@oh-my-pi/pi-catalog 16.1.2 → 16.1.4

package/CHANGELOG.md

@@ -2,6 +2,19 @@
 
 ## [Unreleased]
 
+## [16.1.4] - 2026-06-19
+
+### Fixed
+
+- Fixed Claude 4.6 routing on the `google-antigravity` (and `google-gemini-cli`) Cloud Code Assist providers, whose backend exposes the models asymmetrically: `claude-sonnet-4-6` has no `-thinking` twin and `claude-opus-4-6` has only the `-thinking` twin. The shared `thinkingPair` family was routing thinking efforts on `claude-sonnet-4-6` to a non-existent `claude-sonnet-4-6-thinking` wire id (404 `Requested entity was not found`); replaced both 4.6 entries with bespoke single-wire families so every effort and off resolve to the live wire id. Added `claude-sonnet-4-6` and `claude-opus-4-6-thinking` entries to `ANTIGRAVITY_MODEL_WIRE_PROFILES` capped at the backend's 64000-output-token limit (over-cap requests 400'd with `Request contains an invalid argument`); `modelEnum` is now optional on `AntigravityModelWireProfile` since the Claude wire ids are accepted without a captured `labels.model_enum`. ([#3067](https://github.com/can1357/oh-my-pi/issues/3067))
+
+## [16.1.3] - 2026-06-19
+
+### Fixed
+
+- Marked Ollama Cloud catalog models to omit on-the-wire output-token caps, preventing context-window-sized `num_predict` values from causing HTTP 400s for models whose true output cap is not discoverable. ([#2984](https://github.com/can1357/oh-my-pi/issues/2984))
+- Fixed `readModelCache`/`writeModelCache` using a process-global shared database even when a custom `dbPath` was provided. Custom-path cache operations now open and close a per-call database via `withModelCacheDb`, preventing leaked SQLite handles on Windows
+
 ## [16.1.2] - 2026-06-19
 
 ### Added

package/dist/types/types.d.ts

@@ -477,6 +477,11 @@ export interface Model<TApi extends Api = Api> {
     baseUrl: string;
     reasoning: boolean;
     input: ("text" | "image")[];
+    /**
+     * Decoder family used for image inputs when it has narrower format support
+     * than OMP's general image pipeline. `stb` local backends reject WebP.
+     */
+    imageInputDecoder?: "stb";
     /**
      * Native provider tool-call support. `false` is the only unsupported signal:
      * `true` and `undefined` both mean callers may use native tools. Catalog and

package/dist/types/wire/gemini-headers.d.ts

@@ -19,15 +19,18 @@ export declare let getAntigravityUserAgent: () => string;
 /**
  * Per-wire-id Antigravity Cloud Code Assist request constants, captured from the
  * real `antigravity/hub` client against `daily-cloudcode-pa`. `modelEnum` is the
- * opaque `labels.model_enum` token the client tags each request with;
- * `maxOutputTokens` is the fixed `generationConfig.maxOutputTokens` it sends
- * regardless of the thinking budget. Keyed by the routed upstream wire id
- * (post effort-routing), not the collapsed logical id. Checkpoint-only ids
+ * opaque `labels.model_enum` token the client tags each request with — optional
+ * because Anthropic-backed wire ids (e.g. `claude-sonnet-4-6`,
+ * `claude-opus-4-6-thinking`) are accepted without one; the label is purely
+ * telemetry. `maxOutputTokens` is the fixed `generationConfig.maxOutputTokens`
+ * the backend enforces regardless of the thinking budget (Claude caps at
+ * 64000, Gemini accepts the discovered cap). Keyed by the routed upstream wire
+ * id (post effort-routing), not the collapsed logical id. Checkpoint-only ids
  * (e.g. `gemini-3.1-flash-lite`) are intentionally absent — this provider only
  * emits agent requests.
  */
 export interface AntigravityModelWireProfile {
-    modelEnum: string;
+    modelEnum?: string;
     maxOutputTokens: number;
 }
 export declare const ANTIGRAVITY_MODEL_WIRE_PROFILES: Readonly<Record<string, AntigravityModelWireProfile>>;

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "16.1.2",
+	"version": "16.1.4",
 	"description": "Model catalog for omp: bundled model database, provider discovery descriptors, model identity, classification, and equivalence",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
@@ -34,12 +34,12 @@
 	},
 	"dependencies": {
 		"@bufbuild/protobuf": "^2.12.0",
-		"@oh-my-pi/pi-utils": "16.1.2",
+		"@oh-my-pi/pi-utils": "16.1.4",
 		"arktype": "^2.2.0",
 		"zod": "^4"
 	},
 	"devDependencies": {
-		"@oh-my-pi/pi-ai": "16.1.2",
+		"@oh-my-pi/pi-ai": "16.1.4",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/model-cache.ts

@@ -46,14 +46,7 @@ interface CacheEntry<TApi extends Api = Api> {
 let sharedDb: Database | null = null;
 let sharedDbPath: string | null = null;
 
-function getDb(dbPath?: string): Database {
-	const resolvedPath = dbPath ?? getModelDbPath();
-	if (sharedDb && sharedDbPath === resolvedPath) {
-		return sharedDb;
-	}
-	if (sharedDb) {
-		sharedDb.close();
-	}
+function openDb(resolvedPath: string): Database {
 	const db = new Database(resolvedPath, { create: true });
 	// Install the busy handler BEFORE any lock-taking statement. See
 	// https://github.com/can1357/oh-my-pi/issues/2421.
@@ -70,16 +63,42 @@ function getDb(dbPath?: string): Database {
 		)
 	`);
 	migrateCacheSchema(db);
+	return db;
+}
 
+function getSharedDb(): Database {
+	const resolvedPath = getModelDbPath();
+	if (sharedDb && sharedDbPath === resolvedPath) {
+		return sharedDb;
+	}
+	if (sharedDb) {
+		sharedDb.close();
+	}
+	const db = openDb(resolvedPath);
 	sharedDb = db;
 	sharedDbPath = resolvedPath;
 	return db;
 }
 
+function withModelCacheDb<T>(dbPath: string | undefined, useDb: (db: Database) => T): T {
+	if (!dbPath) return useDb(getSharedDb());
+	const db = openDb(dbPath);
+	try {
+		return useDb(db);
+	} finally {
+		db.close();
+	}
+}
+
 function migrateCacheSchema(db: Database): void {
-	const columns = db.prepare("PRAGMA table_info(model_cache)").all() as TableInfoRow[];
-	if (!columns.some(column => column.name === "static_fingerprint")) {
-		db.run("ALTER TABLE model_cache ADD COLUMN static_fingerprint TEXT NOT NULL DEFAULT ''");
+	const stmt = db.prepare("PRAGMA table_info(model_cache)");
+	try {
+		const columns = stmt.all() as TableInfoRow[];
+		if (!columns.some(column => column.name === "static_fingerprint")) {
+			db.run("ALTER TABLE model_cache ADD COLUMN static_fingerprint TEXT NOT NULL DEFAULT ''");
+		}
+	} finally {
+		stmt.finalize();
 	}
 	db.run("UPDATE model_cache SET version = ? WHERE version = 2", [CACHE_SCHEMA_VERSION]);
 }
@@ -91,21 +110,27 @@ export function readModelCache<TApi extends Api>(
 	dbPath?: string,
 ): CacheEntry<TApi> | null {
 	try {
-		const db = getDb(dbPath);
-		const row = db.query<CacheRow, [string]>("SELECT * FROM model_cache WHERE provider_id = ?").get(providerId);
-		if (!row || row.version !== CACHE_SCHEMA_VERSION) {
-			return null;
-		}
-		const models = JSON.parse(row.models) as ModelSpec<TApi>[];
-		const ageMs = now() - row.updated_at;
-		const fresh = Number.isFinite(ageMs) && ageMs >= 0 && ageMs <= ttlMs;
-		return {
-			models,
-			fresh,
-			authoritative: row.authoritative === 1,
-			updatedAt: row.updated_at,
-			staticFingerprint: row.static_fingerprint ?? "",
-		};
+		return withModelCacheDb(dbPath, db => {
+			const stmt = db.query<CacheRow, [string]>("SELECT * FROM model_cache WHERE provider_id = ?");
+			try {
+				const row = stmt.get(providerId);
+				if (!row || row.version !== CACHE_SCHEMA_VERSION) {
+					return null;
+				}
+				const models = JSON.parse(row.models) as ModelSpec<TApi>[];
+				const ageMs = now() - row.updated_at;
+				const fresh = Number.isFinite(ageMs) && ageMs >= 0 && ageMs <= ttlMs;
+				return {
+					models,
+					fresh,
+					authoritative: row.authoritative === 1,
+					updatedAt: row.updated_at,
+					staticFingerprint: row.static_fingerprint ?? "",
+				};
+			} finally {
+				stmt.finalize();
+			}
+		});
 	} catch {
 		return null;
 	}
@@ -120,19 +145,20 @@ export function writeModelCache<TApi extends Api>(
 	dbPath?: string,
 ): void {
 	try {
-		const db = getDb(dbPath);
-		db.run(
-			`INSERT OR REPLACE INTO model_cache (provider_id, version, updated_at, authoritative, static_fingerprint, models)
-			 VALUES (?, ?, ?, ?, ?, ?)`,
-			[
-				providerId,
-				CACHE_SCHEMA_VERSION,
-				updatedAt,
-				authoritative ? 1 : 0,
-				staticFingerprint,
-				JSON.stringify(models.map(model => ({ ...model, compat: model.compatConfig, compatConfig: undefined }))),
-			],
-		);
+		withModelCacheDb(dbPath, db => {
+			db.run(
+				`INSERT OR REPLACE INTO model_cache (provider_id, version, updated_at, authoritative, static_fingerprint, models)
+				 VALUES (?, ?, ?, ?, ?, ?)`,
+				[
+					providerId,
+					CACHE_SCHEMA_VERSION,
+					updatedAt,
+					authoritative ? 1 : 0,
+					staticFingerprint,
+					JSON.stringify(models.map(model => ({ ...model, compat: model.compatConfig, compatConfig: undefined }))),
+				],
+			);
+		});
 	} catch {
 		// Cache writes are best-effort; failures should not break model resolution.
 	}

package/src/models.json

@@ -53624,6 +53624,7 @@
 			},
 			"contextWindow": 163840,
 			"maxTokens": 32000,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -53652,6 +53653,7 @@
 			},
 			"contextWindow": 163840,
 			"maxTokens": 163840,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -53680,6 +53682,7 @@
 			},
 			"contextWindow": 163840,
 			"maxTokens": 65536,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -53708,6 +53711,7 @@
 			},
 			"contextWindow": 1048576,
 			"maxTokens": 1048576,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -53737,6 +53741,7 @@
 			},
 			"contextWindow": 1048576,
 			"maxTokens": 1048576,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -53765,7 +53770,8 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 262144
+			"maxTokens": 262144,
+			"omitMaxOutputTokens": true
 		},
 		"devstral-small-2:24b": {
 			"id": "devstral-small-2:24b",
@@ -53785,7 +53791,8 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 262144
+			"maxTokens": 262144,
+			"omitMaxOutputTokens": true
 		},
 		"gemini-3-flash-preview": {
 			"id": "gemini-3-flash-preview",
@@ -53806,6 +53813,7 @@
 			},
 			"contextWindow": 1048576,
 			"maxTokens": 65536,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -53836,6 +53844,7 @@
 			},
 			"contextWindow": 262144,
 			"maxTokens": 262144,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -53864,6 +53873,7 @@
 			},
 			"contextWindow": 202752,
 			"maxTokens": 131072,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -53892,6 +53902,7 @@
 			},
 			"contextWindow": 202752,
 			"maxTokens": 131072,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -53920,6 +53931,7 @@
 			},
 			"contextWindow": 202752,
 			"maxTokens": 131072,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -53948,6 +53960,7 @@
 			},
 			"contextWindow": 202752,
 			"maxTokens": 131072,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -53976,6 +53989,7 @@
 			},
 			"contextWindow": 976000,
 			"maxTokens": 131072,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54005,6 +54019,7 @@
 			},
 			"contextWindow": 131072,
 			"maxTokens": 32768,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54033,6 +54048,7 @@
 			},
 			"contextWindow": 131072,
 			"maxTokens": 32768,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54061,6 +54077,7 @@
 			},
 			"contextWindow": 262144,
 			"maxTokens": 262144,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54089,7 +54106,8 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 262144
+			"maxTokens": 262144,
+			"omitMaxOutputTokens": true
 		},
 		"kimi-k2.5": {
 			"id": "kimi-k2.5",
@@ -54110,6 +54128,7 @@
 			},
 			"contextWindow": 262144,
 			"maxTokens": 262144,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54139,6 +54158,7 @@
 			},
 			"contextWindow": 262144,
 			"maxTokens": 262144,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54168,6 +54188,7 @@
 			},
 			"contextWindow": 262144,
 			"maxTokens": 262144,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54196,6 +54217,7 @@
 			},
 			"contextWindow": 204800,
 			"maxTokens": 128000,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54225,6 +54247,7 @@
 			},
 			"contextWindow": 204800,
 			"maxTokens": 131072,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54254,6 +54277,7 @@
 			},
 			"contextWindow": 204800,
 			"maxTokens": 131072,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54283,6 +54307,7 @@
 			},
 			"contextWindow": 196608,
 			"maxTokens": 196608,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54313,6 +54338,7 @@
 			},
 			"contextWindow": 512000,
 			"maxTokens": 131072,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54341,7 +54367,8 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 128000
+			"maxTokens": 128000,
+			"omitMaxOutputTokens": true
 		},
 		"ministral-3:3b": {
 			"id": "ministral-3:3b",
@@ -54361,7 +54388,8 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 128000
+			"maxTokens": 128000,
+			"omitMaxOutputTokens": true
 		},
 		"ministral-3:8b": {
 			"id": "ministral-3:8b",
@@ -54381,7 +54409,8 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 128000
+			"maxTokens": 128000,
+			"omitMaxOutputTokens": true
 		},
 		"mistral-large-3:675b": {
 			"id": "mistral-large-3:675b",
@@ -54401,7 +54430,8 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 262144
+			"maxTokens": 262144,
+			"omitMaxOutputTokens": true
 		},
 		"nemotron-3-nano:30b": {
 			"id": "nemotron-3-nano:30b",
@@ -54421,6 +54451,7 @@
 			},
 			"contextWindow": 1048576,
 			"maxTokens": 131072,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54449,6 +54480,7 @@
 			},
 			"contextWindow": 262144,
 			"maxTokens": 65536,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54477,6 +54509,7 @@
 			},
 			"contextWindow": 262144,
 			"maxTokens": 128000,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54504,7 +54537,8 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 65536
+			"maxTokens": 65536,
+			"omitMaxOutputTokens": true
 		},
 		"qwen3-coder:480b": {
 			"id": "qwen3-coder:480b",
@@ -54523,7 +54557,8 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 65536
+			"maxTokens": 65536,
+			"omitMaxOutputTokens": true
 		},
 		"qwen3-next:80b": {
 			"id": "qwen3-next:80b",
@@ -54543,6 +54578,7 @@
 			},
 			"contextWindow": 262144,
 			"maxTokens": 32768,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54572,6 +54608,7 @@
 			},
 			"contextWindow": 262144,
 			"maxTokens": 32768,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54600,7 +54637,8 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 131072
+			"maxTokens": 131072,
+			"omitMaxOutputTokens": true
 		},
 		"qwen3.5:397b": {
 			"id": "qwen3.5:397b",
@@ -54621,6 +54659,7 @@
 			},
 			"contextWindow": 262144,
 			"maxTokens": 65536,
+			"omitMaxOutputTokens": true,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -54648,7 +54687,8 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 32768,
-			"maxTokens": 4096
+			"maxTokens": 4096,
+			"omitMaxOutputTokens": true
 		}
 	},
 	"openai": {
@@ -84815,4 +84855,4 @@
 			}
 		}
 	}
-}
+}

package/src/provider-models/ollama.ts

@@ -159,6 +159,7 @@ export function ollamaCloudModelManagerOptions(
 							discoveredContextWindow !== null && discoveredContextWindow !== undefined
 								? (providerReference?.maxTokens ?? Math.min(contextWindow, 8192))
 								: Math.min(contextWindow, 8192),
+						omitMaxOutputTokens: true,
 					};
 				}),
 			);

package/src/types.ts

@@ -594,6 +594,11 @@ export interface Model<TApi extends Api = Api> {
 	baseUrl: string;
 	reasoning: boolean;
 	input: ("text" | "image")[];
+	/**
+	 * Decoder family used for image inputs when it has narrower format support
+	 * than OMP's general image pipeline. `stb` local backends reject WebP.
+	 */
+	imageInputDecoder?: "stb";
 	/**
 	 * Native provider tool-call support. `false` is the only unsupported signal:
 	 * `true` and `undefined` both mean callers may use native tools. Catalog and

package/src/variant-collapse.ts

@@ -220,8 +220,26 @@ const SHARED_CCA_FAMILIES: readonly EffortVariantFamily[] = [
 		routing: {},
 		thinking: { mode: "budget", efforts: [Effort.Minimal, Effort.Low, Effort.Medium, Effort.High] },
 	},
-	thinkingPair("claude-sonnet-4-6", "Claude Sonnet 4.6"),
-	thinkingPair("claude-opus-4-6", "Claude Opus 4.6"),
+	// Antigravity Cloud Code Assist exposes Claude 4.6 asymmetrically: only the
+	// bare `claude-sonnet-4-6` wire id (no `-thinking` twin) and only the
+	// `claude-opus-4-6-thinking` wire id (no bare twin). Per-effort thinking is
+	// carried in the request body via `thinkingBudget`, so both ids accept on/off
+	// requests. Listing both candidates in `members` (priority order) keeps the
+	// collapse correct if the backend mix ever rebalances.
+	{
+		id: "claude-sonnet-4-6",
+		name: "Claude Sonnet 4.6",
+		members: ["claude-sonnet-4-6", "claude-sonnet-4-6-thinking"],
+		routing: {},
+		thinking: { mode: "budget", efforts: [Effort.Minimal, Effort.Low, Effort.Medium, Effort.High] },
+	},
+	{
+		id: "claude-opus-4-6",
+		name: "Claude Opus 4.6",
+		members: ["claude-opus-4-6-thinking", "claude-opus-4-6"],
+		routing: {},
+		thinking: { mode: "budget", efforts: [Effort.Minimal, Effort.Low, Effort.Medium, Effort.High] },
+	},
 	thinkingPair("claude-sonnet-4-5", "Claude Sonnet 4.5"),
 	thinkingPair("claude-opus-4-5", "Claude Opus 4.5"),
 	thinkingPair("gemini-2.5-flash", "Gemini 2.5 Flash"),

package/src/wire/gemini-headers.ts

@@ -43,15 +43,18 @@ export let getAntigravityUserAgent = () => {
 /**
  * Per-wire-id Antigravity Cloud Code Assist request constants, captured from the
  * real `antigravity/hub` client against `daily-cloudcode-pa`. `modelEnum` is the
- * opaque `labels.model_enum` token the client tags each request with;
- * `maxOutputTokens` is the fixed `generationConfig.maxOutputTokens` it sends
- * regardless of the thinking budget. Keyed by the routed upstream wire id
- * (post effort-routing), not the collapsed logical id. Checkpoint-only ids
+ * opaque `labels.model_enum` token the client tags each request with — optional
+ * because Anthropic-backed wire ids (e.g. `claude-sonnet-4-6`,
+ * `claude-opus-4-6-thinking`) are accepted without one; the label is purely
+ * telemetry. `maxOutputTokens` is the fixed `generationConfig.maxOutputTokens`
+ * the backend enforces regardless of the thinking budget (Claude caps at
+ * 64000, Gemini accepts the discovered cap). Keyed by the routed upstream wire
+ * id (post effort-routing), not the collapsed logical id. Checkpoint-only ids
  * (e.g. `gemini-3.1-flash-lite`) are intentionally absent — this provider only
  * emits agent requests.
  */
 export interface AntigravityModelWireProfile {
-	modelEnum: string;
+	modelEnum?: string;
 	maxOutputTokens: number;
 }
 export const ANTIGRAVITY_MODEL_WIRE_PROFILES: Readonly<Record<string, AntigravityModelWireProfile>> = {
@@ -60,6 +63,11 @@ export const ANTIGRAVITY_MODEL_WIRE_PROFILES: Readonly<Record<string, Antigravit
 	"gemini-3-flash-agent": { modelEnum: "MODEL_PLACEHOLDER_M132", maxOutputTokens: 65536 },
 	"gemini-3.1-pro-low": { modelEnum: "MODEL_PLACEHOLDER_M36", maxOutputTokens: 65535 },
 	"gemini-pro-agent": { modelEnum: "MODEL_PLACEHOLDER_M16", maxOutputTokens: 65535 },
+	// Claude on `daily-cloudcode-pa` rejects `maxOutputTokens > 64000` with a
+	// 400 (`Request contains an invalid argument`). The model_enum label is
+	// untracked for these ids; the backend does not require it.
+	"claude-sonnet-4-6": { maxOutputTokens: 64000 },
+	"claude-opus-4-6-thinking": { maxOutputTokens: 64000 },
 };
 export function getAntigravityModelWireProfile(wireModelId: string): AntigravityModelWireProfile | undefined {
 	return ANTIGRAVITY_MODEL_WIRE_PROFILES[wireModelId];