add gemini 2.0 flash

[wintonzheng]

Important

Add GEMINI_FLASH_2_0 and GEMINI_FLASH_2_0_LITE configurations to LLMConfigRegistry with vision support and 8192 max tokens.

• Behavior:
  • Adds GEMINI_FLASH_2_0 and GEMINI_FLASH_2_0_LITE configurations to LLMConfigRegistry in config_registry.py.
  • Both configurations support vision, do not add assistant prefix, and have a max of 8192 completion tokens.
• Conditions:
  • Configurations are registered only if settings.ENABLE_GEMINI is true.

^{This description was created by for ded4688. It will automatically update as commits are pushed.}

[ellipsis-dev]

👍 Looks good to me! Incremental review on ded4688 in 26 seconds

More details

• Looked at 31 lines of code in 1 files
• Skipped 0 files when reviewing.
• Skipped posting 3 drafted comments based on config settings.

1. skyvern/forge/sdk/api/llm/config_registry.py:309

• Draft comment:
  Ensure the new GEMINI_FLASH_2_0 config is unique and consistent with other flash models.
• Reason this comment was not posted:
  Confidence changes required: 0% <= threshold 50%
  None

2. skyvern/forge/sdk/api/llm/config_registry.py:318

• Draft comment:
  Consider adding a brief comment documenting differences for GEMINI_FLASH_2_0_LITE.
• Reason this comment was not posted:
  Confidence changes required: 0% <= threshold 50%
  None

3. skyvern/forge/sdk/api/llm/config_registry.py:307

• Draft comment:
  New Gemini Flash 2.0 configs added. Ensure the new model IDs (gemini-2.0-flash-001 and gemini-2.0-flash-lite-preview-02-05) and their parameters (e.g., supports_vision, max_completion_tokens) match the intended spec.
• Reason this comment was not posted:
  Confidence changes required: 0% <= threshold 50%
  None

Workflow ID: wflow_7KJcibySI41lDMPa

---

You can customize Ellipsis with 👍 / 👎 feedback, review rules, user-specific overrides, quiet mode, and more.

[ellipsis-dev]

👍 Looks good to me! Reviewed everything up to ded4688 in 1 minute and 46 seconds

More details

• Looked at 31 lines of code in 1 files
• Skipped 0 files when reviewing.
• Skipped posting 4 drafted comments based on config settings.

1. skyvern/forge/sdk/api/llm/config_registry.py:309

• Draft comment:
  Consider adding a brief inline comment or docstring for the new GEMINI_FLASH_2_0 config to clarify its purpose/differences vs earlier versions.
• Reason this comment was not posted:
  Confidence changes required: 33% <= threshold 50%
  None

2. skyvern/forge/sdk/api/llm/config_registry.py:317

• Draft comment:
  Verify the naming and parameter (max_completion_tokens=8192) for GEMINI_FLASH_2_0_LITE match expected behavior for a lite variant.
• Reason this comment was not posted:
  Confidence changes required: 33% <= threshold 50%
  None

3. skyvern/forge/sdk/api/llm/config_registry.py:308

• Draft comment:
  Add new GEMINI_FLASH_2_0 config registration. Consider adding inline comments to document how this differs from the legacy flash (v1.5) configuration.
• Reason this comment was not posted:
  Decided after close inspection that this draft comment was likely wrong and/or not actionable: usefulness confidence = 20% vs. threshold = 50%
  While the comment identifies a real difference between configurations, adding inline documentation about version differences is not strictly necessary. The model names themselves (2.0 vs 1.5) are self-documenting. The configurations are simple and their differences are clear from the code. Documentation about model version differences belongs in external documentation, not inline comments.
  The comment raises a valid point about helping developers understand version differences. Without documentation, developers might not know the practical implications of choosing one version over another.
  However, inline code comments are not the right place for version comparison documentation. This belongs in API docs, release notes, or other external documentation. The code itself is clear and self-documenting.
  Delete this comment. While documentation comparing versions would be useful, it belongs in external documentation, not inline code comments. The code is clear as-is.

4. skyvern/forge/sdk/api/llm/config_registry.py:317

• Draft comment:
  GEMINI_FLASH_2_0_LITE registration uses a preview identifier ('lite-preview-02-05'). Consider clarifying the versioning or adding documentation for future maintainability.
• Reason this comment was not posted:
  Decided after close inspection that this draft comment was likely wrong and/or not actionable: usefulness confidence = 10% vs. threshold = 50%
  While the comment is technically correct in identifying a preview marker, it's not clear what action should be taken. The model ID likely comes from Gemini's official naming, and we probably can't change it. Documentation about versioning would be nice but isn't strictly necessary since the model ID is self-documenting. The comment doesn't identify a clear problem that needs fixing.
  The model ID might be intentionally marked as preview by Gemini to indicate its experimental status. Having this marker could actually be helpful for maintainers.
  Even if the preview marker is intentional and helpful, the comment doesn't suggest any concrete action items or identify any real problems that need fixing.
  Delete the comment. While it makes an observation about the model name, it doesn't identify any actual issues that need to be fixed and doesn't provide actionable feedback.

Workflow ID: wflow_giFA8s50RCCUAM91

---

You can customize Ellipsis with 👍 / 👎 feedback, review rules, user-specific overrides, quiet mode, and more.