Vea - Local AI Copilot 🤖

Learn more about Ollama at https://ollama.com/ for deploying LLMs locally

Table of Contents

• Overview
• Features
• Prerequisites
• Installation
• Configuration
• Usage
• Project Structure
• Logging
• Limitations
• License

Overview

Vea is a local AI copilot that seamlessly integrates with your Ollama installations. It provides a modern, web-based interface for interacting with local AI models while leveraging additional capabilities such as web search, mathematical operations, weather information, and image analysis. You can switch to closed-source models like OpenAI and Anthropic by editing the configuration file at backend/config/agent.yaml or through the web-based configuration interface. Please complete the setup on the configuration page (accessible at http://localhost:3000/configure) before submitting your first query.

Features

• Configurable Tools - Enable/disable web search, weather, and math tools based on your needs
• Web search capabilities powered by Tavily API
• Real-time weather information using OpenWeather API
• Mathematical calculations including arithmetic and trigonometric functions
• Multi-modal vision capabilities
• Context-aware responses with current date/time integration
• Modular architecture with LangGraph support and LangSmith observability
• Support markdown for programming languages
• Visible thinking traces when using a reasoning model
• Web-based configuration interface for model and tool selection

Tech Stack

• Backend: FastAPI, LangGraph, LangSmith, and Ollama
• Frontend: React, TypeScript, TailwindCSS, and Vite

Prerequisites

To install and run Vea, ensure you have the following:

• Tavily API key
• OpenWeather API key
• Ollama installed and running

Installation

1. Clone the repository:

git clone https://github.com/therealcyberlord/Vea.git
cd Vea

2. Install Python dependencies using UV:

cd backend
uv sync
source .venv/bin/activate

3. Install frontend dependencies:

cd frontend
npm install

Configuration

Create a .env file in the backend directory with the following configuration:

TAVILY_API_KEY=your_tavily_api_key
OPENWEATHER_API_KEY=your_openweather_api_key
LANGSMITH_API_KEY=your_langsmith_api_key
LANGSMITH_TRACE=true

You can obtain API keys from:

• Tavily API: https://tavily.com/
• OpenWeather API: https://openweathermap.org/api
• LangSmith: https://www.langchain.com/langsmith

Tools Configuration

Vea supports configurable tools that can be enabled or disabled based on your needs. The backend/config/agent.yaml file contains the following configuration:

llm_config:
  tool_llm:
    name: qwen3:4b
    provider: ollama
    temperature: 0.3
  vision_llm:
    name: gemma3:12b
    provider: ollama
    temperature: 0.2
tools:
  web_search: false
  weather: true
  math: false

• web_search: Enable/disable web search capabilities (requires Tavily API key)
• weather: Enable/disable weather information lookup (requires OpenWeather API key)
• math: Enable/disable mathematical calculation tools

You can also configure your models through the web interface at http://localhost:3000/configure.

Usage

1. Start the FastAPI backend server:

cd backend
uvicorn main:app --reload

2. Start the frontend development server:

cd frontend
npm run dev

The application will be accessible at http://localhost:3000

Project Structure

Vea/
├── backend/           # Backend implementation
│   ├── agent/         # Main AI agent
│   ├── tools/         # Tools for function-calling
│   ├── config/        # Configuration files
│   ├── utils/         # Utility functions
│   ├── models/        # Pydantic data models
│   └── .env           # Environment variables
├── frontend/          # Frontend application
│   ├── src/           # TypeScript source code
│   ├── public/        # Static assets
│   └── vite.config.ts # Vite configuration

Screenshots

Here's a look at Vea's chat interface:

The configuration page allows you to customize your models:

Vea supports image inputs:

It also supports coding examples with markdown:

You can view the AI's thinking process if you are using a reasoning model:

Logging

Vea implements comprehensive logging for both development and production environments:

• Console Output: INFO level and above messages are displayed in the console
• File Logging: Detailed DEBUG level logs are written to app.log with rotation
• Log Rotation: Log files are automatically rotated when they reach 10MB, with up to 5 backup files
• Structured Format: All logs follow the format: timestamp - logger_name - level - message

The logging configuration can be customized by modifying backend/config/logging.conf.

Limitations

• Memory persistence is stored in-memory and is not persisted to an external database
• Vision capabilities depend on the underlying model's capabilities
• Temperature configuration functionality is currently a work-in-progress
• Better handling of long-term memory

Contributing

Feel free to contribute to Vea! Whether it's adding new features, improving existing ones, or fixing bugs, your contributions are welcome. Please feel free to submit a pull request or open an issue for any enhancements you'd like to see.

License

This project is licensed under the MIT License - see the LICENSE file for details.