Add GPT-5.5 and Gemini 3.1 Pro model selectors#176
Open
akseljoonas wants to merge 3 commits intomainfrom
Open
Conversation
The web selector and CLI suggestions now expose GPT-5.5 with the product-level high reasoning default and Gemini 3.1 Pro through LiteLLM's Gemini provider path. A small shared default-effort helper keeps frontend sessions and CLI probes aligned when the global preference is max, while preserving explicit user effort choices. Constraint: LiteLLM routes Gemini API-key usage with the gemini/ provider prefix, not through the Hugging Face router. Constraint: The local sandbox blocks writes under .git, so this commit was created through the GitHub API instead of local git refs. Rejected: Treat gemini/gemini-3.1-pro-preview as an HF-router id | that would send it to the wrong base URL and token path. Confidence: high Scope-risk: narrow Directive: Keep frontend MODEL_OPTIONS, backend AVAILABLE_MODELS, and CLI SUGGESTED_MODELS in sync when adding selector models. Tested: UV_CACHE_DIR=/tmp/uv-cache uv run pytest tests/unit/test_llm_params.py tests/unit/test_cli_rendering.py Tested: npm run build Not-tested: Full unit suite is red on existing doom_loop assertion text expectations; frontend lint is red on pre-existing unrelated prefer-const/no-unused-vars issues.
|
Claude encountered an error —— View job I'll analyze this and get back to you. |
A real CLI run against the Gemini selector reached provider auth, but the local environment has no Gemini API key. The existing friendly auth message only named Anthropic, OpenAI, and HF Router, so Gemini users would not see the variable required for the new selector entry. Constraint: No GEMINI_API_KEY is present in the local environment or .env, so live Gemini completion cannot be validated from this machine. Confidence: high Scope-risk: narrow Directive: Keep provider auth hints aligned with every direct-provider model shown in selectors. Tested: UV_CACHE_DIR=/tmp/uv-cache uv run pytest tests/unit/test_llm_params.py tests/unit/test_cli_rendering.py tests/unit/test_llm_error_classification.py Tested: npm run build Tested: UV_CACHE_DIR=/tmp/uv-cache uv run ml-intern --model openai/gpt-5.5 --max-iterations 1 --no-stream 'Reply with exactly: OK' Not-tested: Gemini live completion; GEMINI_API_KEY is not available locally.
Collaborator
Author
|
Live CLI validation update:
Additional checks after the follow-up:
|
|
Claude encountered an error —— View job I'll analyze this and get back to you. |
Live CLI validation with a Gemini key reached Google successfully but free-tier quota for both Gemini 3.1 Pro and Gemini 2.0 Flash was exhausted. The failed 3.1 Pro run showed that raw fallback tracebacks can include provider URLs with API keys, so quota and rate-limit errors now render through a friendly message and fallback tracebacks are scrubbed before surfacing. Constraint: The supplied Gemini key has zero free-tier quota for gemini-3.1-pro and gemini-2.0-flash, so successful Gemini completion could not be validated. Rejected: Leave quota failures as transient tracebacks | provider URLs may contain API keys. Confidence: high Scope-risk: narrow Directive: Any new direct provider should be represented in auth hints and secret redaction patterns before adding it to selectors. Tested: UV_CACHE_DIR=/tmp/uv-cache uv run pytest tests/unit/test_llm_error_classification.py tests/unit/test_redact.py tests/unit/test_llm_params.py tests/unit/test_cli_rendering.py Tested: npm run build Tested: GEMINI_API_KEY=<redacted> UV_CACHE_DIR=/tmp/uv-cache uv run ml-intern --model gemini/gemini-2.0-flash --max-iterations 1 --no-stream 'Reply with exactly: OK' | provider returned RESOURCE_EXHAUSTED; CLI ended with friendly quota error Not-tested: Successful Gemini completion; key quota is exhausted / limit 0.
Collaborator
Author
|
Additional Gemini live validation:
Checks after the fix:
Security note: the Gemini key used for validation was pasted into chat and appeared in an upstream traceback before the redaction fix; it should be rotated. |
|
Claude encountered an error —— View job I'll analyze this and get back to you. |
|
closed per maintainer request |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Verification
UV_CACHE_DIR=/tmp/uv-cache uv run pytest tests/unit/test_llm_params.py tests/unit/test_cli_rendering.pynpm run buildKnown unrelated failures
UV_CACHE_DIR=/tmp/uv-cache uv run pytest tests/unitfails two existingtest_doom_loop.pyassertion text expectations.npm run lintfails existing unrelated lint issues inActivityStatusBar.tsx,agentStore.ts,main.tsx, andlogger.ts.