lf rag

Query your knowledge base and access RAG maintenance utilities.

Querying Documents

lf rag query "question" [flags]

Flag	Purpose
--database	Select a database (defaults to config default).
--data-processing-strategy	Filter results to a strategy.
--retrieval-strategy	Override retrieval behaviour (vector, hybrid, metadata filtered, etc.).
--top-k	Number of chunks to return.
--score-threshold	Minimum similarity score.
--filter	Apply metadata filters (key:value). Repeatable.
--include-metadata, --include-score	Show metadata/score columns.
--distance-metric, --hybrid-alpha, --rerank-model, --query-expansion, --max-tokens	Advanced knobs matching server capabilities.

Example:

lf rag query --database main_db --filter "doc_type:letter" --include-metadata \
  "Which letters mention additional clinical trials?"

Maintenance Commands

Some subcommands are hidden from --help but available for operators:

Command	Description
lf rag stats	Vector/document counts, storage usage (JSON or table).
lf rag health	Embedder/store health summary.
lf rag list	List ingested documents and metadata.
lf rag compact	Compact/optimize the vector store.
lf rag reindex	Reindex all documents using a given strategy.
lf rag clear	Delete all documents from a database (dangerous).
lf rag delete	Remove documents by ID, filename, or metadata filter.
lf rag export/import	Move datasets between environments.

⚠️ Destructive commands (clear, delete) prompt for confirmation unless you pass --force.

Troubleshooting

• Empty results – confirm dataset processing succeeded and the retrieval strategy matches your query type.
• Timeouts – large datasets can take time to process; check Celery logs or increase --server-start-timeout before retrying.
• Hybrid/Tool Errors – some smaller models don’t support tool calls; switch to a basic agent handler via configuration.

See Also

• RAG Guide
• Extending RAG