qqqa v0.9.0 - Local LLMs with Ollama

What's Changed

• Local LLM support: qa --init now lets you provision an Ollama provider out of the box. The agent
  keeps the OpenAI-compatible interface but skips cloud key requirements so you can run llama3.1 (or
  any Ollama model) on localhost + change port.
• Streamlined config: setting api_key: "local" (or exporting OLLAMA_API_KEY=local) avoids empty
  Authorization headers, which keeps middleware and reverse proxies happy even when running fully
  offline.