chore(ai-gemini): sync model meta with Google's current model list by season179 · Pull Request #714 · TanStack/ai

season179 · 2026-06-06T04:52:50Z

Summary

Closes #620, closes #621.

Syncs @tanstack/ai-gemini model metadata with Google's current published model list, migrates every workspace consumer off the retired ids, and fixes the structured-output demo dropdown.

All removals/additions were verified against the live API (GET /v1beta/models + real generateContent calls on a paid Tier 1 key) on 2026-06-06, not just the docs page.

Added

gemini-3.1-flash-lite (stable GA) to model-meta, all type maps, and GEMINI_COMBINED_TOOLS_AND_SCHEMA_MODELS — it round-trips the native combined tools+schema path. The existing preview const was renamed GEMINI_3_1_FLASH_LITE → GEMINI_3_1_FLASH_LITE_PREVIEW to follow the stable/_PREVIEW naming convention used by the 2.5 line.

Removed (verified retired)

id	live API status	successor
`gemini-3-pro-preview`	404 ("no longer available")	`gemini-3.1-pro-preview`
`gemini-2.5-flash-preview-09-2025`	gone from `/v1beta/models`	`gemini-2.5-flash`
`gemini-2.5-flash-lite-preview-09-2025`	gone from `/v1beta/models`	`gemini-2.5-flash-lite`
`gemini-2.0-flash`	docs-retired	`gemini-2.5-flash`
`gemini-2.0-flash-lite`	docs-retired	`gemini-2.5-flash-lite`
`gemini-2.0-flash-preview-image-generation`	gone from `/v1beta/models`	`gemini-2.5-flash-image`

Note: gemini-2.0-flash / -lite still respond at the API level (deprecation grace), but they're gone from the published docs per the #620 acceptance criteria, so they're removed here.

Drive-by fixes (beyond the strict issue text, all small and disclosed here)

gemini-3.5-flash was missing from GeminiChatModelToolCapabilitiesByName, so its provider-tool typing was broken. One-line fix in a map this PR rewrites anyway.
gemini-3.5-flash was grouped under "no document" in GeminiModelInputModalitiesByName despite supports.input including document.
Display labels next to migrated model values: the svelte/vue/panel model-selection lists ended up with a duplicate gemini-2.5-flash entry (the old 2.0-flash row pointed at the same id) — duplicates removed, stale "Gemini - 2.0 Flash" labels updated. Same label sync in the panel's structured/summarize provider chips and its OpenRouter row.
Removed two stale commented-out blocks in model-meta.ts: the GEMINI_2_FLASH_LIVE draft (a retired 2.0-line id) and a GEMINI_MODEL_META draft that referenced consts deleted by this PR. Happy to restore if you prefer the tighter diff.

Consumer migration (#620 cascade)

testing/e2e (providers, summarize route, media-providers), testing/panel (model-selection, chat/structured/summarize/image routes, vendor-config), ai-code-mode-skills test CLI, and the react/svelte/vue chat examples.
Panel's OpenRouter entry google/gemini-2.0-flash-001 → google/gemini-2.5-flash (validated against the OpenRouter catalog).
Docs: stale ids corrected in adapters/gemini.md, adapters/openrouter.md (google/gemini-3-pro-preview isn't in OpenRouter's catalog either), advanced/multimodal-content.md, media/image-generation.md, and the ai-core skill reference. Factual fixes only — no addedAt/updatedAt bumps.
Deliberately untouched: ai-openrouter/model-meta.ts and scripts/openrouter.models.json (nightly-synced from OpenRouter), community-adapters/cencori.md (third-party service's own table).

Tests

The retired 2.0 line was the last chat model without thinking support, so the "models WITHOUT thinking" type-test category is gone. Those negative tests were replaced with retired-id rejection tests — geminiText('gemini-2.0-flash') / geminiText('gemini-3-pro-preview') are now compile errors, guarded by @ts-expect-error. The mixed accept/reject tool-gating coverage moved from gemini-2.0-flash-lite (rejected everything) to gemini-2.5-flash-lite (partial tool set, still exercises both directions). Full positive coverage added for the new stable gemini-3.1-flash-lite.

Demo (#621)

Dropdown: removed gemini-3-pro-preview (404s), added gemini-3.1-flash-lite.
Server fallback default: gemini-3-pro-preview → gemini-3.5-flash — picked because it's the newest stable (non-preview) 3.x id and matches the dropdown's first entry; gemini-3.1-pro-preview is still a preview id. Happy to switch if you'd rather default to the pro line.

Open questions for maintainers

imagen-3.0-generate-002 is also gone from the live API (verified), but chore(ai-gemini): remove retired model ids + cascade workspace consumer migration #620 scoped imagen variants as out-of-scope-but-check. It's kept in this PR. Want me to remove it here too, or file a follow-up issue?
Fallback default — gemini-3.5-flash (stable) vs gemini-3.1-pro-preview (newer flagship, but preview). Currently stable wins; say the word and I'll flip it.

Verification

pnpm test:pr — green (sherif/knip/docs/eslint/lib/types/build, 32 projects)
E2E suite — 253 passed, 0 failed (10 flaky-passed on non-Gemini providers, pre-existing)
Live round-trip of every dropdown entry through /api/structured-output on a paid key — all five stream and deliver structured-output.complete:
gemini-3.5-flash
gemini-3-flash-preview
gemini-3.1-pro-preview
gemini-3.1-flash-lite
gemini-3.1-flash-lite-preview

Summary by CodeRabbit

Documentation
- Updated docs and tables to reflect current Gemini models, image-model availability, and multimodal capabilities.
Chores
- Synchronized supported Gemini models with Google's published list: added newer models (notably gemini-3.1-flash-lite and gemini-2.5-flash), removed retired IDs, and updated defaults and UI labels across examples and demos.
Tests
- Expanded tests and type-safety checks to cover new models and assert removal of retired ones.
Bug Fixes
- Fixed a missing capability mapping for a Gemini model.

coderabbitai · 2026-06-06T04:52:58Z

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 283fbdca-07e8-46b8-b711-5b20f477dbcd

📥 Commits

Reviewing files that changed from the base of the PR and between 83510dc and c38bd40.

📒 Files selected for processing (1)

packages/ai/skills/ai-core/adapter-configuration/references/gemini-adapter.md

✅ Files skipped from review due to trivial changes (1)

packages/ai/skills/ai-core/adapter-configuration/references/gemini-adapter.md

📝 Walkthrough

Walkthrough

Refreshes Gemini model metadata and types (adds stable gemini-3.1-flash-lite, removes retired 2.x/preview ids), updates registries/type maps, and migrates workspace consumers (examples, tests, E2E, panel, docs, and CLI) to current Gemini model IDs.

Changes

Gemini Model Metadata Refresh and Consumer Migration

Layer / File(s)	Summary
Core model metadata, types, and registries `packages/ai-gemini/src/model-meta.ts`, `packages/ai-gemini/src/image/image-provider-options.ts`	Model constants and pricing changed (replace GEMINI_3_PRO with GEMINI_3_FLASH, add GEMINI_3_1_FLASH_LITE/preview, remove older 2.x previews). Exported registries (`GEMINI_MODELS`, `GEMINI_IMAGE_MODELS`, `GEMINI_COMBINED_TOOLS_AND_SCHEMA_MODELS`) and type maps (`GeminiChatModelProviderOptionsByName`, `GeminiChatModelToolCapabilitiesByName`, `GeminiModelInputModalitiesByName`) updated accordingly. `GeminiNativeImageModels` union removes the retired image id.
Type-safety and metadata tests `packages/ai-gemini/tests/*`	Compile-time and runtime tests updated: add stable `gemini-3.1-flash-lite` coverage (thinking + structured-output), add `gemini-3.5-flash` assertions, remove retired-model assertions, and update tool-gating and modality assertions.
Adapter tests and expected payloads `packages/ai-gemini/tests/gemini-adapter.test.ts`	Adapter tests updated to expect current model IDs (e.g., `gemini-2.5-flash` for summarize, `gemini-3.1-pro-preview` in combined-mode tests).
Example applications: defaults and dropdowns `examples//src/lib/model-selection.ts`, `examples//src/routes/*`	Example apps migrated from `gemini-2.0-flash` to `gemini-2.5-flash` as defaults; structured-output dropdowns remove `gemini-3-pro-preview` and add `gemini-3.1-flash-lite`. API route defaults updated.
E2E and testing panel defaults `testing/e2e/`, `testing/panel/`	E2E providers and media providers now use `gemini-2.5-flash` / `gemini-2.5-flash-image`. Testing panel model selections, API route defaults, and vendor test config updated to current IDs and OpenRouter Gemini naming.
Docs, JSDoc examples, CLI defaults, changeset `docs/`, `packages/ai-gemini/src/adapters/summarize.ts`, `packages/ai-code-mode-skills/`, `.changeset/*`	Documentation tables and references updated (image models, multimodal docs, adapter reference). JSDoc examples and CLI/env defaults set to `gemini-2.5-flash`. Changeset updated to document removed IDs, added stable model, and migration guidance.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

TanStack/ai#691: touches Gemini model identifiers in multimodal-content docs — related documentation area.
TanStack/ai#609: added native combined tools+schema routing that depends on GEMINI model registries — related to updated registries and maps.

Suggested reviewers

tombeckenham
AlemTuzlak
jherr

"🐰 I hopped through registries, swapped old ids for new,
Three-one-flash-lite now takes its place in view.
Two-point-five skipped along the garden test,
Docs and demos sing — migrations all addressed.
A tiny carrot of change for reviewers to chew!"

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 20.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately reflects the main change: syncing Gemini model metadata with Google's current model list. It is concise, clear, and specific.
Description check	✅ Passed	The description is comprehensive and well-structured, covering changes, migration details, verification, and open questions. It exceeds template requirements with thorough documentation.
Linked Issues check	✅ Passed	The PR directly addresses all requirements from `#620` and `#621`: removes retired model IDs, adds stable gemini-3.1-flash-lite, fixes tool capabilities, migrates consumers, updates dropdown, and includes test coverage.
Out of Scope Changes check	✅ Passed	All changes are directly scoped to syncing Gemini model metadata and migrating consumers. Documentation updates, test modifications, and UI label corrections are all aligned with the stated objectives.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

Remove retired Gemini ids (gemini-3-pro-preview, the 09-2025 2.5 previews, and the 2.0 line), add the stable gemini-3.1-flash-lite, and migrate every workspace consumer off the removed ids. Also fixes the structured-output demo dropdown, which still offered the retired gemini-3-pro-preview as its default. All removals were verified against the live API, not just the docs page. Includes a couple of small fixes found along the way: missing gemini-3.5-flash entries in the tool-capability and modality maps, and stale display labels left next to migrated model values. Closes TanStack#620, closes TanStack#621

…nt-input note Add the stable gemini-3.1-flash-lite row to the skill reference's Key Chat Models table, and fix the modality note that wrongly claimed all Gemini text models accept document input — gemini-2.5-flash does not. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

tombeckenham · 2026-06-09T00:40:14Z

Thanks @season179 !

season179 · 2026-06-09T00:42:30Z

No problem! I love the recent releases and want to be part of the movement.

season179 marked this pull request as ready for review June 6, 2026 04:57

season179 force-pushed the gemini-model-meta-refresh branch from 8ed0409 to 83510dc Compare June 8, 2026 13:33

tombeckenham self-requested a review June 9, 2026 00:09

tombeckenham merged commit efa76c8 into TanStack:main Jun 9, 2026
3 checks passed

season179 deleted the gemini-model-meta-refresh branch June 9, 2026 00:41

github-actions Bot mentioned this pull request Jun 9, 2026

ci: Version Packages #724

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

chore(ai-gemini): sync model meta with Google's current model list#714

chore(ai-gemini): sync model meta with Google's current model list#714
tombeckenham merged 2 commits into
TanStack:mainfrom
season179:gemini-model-meta-refresh

season179 commented Jun 6, 2026 •

edited by coderabbitai Bot

Loading

Uh oh!

coderabbitai Bot commented Jun 6, 2026 •

edited

Loading

Walkthrough

Changes

Estimated code review effort

Possibly related PRs

Suggested reviewers

❌ Failed checks (1 warning)

Uh oh!

Uh oh!

tombeckenham commented Jun 9, 2026

Uh oh!

season179 commented Jun 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

season179 commented Jun 6, 2026 • edited by coderabbitai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Added

Removed (verified retired)

Drive-by fixes (beyond the strict issue text, all small and disclosed here)

Consumer migration (#620 cascade)

Tests

Demo (#621)

Open questions for maintainers

Verification

Summary by CodeRabbit

Uh oh!

coderabbitai Bot commented Jun 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Changes

Estimated code review effort

Possibly related PRs

Suggested reviewers

❌ Failed checks (1 warning)

Uh oh!

Uh oh!

tombeckenham commented Jun 9, 2026

Uh oh!

season179 commented Jun 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

season179 commented Jun 6, 2026 •

edited by coderabbitai Bot

Loading

coderabbitai Bot commented Jun 6, 2026 •

edited

Loading