fix(cli): add gemini-3.5-flash to model selector for all users (#28483)

[kunalrawat425]

Problem

Fixes #28483 — users on v0.51.0 cannot select gemini-3.5-flash or gemini-3.6-flash from the model selector.

Root cause: buildAvailableModels and ModelDialog (legacy paths) only surface DEFAULT_GEMINI_FLASH_MODEL which stays as gemini-2.5-flash until the backend GEMINI_3_5_FLASH_GA_LAUNCHED experiment flag fires. Users not yet enrolled in the rollout cannot see gemini-3.5-flash as a selectable option, and gemini-3.6-flash was not defined in the codebase at all.

Fix

1. Add DEFAULT_GEMINI_3_5_FLASH_MODEL (gemini-3.5-flash) to the legacy-path manualOptions list in both acpUtils.ts and ModelDialog.tsx, guarded by a deduplication check so it does not appear twice once the GA flag promotes it to DEFAULT_GEMINI_FLASH_MODEL.

2. Add gemini-3.6-flash and gemini-3.5-flash-lite as new model constants and surface them in the legacy model selector in both acpUtils.ts and ModelDialog.tsx, defaultModelConfigs.ts, VALID_GEMINI_MODELS, and documentation.

3. Guard resolveModel() to prevent overriding an explicitly-set gemini-3.6-flash back to the default flash model.

Changes

File	Change
packages/core/src/config/models.ts	Add DEFAULT_GEMINI_3_5_FLASH_MODEL, DEFAULT_GEMINI_3_6_FLASH_MODEL, DEFAULT_GEMINI_3_5_FLASH_LITE_MODEL constants; add to VALID_GEMINI_MODELS; guard resolveModel
packages/core/src/config/defaultModelConfigs.ts	Add model configs for gemini-3.5-flash, gemini-3.6-flash, gemini-3.5-flash-lite
packages/core/index.ts	Export new model constants from public package index
packages/cli/src/acp/acpUtils.ts	Add new models to manualOptions with dedup guard for gemini-3.5-flash
packages/cli/src/ui/components/ModelDialog.tsx	Add new models to legacy selector with dedup guard for gemini-3.5-flash
packages/cli/src/acp/acpUtils.test.ts	New: 3 tests covering selector contents
packages/core/src/config/models.test.ts	Tests for resolveModel guard
schemas/settings.schema.json	Add new model IDs to allowed values
docs/reference/configuration.md	Document new model options

Test plan

• npx vitest run packages/cli/src/acp/acpUtils.test.ts packages/core/src/config/models.test.ts — 98 tests pass
• npm run build — clean build, no type errors introduced by this PR
• Pre-commit hooks pass (ESLint + Prettier)
• Manual: run gemini → /model → verify gemini-3.5-flash, gemini-3.6-flash, and gemini-3.5-flash-lite appear in the list

[gemini-code-assist]

Summary of Changes

Hello, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request addresses a visibility issue for the gemini-3.5-flash model in the CLI, ensuring it is selectable for all users regardless of experiment rollout status. Additionally, it enhances the CLI's robustness by adding an auto-compression feature for context overflows and improving the usability of error messages by sanitizing embedded URLs.

Highlights

• Model Selector Fix: Added gemini-3.5-flash as an explicit option in the legacy model selector to ensure availability for all users, including a deduplication guard.
• Auto-Compression Feature: Introduced autoCompressOnOverflow setting to automatically compress chat history when context limits are reached, improving the user experience during long sessions.
• URL Sanitization: Implemented sanitizeUrlsInMessage to strip trailing punctuation from URLs in error messages, ensuring links remain clickable.
• Code Cleanup: Refactored vscode-ide-companion extension activation to remove unnecessary wrapping parentheses in subscription pushes.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize the Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counterproductive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[github-actions]

📊 PR Size: size/L

• Lines changed: 355
• Additions: +351
• Deletions: -4
• Files changed: 9

[gemini-code-assist]

Code Review

This pull request introduces several enhancements and bug fixes across the repository. Key changes include adding an autoCompressOnOverflow setting to automatically compress chat history when approaching context limits, ensuring gemini-3.5-flash is always available as a selectable model, sanitizing trailing punctuation from URLs in error messages, and fixing skill description parsing for single-line or colon-containing descriptions. Feedback on these changes highlights that the background auto-compression task in useGeminiStream.ts lacks proper cancellation handling and should accept and propagate an AbortSignal to prevent dangling operations.

[kunalrawat425]

I have split the other unrelated fixes into separate focused pull requests:

• VS Code companion disposables cleanup: fix(vscode-ide-companion): remove comma operator wrapping in activate() #28489
• Error message URL sanitization: fix(core,cli): sanitize URLs in error messages #28490
• Skill description parser regression tests: test(core): add regression tests for skill loader descriptions #28491
• Context window auto-compression: feat(cli): auto-compress chat history on context window overflow #28488

[kunalrawat425]

Good question — the short answer is no, this PR only affects the manual model picker, not the auto router.

The auto mode resolves to concrete models via resolveModel() in packages/core/src/config/models.ts, controlled entirely by backend experiment flags (useGemini3_1, useGemini3_5Flash, hasAccessToPreview) — not by what appears in the picker list. Specifically:

• auto → Pro tier: gemini-3.1-pro-preview (if useGemini3_1) else gemini-3-flash-preview / gemini-2.5-pro
• auto → Flash tier: gemini-3.5-flash (if useGemini3_5Flash flag fires) else DEFAULT_GEMINI_FLASH_MODEL (gemini-2.5-flash)

gemini-3.6-flash and gemini-3.5-flash-lite are not in the auto routing switch — auto will not route to them unless a future backend change adds them there.

What this PR does is narrower: it adds gemini-3.5-flash, gemini-3.6-flash, and gemini-3.5-flash-lite as explicitly selectable options in the manual picker (/model → Manual), so users can explicitly choose them regardless of which experiment flags are active for them.

If the intent is also for auto to consider gemini-3.6-flash, that would require a separate change to resolveModel() and the auto model routing logic.

[kunalrawat425]

Flagging for maintainer attention — this PR is ready for review:

• ✅ All automated checks pass (check-changes, size-label, cla/google)
• ✅ Tests added (98 pass)
• ✅ Docs updated (docs/reference/configuration.md)
• ✅ Confirmed working locally — gemini-3.6-flash selectable and responds correctly
• ✅ Scope is tight — only touches the manual model picker, no impact on auto routing

The scope question from @hi-fox has been answered. Happy to make any changes the team needs.

[kchugh-s]

Is there a way to see where the PR ranks in maintainers' triage? Or are they chosen at random

[gemini-cli]

Hi there! Thank you for your interest in contributing to Gemini CLI.

To ensure we maintain high code quality and focus on our prioritized roadmap, we only guarantee review and consideration of pull requests for issues that are explicitly labeled as 'help wanted'.

This PR will be closed in 7 days if it remains without that designation. We encourage you to find and contribute to existing 'help wanted' issues in our backlog! Thank you for your understanding.

[gemini-cli]

This pull request is being closed as it has been open for 14 days without a 'help wanted' designation. We encourage you to find and contribute to existing 'help wanted' issues in our backlog! Thank you for your understanding.

[kchugh-s]

This type of triage is abysmal. Someone performed free labor to improve a company's product and it's been ignored