CLI reference

Every ragforge subcommand, sourced from ragforge/cli.py.

ragforge info

List all registered plugin components (parsers, chunkers, embedders, metrics) and module status. No flags.

bash

ragforge info

ragforge parse

Parse a file into clean text. Auto-detects format by extension.

bash

ragforge parse notes.md
ragforge parse report.pdf --parser docling --preview 1000

Flag	Description
`--parser`	Parser backend: text, html, pdf, docling. Default: auto-detect.
`--preview N`	Characters of text to show (default: 500).
`--json`	Output as JSON.

ragforge chunk

Parse then chunk a file.

bash

ragforge chunk notes.md --strategy structure --max-tokens 384

Flag	Description
`--parser`	Parser backend (same as parse).
`--strategy`	fixed \| structure \| docling. Default: structure.
`--max-tokens N`	Target chunk size in tokens.
`--show-text`	Print each chunk's text.
`--json`	Output as JSON.

ragforge knowledge build

Build a knowledge base from files or directories.

bash

ragforge knowledge build my-kb ./docs/ faq.md
ragforge knowledge build my-kb ./docs/ --embedder sentence-transformers

Flag	Description
`--strategy`	fixed \| structure \| docling (default: structure).
`--parser`	Parser backend.
`--embedder`	default \| sentence-transformers \| openai.
`--json`	Output as JSON.

ragforge query

Query a built knowledge base with hybrid search.

bash

ragforge query my-kb "How do refunds work?"
ragforge query my-kb "refunds" -k 3 --mode hybrid --rerank
ragforge query my-kb "refunds" --generate --llm ollama --model llama3

Flag	Description
`-k N`	Number of results (default: 5).
`--mode`	dense \| bm25 \| hybrid (default: hybrid).
`--rerank`	Apply cross-encoder reranking.
`--generate`	Generate a grounded LLM answer.
`--llm`	LLM provider: openai, anthropic, ollama.
`--model`	Override default model for the LLM provider.
`--json`	Output as JSON.

ragforge eval

ragforge eval run

Evaluate a knowledge base against a golden dataset (JSON or CSV).

bash

ragforge eval run my-kb golden.json
ragforge eval run my-kb golden.json --metrics hit_rate,mrr --rerank

Flag	Description
`-k N`	Top-k for retrieval (default: 5).
`--mode`	Retrieval mode (default: hybrid).
`--rerank`	Apply reranking.
`--generate`	Generate answers (required for judge metrics).
`--llm`	LLM provider.
`--metrics`	Comma-separated metrics.
`--json`	Output as JSON.

ragforge eval compare

A/B compare two knowledge bases on the same golden dataset.

bash

ragforge eval compare my-kb-v1 my-kb-v2 golden.json

ragforge eval bootstrap

Generate a draft golden dataset from an existing KB using an LLM. Human review required before use.

bash

ragforge eval bootstrap my-kb -n 20 --llm ollama --out draft_golden.json

ragforge agents

ragforge agents run

Run a multi-agent orchestration task from a Python config file. The config must defineagents and optionally goal, max_steps,board_name, and seed.

bash

ragforge agents run config.py --max-steps 20 --persist

Flag	Description
`--max-steps N`	Maximum orchestration steps.
`--persist`	Use SQLite-backed blackboard (default: in-memory).
`--json`	Output as JSON.

ragforge agents benchmark

Run a blackboard-vs-direct-messaging cost comparison.

bash

ragforge agents benchmark config.py

ragforge agents board

Inspect the current state of a named SQLite-backed blackboard.

bash

ragforge agents board research-task

ragforge serve

Start the HTTP/JSON API server.

bash

ragforge serve --host 0.0.0.0 --port 8000

Flag	Description
`--host`	Bind host (default: 0.0.0.0).
`--port N`	Port (default: 8000).
`--reload`	Enable auto-reload (development mode).

ragforge ui

Launch the local web dashboard (tracing, evaluation viewer, chat). Requires ragforge[ui].

bash

ragforge ui --port 8000

Flag	Description
`--host`	Bind host.
`--port N`	Port.
`--no-browser`	Don't auto-open browser.

← Previous

Architecture

HTTP API