Skip to main content

rag

Knowledge bases you host yourself: ingest documents, then retrieve relevant passages at query time (hybrid BM25 + semantic). Reach for this when you want control over the vector backend and chunking.

On digitorn.ai (managed)? You don't need this module.

To give an agent a knowledge base on the hosted platform, open the Knowledge manager in Studio — upload files or add a website, attach it to your agent, and it searches automatically through the built-in knowledge.search tool. No YAML, no vector store to run, nothing to grant. This page documents the self-hosted rag module — reach for it only when you run your own daemon and vector backend (Qdrant/pgvector/Elasticsearch/…).

Exact tools and parameters

catalog.describe_module rag returns the real tools and every parameter, from the live registry - the authority for what actually runs.

Tools, at a glance: create_knowledge_base, list_knowledge_bases, knowledge_base_stats, delete_knowledge_base, ingest, ingest_file, ingest_directory, query, reindex, index_stats, migrate_embeddings.

Gotchas​

  • query retrieves; it does not write the answer. The agent's own model writes the answer from what query returned.
  • Database/web sources are configured on the module and driven by the indexer as background indexing - that's config, not agent tool calls. auto_index.schedule is a cron string handled by the indexer directly.

A working app​

Compiles as-is. The brain uses a Digitorn gateway model (the default - no API key of your own needed); see the note below to use your own key or a local model instead.

yaml
app:
  app_id: kb_assistant
  name: Knowledge Assistant
runtime:
  entry_agent: main
agents:
  - id: main
    system_prompt: |
      Answer from the knowledge bases: query for relevant passages, then write
      the answer from what you retrieved. Ingest new documents when asked.
    brain:
      provider: openai
      backend: openai_compat
      model: mimo-v2.5-free   # a model the Digitorn gateway serves
      config:
        api_key: placeholder  # ignored in gateway mode
tools:
  modules:
    rag: {}
  capabilities:
    grant:
      - module: rag
        tools: [query, create_knowledge_base, ingest, ingest_file, list_knowledge_bases]
  • Longer YAML / concepts: RAG