A lightweight local chatbot that sends prompts directly to Ollama.

• Direct model chat (no retrieval, no vector database)
• Web UI for chatting
• Conversation history: create, resume, rename, delete, export
• Per-conversation context profile overrides
• CLI chat mode

• Ollama running locally
• Python 3.8+

python3 -m venv env
source env/bin/activate
pip install -r requirements.txt

cp .env.example .env
ollama pull llama3.1:8b

python scripts/chat.py --question "Hello"
python scripts/chat.py

python scripts/server.py

Open: http://localhost:8000

• Docker + Docker Compose
• Ollama running on your host machine

If your app runs in Docker and Ollama runs on your host, set:

OLLAMA_BASE_URL=http://host.docker.internal:11434
GENERATION_MODEL=llama3.1:8b
GENERATION_TEMPERATURE=0.2
GENERATION_TOP_P=0.9

docker compose up --build

Open: http://localhost:8000

docker compose down

Notes:

• ./data and ./config are mounted into the container, so conversations/profiles persist.
• On Linux, host.docker.internal is mapped via extra_hosts in docker-compose.yml.

Edit .env:

OLLAMA_BASE_URL=http://localhost:11434
GENERATION_MODEL=llama3.1:8b
GENERATION_TEMPERATURE=0.2
GENERATION_TOP_P=0.9

Profiles are in config/chat_profiles.json. Each profile supports:

• system_prompt
• answer_instruction
• history_turns

• GET / - Web UI
• POST /query - Ask a question
• GET /chat/profiles - List profiles
• POST /conversations - Create conversation
• GET /conversations - List conversations
• GET /conversations/{conversation_id} - Load conversation
• PATCH /conversations/{conversation_id}/context - Update profile/history/prompt overrides
• PATCH /conversations/{conversation_id} - Rename conversation
• DELETE /conversations/{conversation_id} - Delete conversation
• GET /conversations/{conversation_id}/export - Export conversation JSON
• GET /health - Health check