Custom Provider Hangs After "build · model" in custom RuAPI provider

Description

Opencode hangs indefinitely after displaying "build · [model]" when using a custom provider with @ai-sdk/openai-compatible SDK. The API works correctly via curl but opencode never receives a response.

Environment

Opencode version: 1.4.7
OS: Linux (Arch-based)
Node/Bun version: (bun)

Steps to Reproduce

Add custom provider to ~/.config/opencode/opencode.jsonc:

{
  "model": "RuAPI/claude-opus-4.6",
  "provider": {
    "RuAPI": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "RuAPI",
      "options": {
        "baseURL": "https://api.stepanovikov.uno"
      }
    }
  }
}

Add API key: opencode providers login (or auth.json)
Run: opencode run "hi"
Result: Hangs at "build · claude-opus-4.6" forever

Expected Behavior

Should receive response from API within seconds.

Actual Behavior

Logs show provider loaded, SDK initialized
Logs show "small=false agent=build mode=primary stream"
Then hangs indefinitely - no timeout, no error

Additional Tests

curl to API endpoint works fine (both sync & streaming)
Built-in opencode/minimax-m2.5-free works correctly
Other models on same provider (gpt-5.4, claude-sonnet-4.6) also hang
Tried adding apiKey directly in config - same result

Logs

INFO  provider providerID=RuAPI found
INFO  provider providerID=RuAPI pkg=@ai-sdk/openai-compatible using bundled provider
INFO  provider status=completed duration=4 providerID=RuAPI getSDK
INFO  llm providerID=RuAPI modelID=claude-opus-4.6 ... small=false agent=build mode=primary stream
INFO  bus type=session.status publishing
INFO  session.updated publishing
[then hangs forever]

Workaround

Have to use built-in free model: `opencode run "hi" -m opencode/minimax-m2.5-free" for my tasks

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

OpenCode 1.4.3 hangs with local Ollama provider on simple prompts, while /v1/chat/completions works #22132: Same symptom — OpenCode hangs indefinitely after "build · [model]" when using a custom @ai-sdk/openai-compatible provider, while direct API calls (curl) work fine. The log pattern is identical.
Opencode hangs when communicating with venice #11056: OpenCode hangs with a custom provider (venice); a workaround involving bunfig.toml disabling the Bun install cache was found to resolve it.

Please check if either of those resolves your issue, and consider adding a comment there instead.

Also asked AI to figure something out and got this:

The Problem

Opencode hangs at "build · [model]" when using @ai-sdk/openai-compatible custom providers.
The API works via curl, but opencode never receives/processes the response.

Confirmed Working

✅ curl to API endpoint
✅ SDK standalone test (non-streaming)
✅ opencode built-in models (opencode/minimax-m2.5-free)

Not Working

❌ opencode with @ai-sdk/openai-compatible (any model, any provider)
❌ All 3 reported issues (Custom Provider Hangs After "build · model" in custom RuAPI provider #23094, OpenCode 1.4.3 hangs with local Ollama provider on simple prompts, while /v1/chat/completions works #22132, Opencode hangs when communicating with venice #11056) have same symptoms

Root Cause Hypothesis

The issue is NOT in:

Network connectivity (curl works)
API authentication (standalone SDK works)
Model configuration (works with built-in models)
The issue IS in:
How opencode processes the streaming response from @ai-sdk/openai-compatible
OR how the SDK handles the response in the opencode context

Evidence

Logs show: llm providerID=RuAPI ... stream
Then: session.updated publishing
Then: HANGS - no error, no timeout, no response

Things That Didn't Fix It

Adding apiKey to config
Adding timeout to config
Adding bunfig.toml (workaround for different issue)
Different SDK versions (1.0.0, 2.0.41)
Different models (claude-opus, gpt-5.4, claude-sonnet)
Different provider options

What Would Help Debug

Network trace showing if request is made and if response is received
Comparing working builtin model vs broken custom provider at code level
SDK version compatibility with opencode version

Related Issues (all same root cause - likely)

Custom Provider Hangs After "build · model" in custom RuAPI provider #23094 - RuAPI provider (my case)
OpenCode 1.4.3 hangs with local Ollama provider on simple prompts, while /v1/chat/completions works #22132 - Ollama provider
Opencode hangs when communicating with venice #11056 - Venice provider

For Developers

The fix needs to be in opencode's LLM response handling for
@ai-sdk/openai-compatible providers. The SDK itself works fine.

To stay organized issues are automatically closed after 60 days of no activity. If the issue is still relevant please open a new one.

Navigation Menu

Repository navigation

Custom Provider Hangs After "build · model" in custom RuAPI provider #23094

Description

Description

Environment

Steps to Reproduce

Expected Behavior

Actual Behavior

Additional Tests

Logs

Workaround

The Problem

Confirmed Working

Not Working

Root Cause Hypothesis

Evidence

Things That Didn't Fix It

What Would Help Debug

Related Issues (all same root cause - likely)

For Developers

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Custom Provider Hangs After "build · model" in custom RuAPI provider #23094

Description

Description

Description

Environment

Steps to Reproduce

Expected Behavior

Actual Behavior

Additional Tests

Logs

Workaround

Activity

github-actions commented on Apr 17, 2026

fadingawayzy commented on Apr 17, 2026

The Problem

Confirmed Working

Not Working

Root Cause Hypothesis

Evidence

Things That Didn't Fix It

What Would Help Debug

Related Issues (all same root cause - likely)

For Developers

github-actions commented on Jul 6, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions