CLI

rag-server

Starts the HTTP API server.

rag-server [--config <path>] [--log-level <debug|info|warn|error>]

• --config: path to the YAML configuration file.
• --log-level: overrides log.level from the config file.

rag-cli

Synchronizes repositories and ingests markdown files into the server via API.

rag-cli --config <path> [--file <path>] [--log-level <level>]

Environment:

• SERVER_URL: base URL for the API (default: from config, or http://localhost:8080).

Behavior:

• If --file is provided, only that file is embedded and upserted. The file must be inside a datasource working directory (typically /tmp/xcontrol/<datasource>/...) created by rag-cli during sync.
• Otherwise, it iterates over global.datasources in config, syncs each repo, and ingests the markdown files it finds.

ingest (batch tool)

Direct ingestion into Postgres (no HTTP) using the RAG config:

ingest --config <path> [--only-repo <name>] [--dry-run] [--max-files <n>] \
  [--migrate-dim] [--concurrency <n>]

This tool uses the same chunking and embedding config as the server. Prefer passing an explicit --config path when running from the repo root.

ragbench (optional)

Benchmark runner for retrieval quality. It expects a YAML input file describing queries.

ragbench --in queries.yaml --api http://localhost:8080 --out report.md

Note: ragbench assumes a response shape that is not identical to the current /api/rag/query response. Use it with a compatible adapter or update the tool before relying on the metrics.