Ollama

Run AI agents with local models using Ollama. No API key needed — everything runs on your machine.

Setup

1. Install Ollama: ollama.com/download
2. Pull a model:

ollama pull llama3.2

3. Ollama starts automatically and serves on localhost:11434. Configure in droid.yaml:

ollama:
  base_url: "http://localhost:11434/v1"

No API key required.

Agent Config

name: "local-agent"
model:
  provider: "ollama"
  name: "llama3.2"
  max_tokens: 2048
  temperature: 0.7

Popular Models

Model	Size	Best for
llama3.2	3B	Fast, general purpose
llama3.1:8b	8B	Good quality/speed balance
llama3.1:70b	70B	High quality (needs GPU)
mistral	7B	Fast, multilingual
codellama	7-34B	Code generation
qwen2.5	7B	Multilingual

See ollama.com/library for all available models.

Docker Networking

If Sinaptic® DROID+ runs in Docker and Ollama runs on the host:

ollama:
  base_url: "http://host.docker.internal:11434/v1"

If both run in Docker, use Docker Compose networking:

services:
  droid:
    image: sinapticai/droid:latest
    # ...
  ollama:
    image: ollama/ollama:latest
    volumes:
      - ollama-data:/root/.ollama

Then use http://ollama:11434/v1 as the base URL.

Notes

• Ollama serves an OpenAI-compatible API, so it integrates seamlessly with Sinaptic® DROID+.
• Tool use (function calling) support depends on the model. Llama 3.1+ and Mistral support function calling.
• Performance depends entirely on your hardware. GPU is recommended for models larger than 7B.
• Models are downloaded once and cached locally by Ollama.