Bug Description
When logging into a provider (e.g. forge provider login google_ai_studio), the login itself succeeds and prints "configured successfully", but then an unrelated error from a different configured provider is shown. For example, if Bedrock has stale credentials, you get:
$ forge provider login google_ai_studio
? Enter your GoogleAIStudio API key: AIza...
● [12:11:25] GoogleAIStudio configured successfully
● [12:11:25] ERROR: Bearer token is required in API key field
The "Bearer token is required" error comes from the Bedrock provider, not from Google AI Studio.
Root Cause
After a successful login, finalize_provider_activation() calls get_all_provider_models() to check model compatibility. That function fetches models from all configured providers concurrently, then collects results with:
futures::future::join_all(futures)
.await
.into_iter()
.collect::<anyhow::Result<Vec<_>>>()This .collect::<Result<Vec<_>>>() fails on the first error from any provider, discarding all successful results.
The function's own docstring already describes the intended behavior:
Returns a list of ProviderModels for each configured provider that successfully returned models. If every configured provider fails (e.g. due to an invalid API key), the first error encountered is returned so the caller receives the real underlying cause rather than an empty list.
But the implementation doesn't match. One stale provider blocks the entire model-fetching flow for all providers.
Steps to Reproduce
- Configure two providers (e.g. Bedrock and Google AI Studio)
- Let one provider's credentials become stale or invalid
- Run
forge provider login <other_provider> with valid credentials
- Observe the error from the stale provider after the login succeeds
Expected Behavior
The login should succeed and model selection should work, showing models from the providers that responded successfully. Failed providers should be logged as warnings, not block the entire operation.
Actual Behavior
The login prints "configured successfully" but then immediately shows an error from the unrelated stale provider, blocking model selection.
Forge Version
2.9.8
Operating System & Version
Linux (WSL)
Installation Method
npm install -g forgecode
Bug Description
When logging into a provider (e.g.
forge provider login google_ai_studio), the login itself succeeds and prints "configured successfully", but then an unrelated error from a different configured provider is shown. For example, if Bedrock has stale credentials, you get:The "Bearer token is required" error comes from the Bedrock provider, not from Google AI Studio.
Root Cause
After a successful login,
finalize_provider_activation()callsget_all_provider_models()to check model compatibility. That function fetches models from all configured providers concurrently, then collects results with:This
.collect::<Result<Vec<_>>>()fails on the first error from any provider, discarding all successful results.The function's own docstring already describes the intended behavior:
But the implementation doesn't match. One stale provider blocks the entire model-fetching flow for all providers.
Steps to Reproduce
forge provider login <other_provider>with valid credentialsExpected Behavior
The login should succeed and model selection should work, showing models from the providers that responded successfully. Failed providers should be logged as warnings, not block the entire operation.
Actual Behavior
The login prints "configured successfully" but then immediately shows an error from the unrelated stale provider, blocking model selection.
Forge Version
2.9.8
Operating System & Version
Linux (WSL)
Installation Method
npm install -g forgecode