Search and question answering

Search settings

Configure search behavior and context expansion:

search:
  limit: 5                     # Default number of results to return
  max_context_chars: 5000     # Maximum characters in expanded context
  rerank_pool_factor: 10      # Candidates per result a reranker scores

• limit: Default number of search results to return when no limit is specified. Used by CLI, MCP server, and QA. Default: 5
• max_context_chars: Hard limit on total characters in expanded content. Default: 5000.
• rerank_pool_factor: With a reranker configured, a text search fetches limit × rerank_pool_factor candidates and reranks them down to limit. Default: 10.

Context expansion is section-aware. The capability's search tool, the MCP search tools and in-code search() expand every result. HaikuRAG.search returns chunks as matched, and HaikuRAG.expand_context expands them. For structured documents (with section headers), expansion includes the entire section containing the match when it fits the budget. A section larger than the budget grows outward item-by-item from the match, staying inside the section. A section under 20% of the budget (e.g., a title+authors area) grows across section boundaries until the budget is filled, except when the match is a picture or table, which returns its section as-is. Noise labels (page headers, page footers, table of contents) are skipped. For unstructured documents, expansion grows outward item-by-item. Results without doc_item_refs (e.g., custom chunks passed to import_document) pass through unexpanded. Results whose expanded ranges overlap within a document are merged into one.

Question answering configuration

Configure the RAG capability (used by client.ask and haiku-rag ask):

qa:
  model:
    provider: ollama
    name: qwen3.8
    thinking: true
    temperature: 0.3          # Default: 0.3
    vision: true              # Set false for text-only models
  max_searches: 5       # Maximum search units per question
  max_executions: 15    # Maximum execute_code calls per question

• model: LLM configuration (see Providers)
• model.vision: Set to true for vision-capable models (qwen2.5vl, qwen3.6, gpt-4o, claude-sonnet, …). The capability's search tool only attaches picture bytes (BinaryContent) to its ToolReturn when this is true, otherwise picture bytes are withheld. See Pictures × embedder × QA model for the full matrix.
• max_searches: Maximum number of search units a capability can spend per question (default: 5). Up to three searches emitted in the same model response share one unit, so a model that rephrases its query in one response spends one unit. A search in a later response starts a new unit, as does each further group of three within one response. Searches in one response also deduplicate their returns: evidence a sibling search already showed collapses to a reference line, and each picture attaches once per response.
• max_executions: Maximum execute_code calls per question before the capability is told to answer from what it has (default: 15)

Sandbox configuration

Limits of one execute_code call, in the RAG capability and in the MCP execute_code tool:

sandbox:
  code_timeout: 60.0      # Per call: compute stops, no read or search starts past it
  max_output_chars: 50000 # Truncate output after this many chars

• code_timeout: Seconds a single execute_code call has (default: 60). Past it the sandbox starts no further host call, a document read or an in-code search() / list_documents(); one already running finishes. Computation past the limit is stopped by the interpreter, which resets the session: variables from earlier calls are gone.
• max_output_chars: Truncate code output after this many characters (default: 50000)

See RAG capability for usage details.