Skip to main content
The knowledge base is the retrieval backbone of SoftArchitect AI. It stores architecture patterns, tech-stack guides, and best-practice documents as vector embeddings in ChromaDB. The chat endpoints automatically query this base on every request — the endpoints below expose direct access.
The knowledge base endpoints are under active development. The search endpoint currently returns a placeholder response. Ingest and status endpoints are planned for a future phase.

​
GET /api/v1/knowledge/search

Search the knowledge base using a query string. Returns relevant documents ranked by semantic similarity.
This endpoint is not yet implemented. It returns a placeholder response and will be fully built in Phase 2.

​
Current response

{
  "message": "Knowledge search not yet implemented"
}

​
Planned request parameters

The following query parameters are planned for the Phase 2 implementation:
​
q
string
required
The search query string. Will be embedded and used for semantic similarity search against the ChromaDB collection.
​
limit
integer
default:"5"
Maximum number of documents to return.

​
Planned response shape

Based on the existing ChromaDB integration, the completed endpoint is expected to return: 
{
  "status": "ready",
  "documents": 29,
  "vectors": 934,
  "collections": ["softarchitect"],
  "results": [
    {
      "id": "doc_abc123",
      "content": "...",
      "metadata": {},
      "score": 0.87
    }
  ]
}

​
Example

curl "http://localhost:8000/api/v1/knowledge/search?q=microservices+architecture"

​
Upcoming endpoints

The following endpoints are planned for Phase 2 and are not yet available:
MethodPathDescription
POST/api/v1/knowledge/ingestIngest documents into the global knowledge base
GET/api/v1/knowledge/statusReturn collection stats (document count, vector count, status)
The /knowledge/status endpoint will return a response in this format: 
{
  "status": "ready",
  "documents": 29,
  "vectors": 934,
  "collections": ["softarchitect"]
}
To ingest documents into a per-project vector store right now, use the projects ingest endpoint.

Build docs developers (and LLMs) love

Get started for freeTalk to us