Logics Ollama Specialist

Platform scope:

• ollama_install_macos.sh is intentionally macOS-specific.
• ollama_check.sh is a POSIX-shell helper and should not be treated as a Windows workflow entrypoint.
• On Windows, follow the Ollama product's native install path first, then adapt the integration guidance in references/.

Normalize common user variants such as deepseeker-coder-v2, deepseek coder v2, or deepseek-coder to deepseek-coder-v2. Treat Qwen coder, qwen coder, or qwen-coder requests as a Qwen-family coding profile and confirm the exact tag before changing config.

Quick start

1. Clarify OS, target model, editor surface, and whether Ollama runs locally or on another host.
2. If installation is needed, ask for explicit permission before running install commands.
3. For local coding, prefer the repo's curated DeepSeek default profile deepseek-coder-v2:16b unless the user explicitly asks for another supported family or tag.
4. Verify Ollama, daemon reachability, and model availability before editing editor config.
5. If the app is a frontend, prefer a same-origin proxy (/ollama/*) to avoid CORS.
6. If LAN URL causes 403, remove Origin in the proxy or set OLLAMA_ORIGINS.