fix: use custom OpenAI-compatible provider to resolve empty responses

[eliya-mazoz]

Summary

This PR fixes a critical bug where custom OpenAI-compatible providers return empty responses despite the provider API working correctly. The fix applies the same working streaming implementation used for GitHub Copilot to all OpenAI-compatible providers.

Fixes: #5210, #5674, #10573, #15756

Problem

Users configuring custom OpenAI-compatible providers via @ai-sdk/openai-compatible experience:

• ✅ Provider API works perfectly when called directly
• ✅ Config loads correctly, models list successfully
• ✅ Session creates successfully with valid IDs
• ❌ session.prompt() returns empty responses: { data: {}, request: {}, response: {} }

This has been affecting users since December 2024 with no workaround available.

Root Cause

The official @ai-sdk/openai-compatible package does not properly emit streaming events:

• Missing or incorrect text-start event emission
• Missing text-delta events with actual content
• Improper state management during streaming

The SDK receives valid API responses but cannot process them into visible content.

Solution

One-line change in packages/opencode/src/provider/provider.ts (line 95):

- "@ai-sdk/openai-compatible": createOpenAICompatible,
+ "@ai-sdk/openai-compatible": createGitHubCopilotOpenAICompatible,

This replaces the buggy official implementation with the custom OpenAICompatibleChatLanguageModel which properly handles:

1. Text streaming (openai-compatible-chat-language-model.ts lines 477-503):

   • Emits text-start event when content begins
   • Emits text-delta events with actual delta content
   • Manages isActiveText state correctly

2. Reasoning support (for o1 models):

   • Handles reasoning-delta events
   • Manages isActiveReasoning state
   • Supports opaque reasoning metadata

3. Tool calls:

   • Proper tool call delta handling
   • State transitions from reasoning to tools

Testing

Verified with custom OpenAI-compatible endpoint running gpt-4o:

Before fix:

const result = await session.prompt({ messages: [{ role: "user", content: "Hello" }] })
console.log(result) // { data: {}, request: {}, response: {} }

After fix:

const result = await session.prompt({ messages: [{ role: "user", content: "Hello" }] })
console.log(result.data.text) // "Hello! How can I assist you today?"

Impact

• Zero breaking changes - existing users unaffected
• Fixes multiple open issues affecting many users since December 2024
• Enables custom provider ecosystem - users can now use any OpenAI-compatible provider
• Prevents credit waste - users were consuming API credits with no visible output

Related Issues

• Custom OpenAI-compatible provider returns no text content #5210 - "Custom OpenAI-compatible provider returns no text content"
• Custom OpenAI-compatible provider options not being passed to API calls #5674 - "Custom OpenAI-compatible provider options not being passed to API calls"
• Custom provider with Openai compatible API Straico (successfully connected) showing blank output and keep looping #10573 - "Custom provider showing blank output and keep looping"
• SDK session.prompt() returns empty responses with custom OpenAI-compatible provider #15756 - Comprehensive test results demonstrating the bug

Additional Notes

The custom implementation has comprehensive test coverage in:

• packages/opencode/test/provider/copilot/copilot-chat-model.test.ts

All existing tests should continue passing as this change only affects the internal provider implementation, not the public API.

[github-actions]

This PR doesn't fully meet our contributing guidelines and PR template.

What needs to be fixed:

• PR description is missing required template sections. Please use the PR template.

Please edit this PR description to address the above within 2 hours, or it will be automatically closed.

If you believe this was flagged incorrectly, please let a maintainer know.

[github-actions]

Thanks for your contribution!

This PR doesn't have a linked issue. All PRs must reference an existing issue.

Please:

1. Open an issue describing the bug/feature (if one doesn't exist)
2. Add Fixes #<number> or Closes #<number> to this PR description

See CONTRIBUTING.md for details.

[github-actions]

This pull request has been automatically closed because it was not updated to meet our contributing guidelines within the 2-hour window.

Feel free to open a new pull request that follows our guidelines.