CLI Overview - Ragrails

The Ragrails CLI is the terminal interface for building, querying, and maintaining a knowledge base. Use workflow commands when you want the shortest path: ingest, query, and chat. Use stage commands when you want JSON artifacts between steps that you can inspect, commit, diff, or hand off to another process.

Setup Wizard

Run ragrails with no subcommand to launch the setup wizard before using the CLI regularly.

ragrails

The wizard writes .ragrails.toml in the current directory. CLI commands then inherit vector database, embedding, LLM, reranker, chunking, storage, retrieval, and chat defaults from that file. A flag always overrides .ragrails.toml for one run.

Wizard section	Defaults it captures
Vector store	Provider, collection, URL
Embedding	Provider and model
LLM	Provider, model, max tokens
Reranking	Whether reranking is enabled, provider, model
Advanced defaults	Chunking, batch sizes, retrieval limits, query rewrite, intent routing, history compaction

If .ragrails.toml already exists, bare ragrails shows the current config and lets you edit, reset, or exit.

ragrails doctor

There is no ragrails setup subcommand. ragrails setup-url only installs the browser runtime used by website extraction.

Workflows

Workflow	Use it for	Details
`ingest`	Build a searchable knowledge base from sources.	Extracts, chunks, embeds, and stores content.
`query`	Search the indexed knowledge base.	Embeds the query and retrieves matching chunks.
`chat`	Ask grounded questions over the knowledge base.	Retrieves context and asks an LLM for an answer.

ragrails ingest --docs files/refund-policy.pdf \
  --vector-db qdrant --collection support --url http://localhost:6333

ragrails query "How long do refunds take?" \
  --vector-db qdrant --collection support --url http://localhost:6333

ragrails chat "How long do refunds take?" \
  --vector-db qdrant --collection support --url http://localhost:6333

Pipeline

Start with the pipeline overview, or jump directly to a stage:

Stage	Commands	Details
Extraction	`scrape`, `parse`, `fetch`	Turn websites, files, and APIs into extraction JSON.
Chunking	`chunk`	Split extraction JSON into retrievable chunks.
Embedding	`embed`	Generate vectors for chunk JSON.
Storing	`store`, `edit`, `delete`	Upsert, replace, and remove stored chunks.
Retrieval	`retrieve`	Search and optionally rerank stored chunks.
Maintenance	`edit`, `delete`, re-ingest	Refresh or remove stale knowledge base content.

ragrails parse --files files/refund-policy.pdf --output-dir files/output/docs/
ragrails chunk --input-dir files/output/docs/ --output-dir files/output/chunks/
ragrails embed --input-dir files/output/chunks/ --output-dir files/output/embedded/
ragrails store --input-dir files/output/embedded/ \
  --vector-db qdrant --collection support --url http://localhost:6333
ragrails retrieve "How long do refunds take?" \
  --vector-db qdrant --collection support --url http://localhost:6333

Command Index

Command	Purpose
`doctor`	Diagnose config, packages, keys, and optional vector DB connectivity.
`setup-url`	Install the browser runtime for URL scraping.
`scrape`	Extract website pages.
`parse`	Extract documents.
`fetch`	Extract REST API responses.
`chunk`	Split extraction JSON into chunks.
`embed`	Generate vectors for chunk JSON.
`store`	Upsert embedded chunks.
`edit`	Re-embed and replace chunks by ID.
`delete`	Delete chunks by exact ID.
`retrieve`	Search stored chunks.
`ingest`	Full ingest workflow.
`query`	Query embedding plus retrieval workflow.
`chat`	One-shot chat or interactive REPL.

Some SDK capabilities are not exposed as CLI flags: scrape DLQ retry, raw bytes document input, full API pagination/body config, and query rewriting in retrieve. Use SDK or REST API for those.

​Setup Wizard

​Workflows

​Pipeline

​Command Index

Setup Wizard

Workflows

Pipeline

Command Index