Description
The OpenCode interface displays the model's internal "Thinking Process" but fails to render the actual final response text. This creates a loop where the AI appears to "think" about providing an answer but the answer is never displayed.
Environment
- Platform: OpenCode (Linux)
- Date: April 2026
- Interface: Terminal interface
Steps to Reproduce
- Send a query to the AI model
- Wait for response
- Observe - only thinking process is visible, no final response appears
Expected Behavior
- Both the thinking process AND the final response should be visible
- OR: Only the final response should be shown (thinking process is internal)
Actual Behavior
- Only the thinking process block is displayed
- Example: "Let me provide a fresh, clear answer now." appears in thinking, but actual answer text never renders
Workarounds Tested
- Refreshing the session may temporarily fix it
- Very short responses (1-2 words) sometimes appear
- Responses without newlines may appear more reliably
Additional Notes
- The AI model believes it has sent the answer (its thinking confirms this)
- The issue appears to be in how the text block following the thinking block is parsed/rendered
- This is reproducible and happens consistently for certain types of responses
Severity
High - Makes the tool unusable for most tasks
Description
The OpenCode interface displays the model's internal "Thinking Process" but fails to render the actual final response text. This creates a loop where the AI appears to "think" about providing an answer but the answer is never displayed.
Environment
Steps to Reproduce
Expected Behavior
Actual Behavior
Workarounds Tested
Additional Notes
Severity
High - Makes the tool unusable for most tasks