PR #635 returned HTTP 200 with {"input_tokens": 0} when upstream doesn't
support count_tokens (404). This caused Claude Code CLI to trust the zero
value, believing context uses 0 tokens, so auto-compression never triggers.
Fix: return 404 with proper error body so CLI falls back to its local
tokenizer for accurate estimation. Return nil (not error) to avoid
polluting ops error metrics with expected 404s.
Affected paths:
- Passthrough APIKey accounts: upstream 404 now passed through as 404
- Antigravity accounts: same fix (was also returning fake 200)
- Remove Gemini platform exclusion from model restriction UI in
Create/Edit account modals (Gemini now supports model_mapping)
- Remove outdated Gemini model passthrough info cards
- Add model_mapping field to GeminiCredentials type
- Extend warmup request interception toggle to Antigravity platform
- Remove redundant try/catch in API key account creation
- Remove noisy gateway.request_completed debug log
- Reorganize Gemini model mapping sections in constants.go
The ClientSecret constant was left as an empty string after
getClientSecret() was refactored to use defaultClientSecret.
Remove the dead constant and update the test accordingly.
- Fix oauth_test.go and client_test.go to use defaultClientSecret
variable instead of env var (init() already sets the default)
- Align migration 058 gemini-3-pro-high/low/preview mappings with
constants.go (map to 3.1 versions)
Add a built-in default for ANTIGRAVITY_OAUTH_CLIENT_SECRET so the
service works out of the box without requiring environment variable
configuration. The env var can still override the default.
Add claude-sonnet-4-6 to identity injection modelInfoMap and
Antigravity model selector. Add gemini-3.1-pro-high/low to
Antigravity model list and Sonnet 4.6 preset mapping.
- Add claude-sonnet-4-6 to default Antigravity model mapping
- Add antigravity_extra_retries default value in config
- Add cache-dependency-path to CI setup-go for faster builds
- Simplify vitest config to avoid vite plugin compatibility issues
Remove the special case that bypassed model-supported checks for Gemini
API Key accounts, allowing model_mapping to filter requests properly.
Add tests for multiplatform model filtering behavior.
- Use mapped model (billingModel) instead of original request model for billing
- Use resolveFinalAntigravityModelKey for 429 rate limit model key,
ensuring rate limit records match the actual upstream model
- Add regression tests for both fixes
Add a dedicated CheckMixedChannel endpoint that allows the frontend
to pre-validate mixed channel risk before submitting create/update
requests. This improves UX by showing warnings earlier in the flow
instead of only after form submission.
Backend changes:
- Add CheckMixedChannelRequest struct and CheckMixedChannel handler
- Register POST /check-mixed-channel route
- Expose CheckMixedChannelRisk as public method on AdminService
- Simplify Create/Update 409 responses (remove details/require_confirmation)
- Add comprehensive handler tests and stub methods
Frontend changes:
- Add checkMixedChannelRisk API function and TypeScript types
- Refactor CreateAccountModal to precheck before step transition and submission
- Refactor EditAccountModal to precheck before update submission
- Replace pendingPayload pattern with action-based dialog flow
Extract applyInterceptWarmup utility to unify all credential building
call sites:
- Fix upstream account creation missing intercept_warmup_requests write
- Fix apikey edit mode missing else-branch to clear the setting
- Add backend unit test for IsInterceptWarmupEnabled
- Add frontend unit test for credentialsBuilder
Before this change, when a client disconnected mid-request, the error
message was "Upstream request failed after retries", which is misleading
and pollutes error logs. Now we check context.Err() to return a more
accurate "Client disconnected" message for both Claude and Gemini
forward paths.
