The models gemini-2.5-flash-preview-image-generation and
gemini-3.1-flash-image-preview were surfacing in the model catalog
via getAllImageModels() from imageRegistry.ts, not from the live
upstream API. Removed them from the image provider registry.
- Reduce maxOutputTokens from 131072 to 65535 for gemini-3.1-pro-high
and gemini-3.1-pro-low, fixing 400 "invalid argument" errors from
Open WebUI when no max_tokens is specified (upstream limit is 65535)
- Filter non-viable models (gemini-3.1-flash-image-preview,
gemini-2.5-flash-preview-image-generation, gemini-3-pro-high/low)
from the live upstream API response in /api/providers/[id]/models
Models removed from available list (not usable via chat completions):
- gemini-3-pro-high/low — returns empty responses, quota unusable
- gemini-2.5-flash/flash-lite — quota always exhausted on free tier
- gemini-3.1-flash-image-preview — preview variant, not functional
Models hidden from quota UI (in addition to above):
- gemini-3-flash-agent — internal agent model
- gemini-3.1-flash-lite — not usable for chat
Kept gemini-3.1-flash-image in available models (confirmed working).
Removed dead gemini-3-pro from bare Pro ID normalization.
- Standardize cooldown fallback to 2 min (COOLDOWN_MS.rateLimit) in both
chatCore and auth.ts instead of inconsistent 60s/undefined
- Return 504 Gateway Timeout instead of 200 OK when SSE collection times
out, with finish_reason "length" to signal incomplete response
A 429 from one Antigravity model was marking the entire provider connection
as rate-limited, blocking ALL other models on the same account. This happened
in two places: chatCore's error classification (primary) and
markAccountUnavailable (secondary). Both now use model-only lockModel() for
passthrough providers instead of connection-wide rateLimitedUntil.
Also adds:
- Bare Pro model ID normalization (gemini-3-pro → gemini-3-pro-low)
matching OpenClaw convention
- Internal model exclusion list for quota display, matching CLIProxyAPI
- Update model list to match CLIProxyAPI filtered models (remove stale
gemini-2.5-pro, claude-sonnet-4-5, claude-sonnet-4, gemini-2.0-flash;
sort alphabetically)
- Extend 404 model-only lockout to passthrough providers so one missing
model doesn't lock out the entire Antigravity connection
- Always use streaming upstream endpoint (generateContent causes upstream
400 for some models that internally convert to OpenAI format with
stream_options); collect SSE into JSON for non-streaming clients
- Fix unlimited quota models showing 0% by checking q.unlimited before
recalculating percentage
Models without quota data (e.g. tab-completion models) were showing 0%
because remainingFraction defaulted to 0 when absent. Now defaults to 1
so they show 100% remaining instead.
The check-route-validation script now accepts both validateBody()
and .safeParse() as valid body validation methods. This fixes false
positives for routes using Zod schemas with safeParse().
Added 130 missing keys from en.json:
- a2aDashboard: 46 keys
- agents: 6+ keys
- cliTools.guides notes: continue, kiro, opencode
- And all other missing keys from recent additions
Total: All 33 language files now have full key parity.
Detects mismatched placeholders like {count} vs {pocet} between
source (en.json) and translations. Catches cases where raw placeholders
like {# models} are translated without preserving the placeholder format.
Found 14 issues in cs.json as test case.
Added missing cliTools.guides.windsurf.steps[1-5] with title and desc:
- step 1: Open AI Settings
- step 2: Add Custom Provider
- step 3: Base URL (http://127.0.0.1:20128/v1)
- step 4: API Key
- step 5: Select Model
Total: 165 keys across all language files (5 steps × 2 keys × 33 languages)
Added missing i18n keys for 'strict-random' routing strategy:
- combos.strategyGuide.strict-random: {when, avoid, example}
- combos.strategyRecommendations.strict-random: {title, description, tip1, tip2, tip3}
Total: 264 keys across all language files (8 keys × 33 languages)
These keys were already in pt-BR (incorrectly translated) and are now
aligned with the English fallback values from combos/page.tsx
Added missing step 5 'Use Thinking Variant' to all 33 i18n language files
for cliTools.guides.opencode.steps.5
The step was already defined in CLI_TOOLS constant but the i18n
translations were missing, causing the step title/description to
not display in the UI.
- Replace stale model IDs (gemini-3.1-pro-preview, gemini-3.1-flash-lite-preview)
with correct High/Low tier variants from fetchAvailableModels API
(gemini-3-pro-high, gemini-3-pro-low, gemini-3.1-pro-high, gemini-3.1-pro-low, etc.)
- Remove ag/ alias prefix in favor of antigravity/ across registry, providers,
model capabilities, combos, docs, and static model providers
- Make provider alias optional in Zod schema and guard ALIAS_TO_ID/ID_TO_ALIAS maps
- Show raw model IDs in quota display instead of unmapped display names
- Update T28 model catalog test to assert new High/Low tier models
diegosouzapw