Knowledge RAG

Knowledge RAG query (external MCP + local tenant index)

post/api/knowledge-rag/query

Request

  • Base URL: https://api.cernion.de
  • URL: https://api.cernion.de/api/knowledge-rag/query
  • Auth: none declared

Query parameters

tokenstring

Optional Cernion MCP token as URL query parameter. If provided, it overrides CERNION_TOKEN for this request. Prefer Authorization Bearer header in production to reduce token exposure in URLs/logs.

Request body

queryType'semantic' | 'scroll' | 'fetch' | 'collection_info'
querystring
limitnumber
scoreThresholdnumber
dedupeboolean
rerankboolean
diversityPerDocumentnumber
rerankWindownumber

Default: max(30, min(limit * 4, 50))

filterobject
withPayloadboolean
withVectorsboolean
tokenstring

Optional Cernion MCP token for this request. Overrides CERNION_TOKEN when provided.

Example request

{
  "queryType": "semantic",
  "query": "Welche Festlegungen der BNetzA gibt es zum Netzanschluss?",
  "limit": 5,
  "scoreThreshold": 0.4,
  "dedupe": true,
  "rerank": true,
  "diversityPerDocument": 2,
  "rerankWindow": 30,
  "ids": [
    "doc-123"
  ],
  "filter": {
    "must": [
      {
        "key": "metadata.docType",
        "match": {
          "value": "Festlegung"
        }
      },
      {
        "key": "metadata.authority",
        "match": {
          "value": "BNetzA"
        }
      }
    ]
  }
}

Response

RAG query result (internal/direct calls only; external callers receive 202)

successboolean

Example response

{
  "success": true,
  "data": {
    "queryType": "semantic",
    "collection": "cernion_knowledge_v1",
    "returned": 2,
    "results": [
      {
        "pointId": "doc-123",
        "score": 0.91,
        "referenceText_L0": "Die Festlegung regelt ...",
        "vectorText": "Welche Regeln gelten für ...",
        "metadata": {
          "title": "BK6-24-xxx",
          "docType": "Festlegung",
          "authority": "BNetzA",
          "status": "gültig"
        }
      }
    ]
  }
}

Changes

No recorded changes to this operation across all 2 revisions of this API.