Running the CLI¶

The entrypoint passes any recognized command straight through, so the full graphify CLI is available inside the container:

docker run --rm -v "$(pwd):/work" -w /work \
  phillarmonic/graphify-docker graphify --help

Build a graph¶

Mount your project at /work and build:

docker run --rm -v "$(pwd):/work" -w /work \
  phillarmonic/graphify-docker graphify update .

Because GRAPHIFY_OUT=/data is set in the image, you can also build directly into the data volume the MCP server serves:

docker run --rm \
  -v "$(pwd):/work" -w /work \
  -v graphify-data:/data \
  phillarmonic/graphify-docker graphify update .

Anything after the image name that isn't graphify, graphify-mcp, sh, bash, python, or uv is treated as arguments to graphify-mcp — so docker run <image> --help shows the MCP server help, and docker run <image> graphify query ... runs the CLI.

LLM-backed features¶

Pass provider keys as environment variables; nothing is baked into the image:

docker run --rm -v "$(pwd):/work" -w /work \
  -e OPENAI_API_KEY \
  -e OPENAI_BASE_URL \
  -e OPENAI_MODEL \
  phillarmonic/graphify-docker graphify update .

See Environment variables for every supported key.

docker compose¶

The included docker-compose.yml runs the MCP server; override the command to run one-off CLI jobs against the same volume:

docker compose run --rm --entrypoint graphify \
  -v "$(pwd):/work" -w /work \
  graphify-mcp build .