🧠 Feature Request: Ollama Integration with LangChain + API Routes

[abhishek-nexgen-dev]

Implement backend support for Ollama model usage through LangChain, including key features like model verification, chat functionality, and local model status. This allows users to interact with local models securely and only when they’re installed.

---

🔧 Key Features

✅ 1. LangChain + Ollama Integration

• Use LangChain’s Ollama integration to interact with local LLMs.
• Allow any installed Ollama model to be used for chat (e.g., llama3, mistral, etc.)

✅ 2. Check if Ollama Model is Installed

Before processing a chat request:

• Check if the requested Ollama model exists on the local system.
• If not found, return:

{
  "error": true,
  "message": "The Ollama model is not installed on your local system. Please install it before using."
}

Use Ollama's API (GET /api/tags) to check for installed models.

✅ 3. Secure Access via Local Mind API Key

All routes must require the following HTTP header:

local-mind-api-key: <your-local-key>

• Validate this against known or configured secure keys.

• Deny access if missing or invalid.

Ollama API Routes

Chart with Ollama

• Chat with a locally installed Ollama model using LangChain.

Headers:

x-mind-api-key: <your-local-mind-api-key>
Content-Type: application/json

Request Body

{
  "model": "llama3",
  "messages": [
    { "role": "user", "content": "Hello, who are you?" }
  ]
}

Success Response:

{
  "model": "llama3",
  "response": "Hello! I'm a helpful assistant running locally.",
  "timestamp": "2025-10-12T15:30:00Z"
}

Fetch Ollama models

Fetch a list of all locally installed Ollama models.

Headers:

local-mind-api-key: <your-local-key>

Response:

{

  "models": [
    "llama3",
    "mistral",
    "codellama:7b"
  ]
}

Check if model is Install or not

Check if a specific Ollama model is installed locally before allowing usage.

🔐 Headers

x-mind-api-key: <your-local-mind-api-key>   // Required

🌐 Endpoint

GET /api/ollama/check/:modelName

Success Response (if installed):

{
  "model": "llama3",
  "installed": true
}