Skip to main content
DocsGuidesRAG
GuideGrounded Routes

RAG with the current DeepAILab routes

This guide documents the safest RAG surface to use today: the LangChain-backed routes at/api/langchain/rag/*. They are implemented in the live service and are the same routes used by the frontend client in the current codebase.

Add documents as structured content with metadata
Query with question, top_k, score_threshold, and projectId
JWT + tenant-aware retrieval through X-User-ID
Keeps the docs aligned with the routes actually used by the product

Overview

DeepAILab currently exposes two RAG-shaped surfaces in the repository: a LangChain-backed route used by the web product and a gateway proxy for the standalone RAG service. This guide intentionally stays on the LangChain-backed surface because it is grounded in current application code and has a clear request schema in the controller.

export DEEPAILAB_BASE_URL="https://api.deepailab.ai"
export DEEPAILAB_SESSION_TOKEN="your-session-jwt"
export DEEPAILAB_USER_ID="user_123"

Quick start

1. Add documents to the user vector store

curl -X POST "$DEEPAILAB_BASE_URL/api/langchain/rag/add-documents" \
  -H "Authorization: Bearer $DEEPAILAB_SESSION_TOKEN" \
  -H "Content-Type: application/json" \
  -H "X-User-ID: $DEEPAILAB_USER_ID" \
  -d '{
    "documents": [
      {
        "content": "DeepAILab requires Node.js 22 or newer for local development.",
        "metadata": {
          "source": "setup-guide",
          "section": "requirements"
        }
      },
      {
        "content": "Recommended profile installs MCP, skills, and agent-team by default.",
        "metadata": {
          "source": "devtools-guide",
          "section": "profiles"
        }
      }
    ]
  }'

2. Query with grounded retrieval

curl -X POST "$DEEPAILAB_BASE_URL/api/langchain/rag/query" \
  -H "Authorization: Bearer $DEEPAILAB_SESSION_TOKEN" \
  -H "Content-Type: application/json" \
  -H "X-User-ID: $DEEPAILAB_USER_ID" \
  -d '{
    "question": "What profile should I choose for a new engineering team?",
    "top_k": 5,
    "score_threshold": 0.2,
    "projectId": "docs-demo"
  }'

API examples

Use the REST route directly or call it from your own application. The important detail is the request shape: question plus optional retrieval controls, not the older query field shown in historical docs.

cURL

curl -X POST "$DEEPAILAB_BASE_URL/api/langchain/rag/query" \
  -H "Authorization: Bearer $DEEPAILAB_SESSION_TOKEN" \
  -H "Content-Type: application/json" \
  -H "X-User-ID: $DEEPAILAB_USER_ID" \
  -d '{
    "question": "What profile should I choose for a new engineering team?",
    "top_k": 5,
    "score_threshold": 0.2,
    "projectId": "docs-demo"
  }'

JavaScript / TypeScript

const baseUrl = process.env.DEEPAILAB_BASE_URL ?? "https://api.deepailab.ai";
const sessionToken = process.env.DEEPAILAB_SESSION_TOKEN!;
const userId = process.env.DEEPAILAB_USER_ID!;

const response = await fetch(`${baseUrl}/api/langchain/rag/query`, {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    Authorization: `Bearer ${sessionToken}`,
    "X-User-ID": userId,
  },
  body: JSON.stringify({
    question: "Summarize the recommended developer-tool install profile.",
    top_k: 4,
    score_threshold: 0.15,
    projectId: "docs-demo",
  }),
});

const result = await response.json();
console.log(result);

Best practices

Use the LangChain RAG routes documented here

These routes are implemented in langchain-service and are the same ones used by the frontend AI client today.

Send question, not query

The current RAG query controller expects a question field with optional top_k, score_threshold, and projectId.

Always include bearer auth and X-User-ID

The documented routes are JWT-protected and use the user identity for tenant isolation during retrieval.

Treat /api/v1/rag/* as deployment-specific

The API gateway still proxies /api/v1/rag/* to the standalone RAG service, but the collection/document lifecycle depends on how that deployment is wired. This guide stays on the safer LangChain-backed surface.

Continue with the API reference or review the install center if you are also configuring local coding tools against the same DeepAILab gateway.

API ReferenceInstall Center