> # print full session markdown
qrec enrich [--limit N] # enrich sessions with summary/tags/entities
qrec enrich --force # re-enrich all sessions (use after accidental re-index to recover titles)
qrec teardown [--yes] # stop daemon + remove ~/.qrec/
qrec --version
If qrec is not in PATH yet, use the full path:
node $CLAUDE_PLUGIN_ROOT/scripts/qrec-cli.js <command>
| Method | Path | Notes |
|---|---|---|
| GET | /health | Always 200; {status, model, indexedSessions} |
| GET | / | Search UI |
| GET | /audit | Audit log UI |
| GET | /debug | Debug UI |
| POST | /search | {query, k, project?, tag?, dateFrom?, dateTo?} → results; 503 until model ready |
| GET | /sessions | List sessions |
| GET | /sessions/:id | Session detail JSON |
| GET | /sessions/:id/markdown | Session full markdown |
| GET | /audit/entries?limit=N | Audit log entries |
| GET | /activity/entries | Activity log entries |
| GET | /settings | Current config |
| POST | /settings | Update config |
| Variable | Default | Description |
|---|---|---|
QREC_EMBED_PROVIDER | local | local / ollama / openai / stub |
QREC_PORT | 25927 | Daemon port |
QREC_DIR | ~/.qrec | Override data root |
QREC_PROJECTS_DIR | ~/.claude/projects/ | Source sessions directory |
QREC_OLLAMA_HOST | http://localhost:11434 | Ollama URL |
QREC_OLLAMA_MODEL | nomic-embed-text | Ollama model |
QREC_OPENAI_KEY | — | OpenAI-compatible key |
QREC_OPENAI_BASE_URL | https://api.openai.com/v1 | OpenAI-compatible URL |
QREC_OPENAI_MODEL | text-embedding-3-small | OpenAI model |
qrec serve --daemon --no-open~/.qrec/models/)~/.claude/projects/ runs automaticallynpx claudepluginhub dvquy13/qrec --plugin qrecRecalls past Claude Code sessions using qrec CLI via Bash. Useful for queries about prior conversations, implementations, decisions, 'last time we...', or previous work.
Reads and searches OpenCode session logs from the SQLite DB. Use to review past session activity, assess AI work, diagnose session wipes, recover after migration resets.