A web interface for local AI models

A stylish web-based chat interface for Ollama with integrated web search.

Demo

Chat

Web search

Architecture

The browser loads static HTML/CSS/JS files from a Python HTTP server which talks with the Ollama API. Ollama forwards messages to a locally-running LLM, which generates tokens which stream back through the chain to appear in the browser. A seperate Python backend server enables API calls for features like web search.

Minimal example

A bare-bones example is provided in minimal-example.

Quick start

1. Install Ollama and download a model

First, you need to install Ollama and download at least one language model:

# Install Ollama (Linux/macOS)
curl -fsSL https://ollama.ai/install.sh | sh

# Download a model (browse https://ollama.ai/search)
ollama pull llama3.1:latest

# Verify installation
ollama list

Note: Ollama must be running for the chat interface to work. The application defaults to using llama3.1:latest but you can change this in the configuration.

2. Start services

Quick start (recommended):

./start-all.sh  # Sets up environment and starts all services

Manual start (more fun):

Note: Each command needs to run continuously. Use separate terminals, use detached sessions with screen or similar, or run with & to background (as shown).

# Setup logs directory
mkdir logs

# Ollama (if not already running)
OLLAMA_HOST=0.0.0.0:11434 ollama serve > logs/ollama.log 2>&1 &

# Web server
cd frontend && python3 -m http.server 8000 --bind 0.0.0.0 > ../logs/http-server.log 2>&1 &

# ---------------------------------------
# Optional Python backend for web search:
# ---------------------------------------

# Create Python & activate a virtual environment 
python3 -m venv venv
source venv/bin/activate

# Upgrade pip & install dependencies 
pip install --upgrade pip
pip install -r backend/requirements.txt

# Start the search proxy
python3 backend/search-proxy.py > logs/search-proxy.log 2>&1 &

3. Open chat

Navigate to:

http://localhost:8000/chat.html

or

http://<LOCAL_IP>:8000/chat.html

if accessing from another device on your network where LOCAL_IP is the IP address of the device running ollama-chat.

4. Stop services

Quick stop (recommended):

./stop-all.sh  # Will prompt you to stop Ollama

Manual stop:

# Stop HTTP server
pkill -f "python.*http.server 8000"

# Stop search proxy
pkill -f "search-proxy.py"

# Stop Ollama
pkill ollama