Bug type
Behavior bug (incorrect output/state without crash)
Summary
The google-vertex provider fails to resolve gemini-3.1-flash-lite-preview, returning an "Unknown model" error despite the model being available on Vertex AI.
Steps to reproduce
- Run openclaw models set google-vertex/gemini-3.1-flash-lite.
- Run openclaw models list google-vertex to check status.
- Attempt to send a message to the agent.
Expected behavior
The model should be recognized as configured, show text+image input capabilities, and respond to queries.
Actual behavior
The model appears as missing in the CLI list with no capabilities (-). Interaction fails immediately with: Agent failed before reply: Unknown model: google-vertex/gemini-3.1-flash-lite-preview.
OpenClaw version
2026.3.24 (a710366)
Operating system
Linux Ubuntu 24.04.4 LTS
Install method
npm global
Model
google-vertex/gemini-3.1-flash-lite-preview
Provider / routing chain
google-vertex
Additional provider/model setup details
Auth via Application Default Credentials (ADC). GOOGLE_CLOUD_PROJECT and GOOGLE_CLOUD_LOCATION (global) are correctly set in the environment. Normalization in src/agents/model-id-normalization.ts seems to work (mapping gemini-3.1-flash-lite to the preview ID), but the provider registry lacks the allowed entry for Vertex.
Logs, screenshots, and evidence
09:08:59 error diagnostic {"subsystem":"diagnostic"} lane task error: lane=main durationMs=11 error="FailoverError: Unknown model: google-vertex/gemini-3.1-flash-lite-preview"
09:08:59 error diagnostic {"subsystem":"diagnostic"} lane task error: lane=session:agent:main:main durationMs=13 error="FailoverError: Unknown model: google-vertex/gemini-3.1-flash-lite-preview"
09:08:59 warn model-fallback/decision {"subsystem":"model-fallback/decision"} {"event":"model_fallback_decision","tags":["error_handling","model_fallback","candidate_failed"],"runId":"5dfbe794-6313-4016-9ec7-a817addf7f08","decision":"candidate_failed","requestedProvider":"google-vertex","requestedModel":"gemini-3.1-flash-lite-preview","candidateProvider":"google-vertex","candidateModel":"gemini-3.1-flash-lite-preview","attempt":1,"total":1,"reason":"model_not_found","errorPreview":"Unknown model: google-vertex/gemini-3.1-flash-lite-preview","errorHash":"sha256:324e060fd4f5","isPrimary":true,"requestedModelMatched":true,"fallbackConfigured":false} model fallback decision
09:08:59 error Embedded agent failed before reply: Unknown model: google-vertex/gemini-3.1-flash-lite-preview
Impact and severity
-
Affected users: All users utilizing the google-vertex provider for low-latency/low-cost tiers.
-
Severity: Blocks workflow (cannot use the specific Lite tier).
-
Frequency: Always.
-
Consequence: Forced to use the Pro tier (higher cost/latency) as a workaround.
Additional information
This is a parity gap between providers: the google (AI Studio) provider successfully resolves this model ID, but the google-vertex lane does not. Verified that the model is indeed available on Vertex AI in the selected region (global)
Bug type
Behavior bug (incorrect output/state without crash)
Summary
The google-vertex provider fails to resolve gemini-3.1-flash-lite-preview, returning an "Unknown model" error despite the model being available on Vertex AI.
Steps to reproduce
Expected behavior
The model should be recognized as configured, show text+image input capabilities, and respond to queries.
Actual behavior
The model appears as missing in the CLI list with no capabilities (-). Interaction fails immediately with: Agent failed before reply: Unknown model: google-vertex/gemini-3.1-flash-lite-preview.
OpenClaw version
2026.3.24 (a710366)
Operating system
Linux Ubuntu 24.04.4 LTS
Install method
npm global
Model
google-vertex/gemini-3.1-flash-lite-preview
Provider / routing chain
google-vertex
Additional provider/model setup details
Auth via Application Default Credentials (ADC). GOOGLE_CLOUD_PROJECT and GOOGLE_CLOUD_LOCATION (global) are correctly set in the environment. Normalization in src/agents/model-id-normalization.ts seems to work (mapping gemini-3.1-flash-lite to the preview ID), but the provider registry lacks the allowed entry for Vertex.
Logs, screenshots, and evidence
09:08:59 error diagnostic {"subsystem":"diagnostic"} lane task error: lane=main durationMs=11 error="FailoverError: Unknown model: google-vertex/gemini-3.1-flash-lite-preview" 09:08:59 error diagnostic {"subsystem":"diagnostic"} lane task error: lane=session:agent:main:main durationMs=13 error="FailoverError: Unknown model: google-vertex/gemini-3.1-flash-lite-preview" 09:08:59 warn model-fallback/decision {"subsystem":"model-fallback/decision"} {"event":"model_fallback_decision","tags":["error_handling","model_fallback","candidate_failed"],"runId":"5dfbe794-6313-4016-9ec7-a817addf7f08","decision":"candidate_failed","requestedProvider":"google-vertex","requestedModel":"gemini-3.1-flash-lite-preview","candidateProvider":"google-vertex","candidateModel":"gemini-3.1-flash-lite-preview","attempt":1,"total":1,"reason":"model_not_found","errorPreview":"Unknown model: google-vertex/gemini-3.1-flash-lite-preview","errorHash":"sha256:324e060fd4f5","isPrimary":true,"requestedModelMatched":true,"fallbackConfigured":false} model fallback decision 09:08:59 error Embedded agent failed before reply: Unknown model: google-vertex/gemini-3.1-flash-lite-previewImpact and severity
Affected users: All users utilizing the google-vertex provider for low-latency/low-cost tiers.
Severity: Blocks workflow (cannot use the specific Lite tier).
Frequency: Always.
Consequence: Forced to use the Pro tier (higher cost/latency) as a workaround.
Additional information
This is a parity gap between providers: the google (AI Studio) provider successfully resolves this model ID, but the google-vertex lane does not. Verified that the model is indeed available on Vertex AI in the selected region (global)