Skip to main content
Ask a natural language question about a project. Retrieves relevant context via hybrid search and synthesizes a concise answer with source citations. Use this for direct questions; use Search when you need raw results to reason over.

​
When to Use

  • The user asks a direct question about their project (e.g., “What fire rating does door 104 have?”)
  • You want a concise, sourced answer without manually processing search results
  • Quick fact lookups where a synthesized answer is more useful than raw chunks
Use Ask when the user wants a direct answer. Use Search when you need the raw text chunks to do your own analysis or cross-referencing.

​
Parameters

ParameterTypeRequiredDescription
project_idstringyesProject ID to ask about
questionstringyesNatural language question (2-2000 chars)

​
Response

{
  "answer": "Door 104 has a 45-minute fire rating according to the Door Schedule on sheet A-501. It is a hollow metal door with hardware set HW-3.",
  "sources": [
    {
      "source_type": "block",
      "source_id": "blk_door_sched",
      "label": "Sheet A-501",
      "score": 0.94
    },
    {
      "source_type": "block",
      "source_id": "blk_a101_notes",
      "label": "Sheet A-101",
      "score": 0.87
    }
  ]
}
FieldTypeDescription
answerstringSynthesized answer (2-4 sentences)
sourcesarrayTop sources used to generate the answer
sources[].source_typestring"block", "feature", or "file"
sources[].source_idstringEntity ID for follow-up queries
sources[].labelstringHuman-readable source label (e.g., sheet number or file name)
sources[].scorenumberRelevance score

​
Examples

​
Ask about a specific feature

{
  "project_id": "prj_abc",
  "question": "What fire rating does door 104 have?"
}

​
Ask about project-wide specifications

{
  "project_id": "prj_abc",
  "question": "What concrete mix design is specified for the foundation?"
}
  • Search - Get raw search results instead of a synthesized answer
  • Query - Structured queries when you know the exact entity and filters