Unable to generate files when using local Ollama

[StrongKelvin]

Question

OpenCode 1.15.12
I self-hosted Ollama locally using Docker and installed the gemma4 model. I also configured the Ollama provider in OpenCode under C:\Users<UserName>.config\opencode.
{ "$schema": "https://opencode.ai/config.json", "plugin": ["opencode-antigravity-auth@latest"], "provider": { "ollama": { "npm": "@ai-sdk/openai-compatible", "name": "Ollama (local)", "options": { "baseURL": "http://localhost:11434/v1" }, "models": { "gemma4:26b": { "name": "gemma4:26b" }, "deepseek-r1:8b": { "name": "deepseek-r1:8b" } } } } }

In OpenCode, I can successfully connect to the self-hosted Ollama and use the gemma4 model. However, there is one issue: in an OpenCode conversation, I want it to help me build a web page, but instead of generating files, it only displays the code in the chat window.

Am I missing some configuration, or is there an additional setting required?

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• Ollama (qwen2.5-coder): tool calls (edit/write) show as executed but no files are created/modified #7030: Ollama tool calls (edit/write) show as executed but no files are created/modified — same symptom of file operations not being applied to disk
• local model keeps returning json and doesn't create any files #18454: Local model keeps returning JSON and doesn't create any files — same core problem with Ollama + local models
• Opencode + Qwen gives JSON output instead of writing to disk #29757: OpenCode + Qwen gives JSON output instead of writing to disk — very similar issue with Ollama local provider and file generation

These issues share the same root cause: local models via Ollama often lack reliable tool-calling support, causing write/edit tool calls to be printed as JSON in the chat instead of being executed. The behavior you're seeing is a known limitation with some local models that don't properly support the function/tool calling protocol required by OpenCode.