🔍Verity

Verity is a Perplexity-style AI search and answer engine that runs fully locally on AI PCs.
It combines SearXNG-powered search, retrieval, and local LLM reasoning to generate grounded, verifiable answers — without relying on cloud-based LLM providers.

• Fully Local, AI PC Ready - Optimized for Intel AI PCs using OpenVINO (CPU / iGPU / NPU), Ollama (CPU / CUDA / Metal)
• No need of any paid APIs
• Privacy by Design - Search and inference can be fully self-hosted
• SearXNG-Powered Search - Self-hosted, privacy-friendly meta search engine
• Designed for fact-grounded, concise answers
• OpenVINO,Ollama models,llama.cpp server or any OpenAI compatible LLM servers supported
• Modular architecture
• CLI and WebUI support
• API server support
• Powered by Jan-nano 4B model,or configure any model

Supported LLM runtimes:

• OpenVINO - CPU / iGPU / NPU
• Ollama - CPU / CUDA / Apple Metal
• llama.cpp server or any OpenAI compatible LLM server

• Python 3.10 or higher
• uv - fast Python package and project manager
• Node.js

Follow the steps to install and run on Windows(Automated).

• Step 1: Install SearXNG by following this guide
• Step 2: Clone/Download verity repo
• Step 3: Double click install.bat to install

To run CLI interactive app double click start.bat. To run WebUI start the webserver by double clicking start-webserver.bat and start frontend by double clicking start-webui.bat. OpenVINO models will be downloaded in the first run.

Ensure that you have Python 3.x and Node.js installed.

• Install SearXNG by following this guide
• Clone Verity repo, enter into verity dirctory
• Create a virtual environment and activate it
• Run the below commands :

pip install -r requirements.txt
playwright install
cd src/frontend
npm install

For backend configurations create a .env file refer the .env.sample file.

To run interactive CLI app run the below python src/app.py

First run the API server python src/api_server.py

Start the WebUI

cd src/frontend
npm run dev

We have tested the following models.

Tested using Intel AI PC with Intel Core Ultra Series 1 processor with CPU/GPU/NPU.

Run the llama.cpp server:

llama-server.exe -m Jan-nano-Q4_K_M.gguf -c 4096 --port 9000

If you are changing port configure in the .env file

OPENAI_LLM_BASE_URL=http://localhost:8000

❗ You can use any OpenAI compatible LLM servers with Verity.

Use the below config in .env file. update LLM provider(LLM_PROVIDER) as ollama and use model(LLM_MODEL_PATH) as mannix/jan-nano:latest.

Run the below command to pull the model.

ollama pull mannix/jan-nano:latest

To build and start the services: Configurations in .env_docker file

docker compose up --build

To open verity webui http://localhost:5000/ You can use llama.cpp server, ollama or any OpenAI compatible LLM server.

Verity API server exposes one API endpoint,it streams model responses using Server-Sent Events (SSE). http://localhost:8000/

GET /ask?question=<query>

text/event-stream

Follow the steps to run Verity CLI

• Step 1: Install SearXNG by following this guide

• Step 2: Build and run verity CLI tool,in the terminal enter into verity/src/cli folder and run the below command :

  go build

• Step 3:Install and run the Verity Web server by double clicking install.bat & followed by start-webserver.bat.

• Step 4: Verity CLI tool can be used as : verity.exe "your question"

To use MCP server

• Install and run start-mcpserver.bat

Add the following configuration :

{
  "mcpServers": {
    "verity": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "http://127.0.0.1:8000/mcp"
      ],
      "timeout": 120000
    }
  }
}