Answer (RAG)

curl -X POST "http://localhost:8080/answer" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Who won the football match?",
    "query_vector": [0.01, -0.02, "... 384 floats ..."],
    "top_k": 5
  }'

{
  "answer": "Manchester United beat Chelsea 2-1.",
  "model": "qwen2.5:0.5b-instruct",
  "query": "Who won the football match?",
  "context_count": 1,
  "sources": [
    {
      "id": "doc-1",
      "score": 0.894123,
      "label": "Close Match",
      "text": "Manchester United beat Chelsea 2-1 in the Premier League on Saturday."
    }
  ]
}

POST

answer

curl -X POST "http://localhost:8080/answer" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Who won the football match?",
    "query_vector": [0.01, -0.02, "... 384 floats ..."],
    "top_k": 5
  }'

{
  "answer": "Manchester United beat Chelsea 2-1.",
  "model": "qwen2.5:0.5b-instruct",
  "query": "Who won the football match?",
  "context_count": 1,
  "sources": [
    {
      "id": "doc-1",
      "score": 0.894123,
      "label": "Close Match",
      "text": "Manchester United beat Chelsea 2-1 in the Premier League on Saturday."
    }
  ]
}

Overview

Run retrieval-augmented generation (RAG) on the local store:

Search using the provided query_vector (same dimension as the store).
Build a prompt from the top matching passages.
Call Ollama on the host with fixed model qwen2.5:0.5b-instruct.

If no passages pass search (empty store or nothing above threshold in kiosk mode), the server does not call the LLM and returns: I don't have enough information to answer that question.

Requires Ollama running on the host. Start the stack with moorcheh-edge up --with-llm (use --skip-ollama for search-only). The CLI and SDK embed query locally before calling this endpoint.

Request body

string

required

Original question text (included in the LLM prompt and echoed in the response).

array

required

JSON array of floats used for similarity search. Length must match the store dimension (384 for text stores).

number

default:"5"

Number of passages to retrieve for context. Capped at 100.

number

default:"0"

Minimum search score when kiosk_mode is true.

boolean

default:"false"

When true, filters retrieved passages below threshold.

string

Optional system instruction (replaces the default RAG system prompt).

string

Optional instruction appended before the user question in the final user message.

array

Prior turns: [{"role": "user"|"assistant", "content": "..."}].

number

default:"0.2"

LLM sampling temperature (0.0–2.0).

curl -X POST "http://localhost:8080/answer" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Who won the football match?",
    "query_vector": [0.01, -0.02, "... 384 floats ..."],
    "top_k": 5
  }'

{
  "answer": "Manchester United beat Chelsea 2-1.",
  "model": "qwen2.5:0.5b-instruct",
  "query": "Who won the football match?",
  "context_count": 1,
  "sources": [
    {
      "id": "doc-1",
      "score": 0.894123,
      "label": "Close Match",
      "text": "Manchester United beat Chelsea 2-1 in the Premier League on Saturday."
    }
  ]
}

Response fields

Field	Description
`answer`	Generated answer text
`model`	LLM model id (`qwen2.5:0.5b-instruct`)
`query`	Echo of the request question
`context_count`	Number of passages passed to the LLM
`sources`	Search hits used as context (same shape as `/search` results)

Errors

Condition	Status	Message (example)
LLM not configured	`400`	`LLM is not configured: start Ollama on the host and run moorcheh-edge up`
LLM unreachable or error	`400`	`LLM request failed`
Empty store / no matches	`200`	Fixed answer: `I don't have enough information to answer that question.` (`context_count`: 0, LLM not called)

System

Search

AI Generation

Data Operations

Delete

Overview

Request body

Response fields

Errors

​Overview

​Request body

​Response fields

​Errors

​Related

Overview

Request body

Response fields

Errors

Related