@oh-my-pi/pi-catalog 16.1.1 → 16.1.3

package/CHANGELOG.md

@@ -2,6 +2,31 @@
 
 ## [Unreleased]
 
+## [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
+
+- Added support for Gemini 2.5 Flash-Lite, 3.1 Flash-Lite, and 3.5 Flash models
+- Added support for Moonshot V1 model family
+
+### Changed
+
+- Updated context window and token limits for various Claude, Gemini, and GPT-OSS models
+- Refined thinking mode behaviors and routing for supported LLM families
+
+### Fixed
+
+- Fixed GLM-5.2 `reasoning_effort` so the top thinking tier reaches each host's genuine maximum instead of 400ing, mapping the internal `xhigh` tier per host dialect (verified against live endpoints): Z.ai/Zhipu collapse onto the model's `none`/`high`/`max` scale (`xhigh → max`); Fireworks, resellers, and Ollama Cloud keep their distinct lower tiers and remap only the top `xhigh → max` (merged over host quirks such as Fireworks' `minimal → none`); and OpenRouter — whose API rejects `max` and treats `xhigh` as its own max tier — now exposes the `xhigh` tier and forwards it verbatim. Dialect detection keys off resolved `compat.thinkingFormat`, so custom OpenRouter/Z.ai-format providers are covered too.
+- Maintained thinking effort routing when discovery only returns the base model ID
+- Improved credential retrieval logic for Antigravity and Codex providers via auth discovery
+
 ## [16.0.9] - 2026-06-18
 
 ### Fixed

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/variant-collapse.d.ts

@@ -40,6 +40,13 @@ export interface EffortVariantFamily {
     thinking: Readonly<Omit<ThinkingConfig, "effortRouting" | "suppressWhenOff">>;
     /** Thinking-off requests must explicitly suppress thinking on the wire. */
     suppressWhenOff?: boolean;
+    /**
+     * Preserve non-off effort routes even when discovery omits the backing member.
+     * Used for Cloud Code Assist `X`/`X-thinking` pairs where upstream accepts
+     * the `-thinking` wire id but the model-list endpoint may advertise only the
+     * bare id.
+     */
+    preserveAbsentEffortRoutes?: boolean;
     /** Retired/recycled selector ids that alias to this family without being members. */
     extraAliases?: readonly string[];
 }

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "16.1.1",
+	"version": "16.1.3",
 	"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": "16.1.1",
+		"@oh-my-pi/pi-utils": "16.1.3",
 		"arktype": "^2.2.0",
 		"zod": "^4"
 	},
 	"devDependencies": {
-		"@oh-my-pi/pi-ai": "16.1.1",
+		"@oh-my-pi/pi-ai": "16.1.3",
 		"@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/model-thinking.ts

@@ -86,7 +86,7 @@ const ZAI_GLM_52_REASONING_EFFORT_MAP: Readonly<EffortMap> = {
 	[Effort.High]: "high",
 	[Effort.XHigh]: "max",
 };
-const OLLAMA_CLOUD_GLM_52_REASONING_EFFORT_MAP: Readonly<EffortMap> = {
+const GLM_52_XHIGH_MAX_EFFORT_MAP: Readonly<EffortMap> = {
 	[Effort.XHigh]: "max",
 };
 
@@ -164,7 +164,7 @@ function fillThinkingWireDefaults<TApi extends Api>(
 	thinking: ThinkingConfig,
 ): ThinkingConfig {
 	const parsed = parseKnownModel(spec.id);
-	const normalizedEfforts = getModelDefinedEfforts(spec) ?? thinking.efforts;
+	const normalizedEfforts = getModelDefinedEfforts(spec, compat) ?? thinking.efforts;
 	const effortsChanged = !sameEffortList(normalizedEfforts, thinking.efforts);
 	const effortMap =
 		thinking.effortMap === undefined
@@ -251,7 +251,7 @@ function inferEffortMap<TApi extends Api>(
 	mode: ThinkingConfig["mode"],
 	efforts: readonly Effort[],
 ): EffortMap | undefined {
-	const detected = inferDetectedEffortMap(spec, parsedModel, mode);
+	const detected = inferDetectedEffortMap(spec, compat, parsedModel, mode);
 	const configured = readCompatEffortMap(compat);
 	const merged =
 		detected === undefined ? configured : configured === undefined ? detected : { ...detected, ...configured };
@@ -281,23 +281,26 @@ function isOpenAICompatReasoningApi(api: Api): boolean {
 	return api === "openai-completions" || api === "openrouter";
 }
 
-function getModelDefinedEfforts<TApi extends Api>(spec: ModelSpec<TApi>): readonly Effort[] | undefined {
-	if (isOpenAICompatReasoningApi(spec.api) && isZaiGlm52ReasoningEffortModel(spec)) {
-		return DEFAULT_REASONING_EFFORTS_WITH_XHIGH;
-	}
-	if (isOllamaCloudGlm52ReasoningEffortModel(spec)) {
-		return GLM_52_HIGH_MAX_REASONING_EFFORTS;
+function getModelDefinedEfforts<TApi extends Api>(
+	spec: ModelSpec<TApi>,
+	compat: CompatOf<TApi>,
+): readonly Effort[] | undefined {
+	if (isGlm52ReasoningEffortModelId(spec.id)) {
+		// Z.ai/Zhipu and OpenRouter both surface GLM-5.2's full effort ladder,
+		// including the top `xhigh` (= "max") tier; Ollama Cloud exposes only
+		// high/xhigh.
+		if (isZaiThinkingFormat(compat) || isOpenRouterThinkingFormat(compat)) {
+			return DEFAULT_REASONING_EFFORTS_WITH_XHIGH;
+		}
+		if (isOllamaCloudGlm52ReasoningEffortModel(spec)) {
+			return GLM_52_HIGH_MAX_REASONING_EFFORTS;
+		}
 	}
 	return isOpenAICompatReasoningApi(spec.api) && (isMinimaxM2FamilyModelId(spec.id) || isOpenAIGptOssModelId(spec.id))
 		? LOW_MEDIUM_HIGH_REASONING_EFFORTS
 		: undefined;
 }
 
-function isZaiGlm52ReasoningEffortModel<TApi extends Api>(spec: ModelSpec<TApi>): boolean {
-	if (!isGlm52ReasoningEffortModelId(spec.id)) return false;
-	return modelMatchesHost(spec, "zai") || modelMatchesHost(spec, "zhipu");
-}
-
 function isOllamaCloudGlm52ReasoningEffortModel<TApi extends Api>(spec: ModelSpec<TApi>): boolean {
 	return spec.api === "ollama-chat" && spec.provider === "ollama-cloud" && isGlm52ReasoningEffortModelId(spec.id);
 }
@@ -314,8 +317,17 @@ function readCompatEffortMap(compat: CompatOf<Api>): EffortMap | undefined {
 	return map && Object.keys(map).length > 0 ? map : undefined;
 }
 
+function isOpenRouterThinkingFormat(compat: CompatOf<Api>): boolean {
+	return compat !== undefined && "thinkingFormat" in compat && compat.thinkingFormat === "openrouter";
+}
+
+function isZaiThinkingFormat(compat: CompatOf<Api>): boolean {
+	return compat !== undefined && "thinkingFormat" in compat && compat.thinkingFormat === "zai";
+}
+
 function inferDetectedEffortMap<TApi extends Api>(
 	spec: ModelSpec<TApi>,
+	compat: CompatOf<TApi>,
 	parsedModel: ParsedModel,
 	mode: ThinkingConfig["mode"],
 ): EffortMap | undefined {
@@ -327,29 +339,42 @@ function inferDetectedEffortMap<TApi extends Api>(
 			? ANTHROPIC_ADAPTIVE_EFFORT_MAP_5_TIER
 			: ANTHROPIC_ADAPTIVE_EFFORT_MAP_4_TIER;
 	}
+	// GLM-5.2 coding SKUs accept `reasoning_effort`, but the effort dialect is
+	// host-specific (verified against live endpoints):
+	//   - Z.ai/Zhipu ("zai" dialect): the model exposes only none/high/max, so
+	//     `xhigh` 400s — collapse minimal->none, low/medium/high->high, xhigh->max.
+	//   - OpenRouter: `max` 400s and `xhigh` IS its max tier, so it passes `xhigh`
+	//     through literally (no map; the tier is exposed via getModelDefinedEfforts).
+	//   - Other openai-compat hosts (Fireworks, resellers) and Ollama Cloud keep
+	//     their distinct lower tiers and host quirks (e.g. Fireworks rejects
+	//     `minimal`, so `minimal->none` stays) and only remap the top `xhigh` UI
+	//     tier onto the genuine `max` budget. Filtered to supported efforts later.
+	const isGlm52 = isGlm52ReasoningEffortModelId(spec.id);
+	if (isGlm52 && isZaiThinkingFormat(compat)) {
+		return ZAI_GLM_52_REASONING_EFFORT_MAP;
+	}
 	if (isOllamaCloudGlm52ReasoningEffortModel(spec)) {
-		return OLLAMA_CLOUD_GLM_52_REASONING_EFFORT_MAP;
+		return GLM_52_XHIGH_MAX_EFFORT_MAP;
 	}
 	if (!isOpenAICompatReasoningApi(spec.api)) {
 		return undefined;
 	}
+	let map: EffortMap | undefined;
 	if (spec.provider === "groq" && spec.id === "qwen/qwen3-32b") {
-		return GROQ_QWEN3_32B_REASONING_EFFORT_MAP;
-	}
-	if (isZaiGlm52ReasoningEffortModel(spec)) {
-		return ZAI_GLM_52_REASONING_EFFORT_MAP;
-	}
-	if (isDeepseekReasoningModel(spec)) {
-		return DEEPSEEK_REASONING_EFFORT_MAP;
-	}
-	if (modelMatchesHost(spec, "openrouter")) {
-		const openRouterAnthropicMap = getOpenRouterAnthropicReasoningEffortMap(spec.id);
-		if (openRouterAnthropicMap !== undefined) return openRouterAnthropicMap;
+		map = GROQ_QWEN3_32B_REASONING_EFFORT_MAP;
+	} else if (isDeepseekReasoningModel(spec)) {
+		map = DEEPSEEK_REASONING_EFFORT_MAP;
+	} else if (modelMatchesHost(spec, "openrouter")) {
+		map = getOpenRouterAnthropicReasoningEffortMap(spec.id);
+	} else if (modelMatchesHost(spec, "fireworks")) {
+		map = FIREWORKS_REASONING_EFFORT_MAP;
 	}
-	if (modelMatchesHost(spec, "fireworks")) {
-		return FIREWORKS_REASONING_EFFORT_MAP;
+	// Overlay GLM-5.2's top-tier `xhigh -> max` on the host base map, except on
+	// OpenRouter (xhigh IS its max tier; `max` 400s there).
+	if (isGlm52 && !isOpenRouterThinkingFormat(compat)) {
+		map = { ...map, ...GLM_52_XHIGH_MAX_EFFORT_MAP };
 	}
-	return undefined;
+	return map;
 }
 
 function isDeepseekReasoningModel<TApi extends Api>(spec: ModelSpec<TApi>): boolean {
@@ -383,7 +408,7 @@ function inferSupportedEfforts<TApi extends Api>(
 	spec: ModelSpec<TApi>,
 	compat: CompatOf<TApi>,
 ): readonly Effort[] {
-	const modelDefinedEfforts = getModelDefinedEfforts(spec);
+	const modelDefinedEfforts = getModelDefinedEfforts(spec, compat);
 	if (modelDefinedEfforts !== undefined) {
 		return modelDefinedEfforts;
 	}