Use Cases › Debugging with Voice in OpenCode

🐛

Debugging with Voice in OpenCode

The Problem

Debugging requires rapid iteration: form a hypothesis, test it, check results, adjust. Typing this loop slows you down.

The Solution

Voice lets you describe bugs naturally. Walkie-talkie mode captures your spoken hypotheses, TTS reads AI analysis, and interruption lets you pivot as new information emerges.

Workflow

  1. Paste the error or problematic code into context.
  2. Type /ptt and speak the symptom: "I'm getting a NullPointerException on line 42 when the config is empty."
  3. Send and hear the AI's analysis via TTS.
  4. As the AI hypothesizes, interrupt when you think of something: "I already checked that — the config loader returns null."
  5. The AI pivots with your correction and continues debugging.
  6. Iterate rapidly — each /ptt or voice interruption is a new debugging cycle.
Tips
• Debugging benefits from OpenAI API transcription (set OPENAI_API_KEY) — it handles code terms like "NullPointerException" better.
• Use /tts-off if you prefer reading the AI's analysis, then /tts-on again for the next round.
• The timing window (Pro: timingWindowMs) is your debugging window — set it longer (10s+) for complex debugging sessions.

Want to try this workflow?

Install interrupt →