pluk 0.3.9

Symbol lookup / search engine CLI

Pluk

Git-commit–aware symbol lookup & impact analysis engine

---

What is a "symbol"?

In Pluk, a symbol is any named entity in your codebase that can be referenced, defined, or impacted by changes. This includes functions, classes, methods, variables, and other identifiers that appear in your source code. Pluk tracks symbols across commits and repositories to enable powerful queries like "go to definition", "find all references", and "impact analysis".

Pluk gives developers “go-to-definition”, “find-all-references”, and “blast-radius” impact queries across one or more Git repositories. Heavy lifting (indexing, querying, storage) runs in Docker containers; a lightweight host shim (pluk) boots the stack and delegates commands into a thin CLI container (plukd) that talks to an internal API.

---

Key Features

• Symbol search (pluk search) and definition lookup (pluk define)
• Impact analysis (pluk impact) to trace downstream dependents
• Commit-aware indexing (pluk diff) across Git history
• Containerized backend: PostgreSQL (graph) + Redis (broker/cache)
• Strict lifecycle: pluk start is required before any commands
• Host controls: pluk status to check, pluk cleanup to stop services

---

Quickstart

1. Install

pip install pluk

2. Start services (required)

pluk start

This creates/updates ~/.pluk/docker-compose.yml, pulls latest images, and brings up: postgres, redis, api (FastAPI), worker (Celery), and cli (idle exec target). The API stays internal to the Docker network.

3. Index and query

pluk init /path/to/repo           # queue full index (API→Redis→Celery→Postgres)
pluk search MyClass               # fuzzy lookup; defs + refs (API→Postgres, cached)
pluk define my_function           # show definition (file:line@commit)
pluk impact computeFoo            # transitive dependents (blast radius; cached)
pluk diff symbol abc123 def456    # symbol changes between commits abc123 → def456

Note: CLI commands that poll for job status (like pluk init) now display real-time output, thanks to unbuffered Python output in the CLI container.

4. Check / stop (host-side)

pluk status     # tells you if services are running
pluk cleanup    # stops services (containers stay; fast restart)

If you want a full teardown (remove containers/network), use:

docker compose -f ~/.pluk/docker-compose.yml down

---

Data Flow

How it works

• Host shim (pluk) writes the Compose file, pulls images, and runs docker compose up.
• CLI container (plukd) is the exec target; it calls the API at http://api:8000.
• API (FastAPI) serves read endpoints (/search, /define, /impact, /diff) and enqueues write jobs (/reindex) to Redis.
• Worker (Celery) consumes jobs from Redis, clones/pulls repos into a volume (/var/pluk/repos), parses deltas, and writes to Postgres.
• Reads never block on indexing; write progress can be polled via job status endpoints (planned).

---

Architecture (current)

• Single image, multiple roles: Compose selects per-service command
  • api → uvicorn pluk.api:app --host 0.0.0.0 --port 8000
  • worker → celery -A pluk.worker worker -l info
  • cli → sleep infinity (keeps container up for docker compose exec)
• Internal networking: API is not exposed to the host; CLI calls it over Docker DNS (PLUK_API_URL=http://api:8000).
• Config: PLUK_DATABASE_URL, PLUK_REDIS_URL injected via Compose; worker uses PLUK_REPOS_DIR=/var/pluk/repos.
• Images: by default the shim uses jorstors/pluk:latest, postgres:16-alpine, and redis-alpine

---

Development

• Project layout (src/pluk):
  • shim.py — host shim entrypoint (pluk)
  • cli.py — container CLI (plukd)
  • api.py — FastAPI app (internal API)
  • worker.py — Celery app & tasks
• Entry points (pyproject.toml):

[project.scripts]
pluk  = "pluk.shim:main"
plukd = "pluk.cli:main"

---

Testing

pytest

Docker must be running; services must be started via pluk start for integration tests.

---

License

MIT License