@oonacode/win32-x64 0.3.200 → 0.3.202

package/oona.exe

@@ -1,4 +1,4 @@
 [diffend] Oversized file quarantined before diffing.
 name: package/oona.exe
-size: 111896064 bytes
-sha256: e1c1e79cfd958c3a000b522936ea14a7941b14dd9acf89093779977761a5bf29
+size: 111949824 bytes
+sha256: a111042daf85fbdaa6541cfe23befaaa857738b7dcabb88bc0932cbd8fc84c98

package/oonacode-release.json

@@ -1,13 +1,13 @@
 {
     "name":  "oonacode",
-    "version":  "0.3.200",
+    "version":  "0.3.202",
     "channel":  "stable",
     "platform":  "win32",
     "arch":  "x64",
     "mode":  "bun",
     "launcherRuntime":  "bun 1.4.1",
-    "commit":  "43006efd",
-    "builtAt":  "2026-10-02T20:50:39.8856110Z",
+    "commit":  "c6065559",
+    "builtAt":  "2026-10-03T16:45:37.8046229Z",
     "engine":  "engine-store",
     "engineVersion":  "0.3.280",
     "siteOrigin":  "https://oonacode.oonak.ai",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@oonacode/win32-x64",
-  "version": "0.3.200",
+  "version": "0.3.202",
   "description": "OonaCode native binaries for Windows x64: oona.exe (CLI + headless host service), the pinned oonaclaude.exe engine sidecar, local web UI assets. Installed automatically as an optional dependency of `oonacode`.",
   "license": "UNLICENSED",
   "homepage": "https://oonacode.oonak.ai",

package/registry.json

@@ -1240,33 +1240,10 @@
         "header": "authorization-bearer"
       },
       "models": [
-        {
-          "id": "gemini-3.1-pro-high",
-          "displayName": "Gemini 3.1 Pro High (Antigravity)",
-          "contextWindow": 1048576,
-          "maxOutputTokens": 65536,
-          "supportsTools": true,
-          "supportsThinking": true,
-          "supportsVision": true,
-          "inputPricePerM": 4,
-          "outputPricePerM": 18,
-          "notes": "Gemini 3.1 Pro at Antigravity's HIGH thinking level (the level is in the id, so no thinkingLevel is sent). Cost = Google's own Gemini API list price for gemini-3.1-pro-preview (ai.google.dev/gemini-api/docs/pricing, updated 2026-10-01): $2 in / $0.20 cached / $12 out up to a 200k-token prompt, $4 / $0.40 / $18 above. This file carries the TOP tier; the cheaper one is `models.price_tier_rule` (migration 0156). The wire id follows Antigravity's naming and is to be confirmed against the account's fetchAvailableModels list, which the admin sign-in writes into the credential's note."
-        },
-        {
-          "id": "gemini-3.1-pro-low",
-          "displayName": "Gemini 3.1 Pro Low (Antigravity)",
-          "contextWindow": 1048576,
-          "maxOutputTokens": 65536,
-          "supportsTools": true,
-          "supportsThinking": true,
-          "supportsVision": true,
-          "inputPricePerM": 4,
-          "outputPricePerM": 18,
-          "notes": "Gemini 3.1 Pro at Antigravity's LOW thinking level. Same Google list price as the High id (gemini-3.1-pro-preview, tiered at a 200k-token prompt — migration 0156); fewer thinking tokens is the whole difference."
-        },
         {
           "id": "gemini-3.8-flash",
           "displayName": "Gemini 3.8 Flash (Antigravity)",
+          "upstreamId": "gemini-3.8-flash-{effort:high}",
           "contextWindow": 1048576,
           "maxOutputTokens": 65536,
           "supportsTools": true,
@@ -1274,56 +1251,23 @@
           "supportsVision": true,
           "inputPricePerM": 0.75,
           "outputPricePerM": 3.75,
-          "notes": "Antigravity's default model (the google-antigravity SDK's DEFAULT_MODEL, docs/antigravity.md section 3). Cost = Google's Gemini API list price for gemini-3.8-flash: $0.75 in / $0.075 cached / $3.75 out per 1M through 2026-12-31, DOUBLING to $1.50 / $0.15 / $7.50 from 2027-01-01 (ai.google.dev/gemini-api/docs/pricing, updated 2026-10-01) — move the cost cells on the model page then."
+          "notes": "Google lists one model per level — gemini-3.8-flash-high, -medium and -low (fetchAvailableModels, 2026-10-03: 1,048,576 context / 65,536 output each); OonaCode shows ONE model and the chat’s effort picker fills the level (upstreamId template; xhigh and max read as high; no effort = high). Cost = Google’s Gemini API list price for gemini-3.8-flash: $0.75 in / $0.075 cached / $3.75 out per 1M through 2026-12-31, DOUBLING to $1.50 / $0.15 / $7.50 from 2027-01-01 (ai.google.dev/gemini-api/docs/pricing, updated 2026-10-01) — move the cost cells on the model page then."
         },
         {
-          "id": "claude-sonnet-4-6",
-          "displayName": "Claude Sonnet 4.6 (Antigravity)",
-          "contextWindow": 1000000,
-          "maxOutputTokens": 64000,
-          "supportsTools": true,
-          "supportsThinking": false,
-          "supportsVision": true,
-          "inputPricePerM": 3,
-          "outputPricePerM": 15,
-          "notes": "Claude Sonnet 4.6 hosted by Google and reached through Antigravity's Code Assist backend in the Gemini request shape. Cost = ANTHROPIC's own list price for Sonnet 4.6 ($3 in / $0.30 cache read / $3.75 5-minute write / $15 out, 1M context at standard rates — platform.claude.com/docs/en/about-claude/pricing, read 2026-10-02): Google publishes no Antigravity price for it."
-        },
-        {
-          "id": "claude-sonnet-4-6-thinking",
-          "displayName": "Claude Sonnet 4.6 Thinking (Antigravity)",
-          "contextWindow": 1000000,
-          "maxOutputTokens": 64000,
-          "supportsTools": true,
-          "supportsThinking": true,
-          "supportsVision": true,
-          "inputPricePerM": 3,
-          "outputPricePerM": 15,
-          "notes": "Claude Sonnet 4.6 with extended thinking, through Antigravity. The thinking budget is sent as thinkingConfig.thinkingBudget; Claude's signed thinking comes back as thought parts and is echoed with its signature (the translator carries it in the Anthropic thinking block). Cost = Anthropic's list price for Sonnet 4.6."
-        },
-        {
-          "id": "claude-opus-4-6-thinking",
-          "displayName": "Claude Opus 4.6 Thinking (Antigravity)",
-          "contextWindow": 1000000,
-          "maxOutputTokens": 128000,
+          "id": "gemini-3.7-flash",
+          "displayName": "Gemini 3.7 Flash (Antigravity)",
+          "upstreamId": "gemini-3.7-flash-{effort:high}",
+          "contextWindow": 1048576,
+          "maxOutputTokens": 65536,
           "supportsTools": true,
           "supportsThinking": true,
           "supportsVision": true,
-          "inputPricePerM": 5,
-          "outputPricePerM": 25,
-          "notes": "Claude Opus 4.6 with extended thinking, through Antigravity. Cost = ANTHROPIC's list price for Opus 4.6 ($5 in / $0.50 cache read / $6.25 5-minute write / $25 out, 1M context at standard rates — read 2026-10-02)."
-        },
-        {
-          "id": "gpt-oss-120b-medium",
-          "displayName": "GPT-OSS 120B (Antigravity)",
-          "contextWindow": 400000,
-          "maxOutputTokens": 32768,
-          "supportsTools": true,
-          "supportsThinking": true,
-          "supportsVision": false,
-          "notes": "OpenAI's open-weight GPT-OSS 120B, hosted by Google, at Antigravity's medium reasoning level (in the id). NO PRICE: Google's Vertex AI list price for gpt-oss-120b could not be read on 2026-10-02, so the row is left unpriced and the pool refuses it (pool_model_unpriced) until the owner sets its cost on the model page. The 400K window is the one docs/antigravity.md section 3 records."
+          "inputPricePerM": 0.75,
+          "outputPricePerM": 3.75,
+          "notes": "Google lists one model per level — gemini-3.7-flash-high, -medium and -low (fetchAvailableModels, 2026-10-03: 1,048,576 context / 65,536 output each); OonaCode shows ONE model and the chat’s effort picker fills the level (upstreamId template; xhigh and max read as high; no effort = high). Cost = Google’s Gemini API list price for gemini-3.7-flash: $0.75 in / $0.075 cached / $3.75 out per 1M through 2026-12-31, DOUBLING to $1.50 / $0.15 / $7.50 from 2027-01-01 (ai.google.dev/gemini-api/docs/pricing, updated 2026-10-01) — move the cost cells on the model page then."
         }
       ],
-      "defaultModel": "gemini-3.1-pro-high",
+      "defaultModel": "gemini-3.8-flash",
       "capabilities": {
         "streaming": true,
         "tools": true,
@@ -1374,9 +1318,7 @@
           "operatorOnly": true,
           "baseUrl": "https://daily-cloudcode-pa.googleapis.com",
           "extraHeaders": {
-            "user-agent": "antigravity/2.19.1 windows/amd64",
-            "x-goog-api-client": "google-cloud-sdk vscode_cloudshelleditor/0.1",
-            "client-metadata": "{\"ideType\":\"IDE_UNSPECIFIED\",\"platform\":\"PLATFORM_UNSPECIFIED\",\"pluginType\":\"GEMINI\"}"
+            "user-agent": "antigravity/cli/1.2.15 (aidev_client; os_type=windows; arch=amd64; cl=992008527; auth_method=consumer)"
           },
           "loopbackLogin": {
             "type": "google-loopback",
@@ -1395,9 +1337,63 @@
             "userinfoUrl": "https://www.googleapis.com/oauth2/v2/userinfo",
             "signInLabel": "Sign in with Google (Antigravity)"
           },
-          "notes": "The operator's Antigravity subscription (Google AI Pro / Ultra or the free tier). The base URL is the daily Code Assist channel the installed Antigravity 2.19.1 talks to (docs/antigravity.md section 1); prod is https://cloudcode-pa.googleapis.com — the row's base URL decides, so a credential can be pointed there. The identity headers are Antigravity's (user-agent with its version — bump it with Antigravity — and the Code Assist client metadata the Antigravity-compatible clients send). Scopes: the six the language server carries minus `aicode`, which the public Antigravity sign-in integrations do not request; add it here if Google ever refuses a Code Assist call for want of it."
+          "notes": "The operator’s Antigravity account. The base URL is the daily Code Assist channel Antigravity itself talks to (its language server runs with --cloud_code_endpoint https://daily-cloudcode-pa.googleapis.com; prod is https://cloudcode-pa.googleapis.com — the row’s base URL decides). The ONE identity header is the user-agent Google’s own agy CLI sends (1.2.15, recorded again on 2026-10-03 after it updated; docs/antigravity.md section 12) — bump its version with Antigravity’s. Scopes: the six the language server carries minus `aicode`, which the public Antigravity sign-in integrations do not request; add it here if Google ever refuses a Code Assist call for want of it."
         }
-      ]
+      ],
+      "effortLevels": ["low", "medium", "high"]
+    },
+    {
+      "id": "microsoft",
+      "displayName": "Microsoft (MAI)",
+      "kind": "openai-compatible",
+      "route": "gateway",
+      "credentialClass": "api_key",
+      "poolOnly": true,
+      "baseUrl": "https://pars5555-2864-resource.services.ai.azure.com/mai",
+      "messagesPath": "/v1/chat/completions",
+      "auth": { "header": "api-key" },
+      "models": [
+        {
+          "id": "MAI-Thinking-1",
+          "displayName": "MAI-Thinking-1 (preview)",
+          "contextWindow": 256000,
+          "maxOutputTokens": 64000,
+          "supportsTools": true,
+          "supportsThinking": false,
+          "supportsVision": false,
+          "inputPricePerM": 2,
+          "outputPricePerM": 8,
+          "notes": "Microsoft AI's own reasoning model (35B active / ~1T total MoE), public preview, sold Direct from Azure. The id is the Foundry DEPLOYMENT name (deployed 2026-10-03 on the operator's resource pars5555-2864-resource, Global Standard, 500K TPM / 500 RPM). It reasons internally and never returns its thoughts (only an encrypted blob on request, not carried yet), so no thinking is shown and there are no effort levels. Cost = Microsoft's list price, $2 in / $8 out per 1M. Version 2026-06-01 retires 2026-11-04."
+        }
+      ],
+      "defaultModel": "MAI-Thinking-1",
+      "capabilities": {
+        "streaming": true,
+        "tools": true,
+        "thinking": false,
+        "vision": false,
+        "countTokens": false,
+        "promptCaching": true
+      },
+      "translate": {
+        "maxTokensParam": "max_completion_tokens",
+        "reasoningEffortFromBudget": false,
+        "reasoningContentField": null,
+        "supportsParallelToolCalls": true,
+        "systemRole": "system",
+        "streamIncludeUsage": true
+      },
+      "phase": 1,
+      "enabled": true,
+      "docsUrl": "https://learn.microsoft.com/en-us/azure/foundry/foundry-models/how-to/use-foundry-models-mai-thinking",
+      "references": [
+        {
+          "label": "Microsoft Foundry pricing (Microsoft models)",
+          "url": "https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/microsoft/",
+          "kind": "pricing"
+        }
+      ],
+      "notes": "OPERATOR POOL ONLY, and HIDDEN (owner, 2026-10-03: 'i wana have it hidden by setting custom model in my sessions, not public for all, i just need to test'): no model row is pooled or on the API list, so nobody is offered it; the owner reaches it with the hidden custom-model command, prefix microsoft/. First-party: Microsoft trains and serves MAI models. Use this resource for Microsoft's own models only: the same Foundry catalogue resells other companies' models, which would make it a broker (guardrail 8). Auth: the resource's key in the api-key header."
     }
   ]
 }