Guides › Using OpenCode Interrupt with Ollama
Using OpenCode Interrupt with Ollama
The interrupt plugin works with any model OpenCode supports, including local models via Ollama. The plugin operates at the harness level, not the model level.
1
Set up Ollama in OpenCode
Configure OpenCode to use your Ollama endpoint (typically http://localhost:11434). Select any Ollama model.
2
Add the interrupt plugin
Add "opencode-interrupt-plugin" to opencode.json. No Ollama-specific configuration needed.
3
Use voice features
All features work identically: /ptt walkie-talkie, streaming TTS, voice interruption. The plugin does not care which backend model is running.
4
Note on TTS
TTS uses edge-tts, not the Ollama model. Edge-tts handles text-to-speech independently. The plugin generates TTS from text chunks as they arrive from Ollama.
Tips
• Ollama models are typically slower than API models, which gives you more time to interrupt mid-generation.
• Whisper transcription is entirely local — your voice data never leaves your machine when using local STT + local LLM.
• The correction injection works with any Ollama model. The plugin adds system context about the interruption.
Troubleshooting
If TTS is choppy with slow Ollama models, the plugin buffers text into sentences. Try reducing EDGE_TTS_RATE to "+0%" for clearer audio with slower text generation.