[Ephraim-9]

Problem

On Usage, most Antigravity tokens show as Unpriced. Two causes:

• The reader's MODEL_IDS table stopped at Gemini 3.1 Pro / 3 Flash, so current Gemini Flash generations fell through to labels like antigravity-model-1319 (#19189).
• Name-based records carry Antigravity's effort and routing suffixes (-low, -medium, -high, -extra-low, -tiered, -control, -thinking, -agent), which the LiteLLM rate table does not list.

Change

apps/server/src/provider/Drivers/antigravityUsageReader.ts:

• Adds the current Gemini Flash ids (1020, 1072, 1073, 1132, 1133, 1187, 1298, 1299, 1301, 1318, 1319, 1320, 1322) and 342 to MODEL_IDS, using the names Antigravity writes in conversation metadata.
• Sets rateModel on every emitted Antigravity record to the base model with the variant suffix stripped, the same mechanism Cursor uses through cursorRateModel. gemini-3.8-flash-high keeps its display name and prices as gemini-3.8-flash; gemini-3-flash-* prices as gemini-3-flash-preview, the only name the table has (matching existing ids 1018/1047/1084).
• rateModel is derived once, in the emission loop, after alias merging has settled the final model. A model-less step that adopts a named generation's model through merging is priced too.

Known limits: id 342 (gpt-oss-120b-medium) gains a readable name but stays unpriced, because LiteLLM has no bare gpt-oss-120b and its provider-prefixed entries disagree on rate. Existing ids 313/329 (gemini-2.5-flash-thinking) now price at gemini-2.5-flash. Names that do not identify a generation (gemini-default) and ids without a recorded name stay unpriced on purpose.

Verification

Headless before/after on Linux: a seeded Antigravity conversation store with 1M uncached input tokens per record, read with readAntigravityUsage and priced with priceUsage against the live LiteLLM table.

Record	Before (main)	After
id 1319	antigravity-model-1319, unpriced	gemini-3.8-flash-medium → gemini-3.8-flash, $0.75
id 1318	antigravity-model-1318, unpriced	gemini-3.8-flash-high → gemini-3.8-flash, $0.75
id 1073	antigravity-model-1073, unpriced	gemini-3.6-flash-low → gemini-3.6-flash, $0.75
id 1020	antigravity-model-1020, unpriced	gemini-3.5-flash-low → gemini-3.5-flash, $1.50
id 1132	antigravity-model-1132, unpriced	gemini-3-flash-agent → gemini-3-flash-preview, $0.50
id 1301	antigravity-model-1301, unpriced	gemini-3.7-flash-tiered → gemini-3.7-flash, $0.75
name "Gemini 3.7 Flash Control"	gemini-3.7-flash-control, unpriced	gemini-3.7-flash-control → gemini-3.7-flash, $0.75

• vp test run src/provider/Drivers/antigravityUsageReader.test.ts src/usage/usagePricing.test.ts in apps/server: 24 passed. The one new reader test (prices Antigravity effort and routing variants at their base model) covers an id-only step, a suffixed name-only generation, the gemini-3-flash-preview remap, and the merge case; it fails with the reader reverted to main. Existing usagePricing tests already prove rateModel is honored.
• vp lint, vp fmt --check, and tsc for apps/server are clean.
• Not verified: the Usage page against a real Antigravity account, macOS, Windows, iOS, Android, or Electron desktop. Headless proof exercises the reader and pricing service on seeded data; no UI layout changes.

Credit

Taken over from @Ephraim-9's original #13809, which found the ids and the suffix set and chose the rateModel approach. The takeover moved the rateModel derivation to a single place after merging and folded the two tests into one.

Fixes #19189.

Rebuilt by Claude Fable 5.1 (Claude Code in T3 Code); tests by Claude Opus 5.5; reviewed by Claude Opus 5.5 and GPT-6 Sol.

🤖 Generated with Claude Code

Resumed and independently reviewed with Claude Opus 5.5 and GPT-6 Sol; verification and preparation by GPT-6.1 Sol through the Codex harness in T3 Code.

[macroscopeapp]

Approvability

Verdict: Needs Human Approval

Macroscope's review found this PR requires human approval — This focused fix changes production usage metering by mapping Antigravity model variants to pricing models and making previously unpriced usage appear with calculated costs. Because it affects pricing and metering behavior, the change warrants human review despite its limited scope and tests.

^{You can add or adjust custom eligibility rules. Learn more.}

[coderabbitai]

No actionable comments were generated in the recent review.

Recent review info

Run configuration

• Source mode: Auto
• Repository: pingdotgg/t3code
• Branch: main
• File: .coderabbit.config.ts
• Commit: 62245d8
• Layers: Repository TS config
• Review profile: CHILL
• Plan: Advanced
• Run ID: be235db8-25b6-480c-b5d1-fd137a3772b6

Commits

Reviewing files that changed from the base of the PR and between 9ffff74 and 765670c.

Files selected for processing: 2

• apps/server/src/provider/Drivers/antigravityUsageReader.test.ts
• apps/server/src/provider/Drivers/antigravityUsageReader.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 1 remain after this review.

Walkthrough

The Antigravity usage reader recognizes additional Gemini model IDs and derives rate models for effort and routing variants. Tests verify four model-to-rate-model mappings and their calculated costs.

Changes: Antigravity model recognition and pricing

Layer / File(s)	Summary
Model recognition and rate pricing apps/server/src/provider/Drivers/antigravityUsageReader.ts, apps/server/src/provider/Drivers/antigravityUsageReader.test.ts	The reader adds Gemini 3.5–3.8 Flash and Gemini 3 Flash Agent ID mappings. It removes additional effort and routing suffixes when deriving rate models and maps gemini-3-flash to gemini-3-flash-preview. Tests check four model-to-rate-model mappings and their costs.

Priority: Normal

Estimated code review effort: 2 (Simple) | ~10 minutes

Change: Bug fix · Severity of issue fixed: Low

Possibly related PRs

• pingdotgg/t3code#24834: Both PRs derive base rate models from Antigravity model variants for pricing.

Suggested reviewers: t3dotgg

Merge Risk: Minimal · up to 76567

No issue requiring a fix before merge was established. Normal checks can proceed.

Security architecture risk: Minimal · up to 9ffff

The change adjusts cost estimates while preserving recorded identities and existing read-only access. No new privilege, access path, or material security regression was identified in the reviewed change.

Retained concerns
No architecture-level concerns identified.

Security review details

Security Blast Radius

• inferred — A party able to modify metadata in scanned conversation stores can influence usage identities and cost estimates. The reviewed change alters pricing interpretation of those records without expanding scanned roots or gaining additional authority.

Trust Boundaries and Controls

• observed — Metadata-derived model names remain data: normalization produces a string key, and rate resolution performs a map lookup. The inspected path does not use that key as executable code, SQL, a filesystem path, or a credential selector.

Resilience and Maintainability Implications

• observed — The existing failure-containment behavior remains: cache entries are published after successful decoding, failed file reads are recorded and skipped, and merged results are assembled in invocation-local state. The PR does not introduce a new partial-publication or recovery transition.

Pre-merge checks: 4

Check name	Status	Explanation
Title check		The title clearly and concisely identifies the main change: pricing Antigravity Gemini Flash effort variants.
Description check		The description is complete and relevant. It explains the problem, implementation, scope, linked issue, verification results, known limits, and unverified environments. The Scope and approval section …
Linked Issues check		Issue #19189 requires readable Antigravity model names instead of IDs such as antigravity-model-1073. MODEL_IDS now maps ID 1073 and the other current Flash IDs to named generations. modelName p…
Out of Scope Changes check		The changes are limited to Antigravity model mappings, base-model rate selection, and automated reader and pricing coverage. The rate selection supports the readable model records required by issue #1…

Full details: Description check

Explanation

The description is complete and relevant. It explains the problem, implementation, scope, linked issue, verification results, known limits, and unverified environments. The Scope and approval section does not state explicit maintainer approval, but the linked issue and focused-fix rationale provide sufficient context.

Full details: Linked Issues check

Explanation

Issue #19189 requires readable Antigravity model names instead of IDs such as antigravity-model-1073. MODEL_IDS now maps ID 1073 and the other current Flash IDs to named generations. modelName preserves recorded names such as gemini-3.7-flash-control. The added reader test verifies readable names, suffix handling, preview remapping, pricing, and model recovery after merging. These changes satisfy the issue's coding requirement.

Full details: Out of Scope Changes check

Explanation

The changes are limited to Antigravity model mappings, base-model rate selection, and automated reader and pricing coverage. The rate selection supports the readable model records required by issue #19189. No unrelated product area or unrelated behavior is changed.

Finishing Touches

🧪 Generate unit tests (beta)

• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

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

[juliusmarminge]

Taking over this one. I'll rebuild the implementation to current conventions, trim the tests to the minimum that proves the fix, and push to this branch.

[maria-rcks]

Note

Written by gpt-6.1-sol on behalf of Maria

updated the original branch and kept @Ephraim-9's commits. merged current main, reused its withRateModel helper, and kept pricing after alias merging. the fix retains the 13 flash id mappings, extends suffix handling, and maps gemini 3 flash to its preview rate. dropped only id 342 (gpt-oss-120b-medium), which remains unpriceable and is outside this flash fix. compared the original diff against the final two-file change.

verification passed:

• reproduced the regression against current main before fixing it.
• 23 reader and pricing tests on blacksmith, including database decoding, merged model recovery, and numeric cost assertions.
• vp run check:changed on blacksmith: format, lint, knip, graphql, affected typechecks, and reader tests.
• real web client → usage service → antigravity sqlite reader → live litellm rates: the same six records with 1m input tokens each change from unpriced to $5.00. five numeric ids cover gemini flash 3, 3.5, 3.6, 3.7, and 3.8; a named control variant covers suffix pricing.
• two independent read-only reviews found no blockers.

unverified: a real antigravity account's current metadata, windows/macOS, native mobile, and electron. the captures use seeded provider-format data, not a live antigravity conversation. all 31 ci checks passed on 765670cbf47ca72357148c6bd28bc9a6db7692fb. coderabbit found no actionable comments; macroscope requires human review for pricing changes. the obsolete alias-pricing thread is resolved. the tokens view still reports the same 6m-token total.