@oh-my-pi/pi-catalog 16.3.10 → 16.3.12

package/CHANGELOG.md

@@ -2,6 +2,27 @@
 
 ## [Unreleased]
 
+## [16.3.12] - 2026-07-08
+
+### Fixed
+
+- Fixed LiteLLM discovery stopping at `/model_group/info` when that endpoint omitted `supports_vision`; it now continues to `/model/info` and preserves `model_info.supports_vision=true` for vision-capable proxy models. ([#4747](https://github.com/can1357/oh-my-pi/issues/4747))
+- Fixed LiteLLM discovery to fall back to bundled catalog metadata when `models.dev` lacks a model reference, preserving reasoning and thinking support for models such as `glm-5.2`. ([#4695](https://github.com/can1357/oh-my-pi/issues/4695))
+- Detected Azure AI Inference / Foundry Anthropic routes as strict-tool-incompatible so resolved Anthropic compat disables strict tools before request construction ([#4679](https://github.com/can1357/oh-my-pi/issues/4679)).
+
+## [16.3.11] - 2026-07-06
+
+### Added
+
+- Added Claude Haiku 4.5 (JP) model support
+- Added tencent/hy3 model support via ZenMux
+
+### Changed
+
+- Updated naming format for various synthetic models to include provider prefix
+- Adjusted context window limit for MiniMax-M3 model
+- Updated pricing for select models
+
 ## [16.3.10] - 2026-07-06
 
 ### Fixed

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "16.3.10",
+	"version": "16.3.12",
 	"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": "16.3.10",
+		"@oh-my-pi/pi-utils": "16.3.12",
 		"arktype": "^2.2.0",
 		"zod": "^4"
 	},
 	"devDependencies": {
-		"@oh-my-pi/pi-ai": "16.3.10",
+		"@oh-my-pi/pi-ai": "16.3.12",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/compat/anthropic.ts

@@ -99,18 +99,15 @@ export function buildAnthropicCompat(spec: ModelSpec<"anthropic-messages">): Res
 	// (issue #4192).
 	const isZenmux = modelMatchesHost(spec, "zenmux");
 	const requiresThinkingEnabled = modelMatchesHost(spec, "moonshotNative") && matchesKimiK27CodeFamily(spec);
+	const isVertex = isVertexAnthropicRoute(baseUrl);
+	const isBedrock = isBedrockAnthropicRoute(baseUrl);
+	const isAzure = isAzureAnthropicRoute(baseUrl);
 	const signingEndpoint =
-		official ||
-		isCopilot ||
-		isZenmux ||
-		isCloudflareAnthropicGateway(baseUrl) ||
-		isVertexAnthropicRoute(baseUrl) ||
-		isBedrockAnthropicRoute(baseUrl) ||
-		isAzureAnthropicRoute(baseUrl);
+		official || isCopilot || isZenmux || isCloudflareAnthropicGateway(baseUrl) || isVertex || isBedrock || isAzure;
 	const compat: ResolvedAnthropicCompat = {
 		officialEndpoint: official,
 		signingEndpoint,
-		disableStrictTools: false,
+		disableStrictTools: isAzure,
 		disableAdaptiveThinking: false,
 		supportsEagerToolInputStreaming: !isCopilot,
 		// Long cache retention is only sent to the official API by default;

package/src/models.json

@@ -9073,6 +9073,35 @@
 			"contextWindow": 128000,
 			"maxTokens": 4096
 		},
+		"jp.anthropic.claude-haiku-4-5-20251001-v1:0": {
+			"id": "jp.anthropic.claude-haiku-4-5-20251001-v1:0",
+			"name": "Claude Haiku 4.5 (JP)",
+			"api": "bedrock-converse-stream",
+			"provider": "amazon-bedrock",
+			"baseUrl": "https://bedrock-runtime.us-east-1.amazonaws.com",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 1,
+				"output": 5,
+				"cacheRead": 0.1,
+				"cacheWrite": 1.25
+			},
+			"contextWindow": 200000,
+			"maxTokens": 64000,
+			"thinking": {
+				"mode": "budget",
+				"efforts": [
+					"minimal",
+					"low",
+					"medium",
+					"high"
+				]
+			}
+		},
 		"jp.anthropic.claude-opus-4-7": {
 			"id": "jp.anthropic.claude-opus-4-7",
 			"name": "Claude Opus 4.7 (JP)",
@@ -51831,6 +51860,25 @@
 			"contextWindow": null,
 			"maxTokens": null
 		},
+		"tencent/hy3": {
+			"id": "tencent/hy3",
+			"name": "tencent/hy3",
+			"api": "openai-completions",
+			"provider": "nanogpt",
+			"baseUrl": "https://nano-gpt.com/api/v1",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 262144,
+			"maxTokens": null
+		},
 		"tencent/hy3-preview": {
 			"id": "tencent/hy3-preview",
 			"name": "Hy3 Preview",
@@ -71923,9 +71971,9 @@
 				"text"
 			],
 			"cost": {
-				"input": 0.56,
-				"output": 1.76,
-				"cacheRead": 0.10400000000000001,
+				"input": 0.9086,
+				"output": 2.8556,
+				"cacheRead": 0.16874,
 				"cacheWrite": 0
 			},
 			"contextWindow": 1048576,
@@ -72145,7 +72193,7 @@
 	"synthetic": {
 		"hf:MiniMaxAI/MiniMax-M3": {
 			"id": "hf:MiniMaxAI/MiniMax-M3",
-			"name": "MiniMax-M3",
+			"name": "MiniMaxAI/MiniMax-M3",
 			"api": "openai-completions",
 			"provider": "synthetic",
 			"baseUrl": "https://api.synthetic.new/openai/v1",
@@ -72160,7 +72208,7 @@
 				"cacheRead": 0.6,
 				"cacheWrite": 0
 			},
-			"contextWindow": 524288,
+			"contextWindow": 262144,
 			"maxTokens": 65536,
 			"thinking": {
 				"mode": "effort",
@@ -72175,7 +72223,7 @@
 		},
 		"hf:moonshotai/Kimi-K2.7-Code": {
 			"id": "hf:moonshotai/Kimi-K2.7-Code",
-			"name": "Kimi K2.7 Code",
+			"name": "moonshotai/Kimi-K2.7-Code",
 			"api": "openai-completions",
 			"provider": "synthetic",
 			"baseUrl": "https://api.synthetic.new/openai/v1",
@@ -72205,7 +72253,7 @@
 		},
 		"hf:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4": {
 			"id": "hf:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4",
-			"name": "Nemotron 3 Super 120B A12B",
+			"name": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4",
 			"api": "openai-completions",
 			"provider": "synthetic",
 			"baseUrl": "https://api.synthetic.new/openai/v1",
@@ -72234,7 +72282,7 @@
 		},
 		"hf:openai/gpt-oss-120b": {
 			"id": "hf:openai/gpt-oss-120b",
-			"name": "GPT OSS 120B",
+			"name": "openai/gpt-oss-120b",
 			"api": "openai-completions",
 			"provider": "synthetic",
 			"baseUrl": "https://api.synthetic.new/openai/v1",
@@ -72261,7 +72309,7 @@
 		},
 		"hf:Qwen/Qwen3.6-27B": {
 			"id": "hf:Qwen/Qwen3.6-27B",
-			"name": "Qwen3.6 27B",
+			"name": "Qwen/Qwen3.6-27B",
 			"api": "openai-completions",
 			"provider": "synthetic",
 			"baseUrl": "https://api.synthetic.new/openai/v1",
@@ -72290,7 +72338,7 @@
 		},
 		"hf:zai-org/GLM-4.7-Flash": {
 			"id": "hf:zai-org/GLM-4.7-Flash",
-			"name": "GLM-4.7-Flash",
+			"name": "zai-org/GLM-4.7-Flash",
 			"api": "openai-completions",
 			"provider": "synthetic",
 			"baseUrl": "https://api.synthetic.new/openai/v1",
@@ -72319,7 +72367,7 @@
 		},
 		"hf:zai-org/GLM-5.2": {
 			"id": "hf:zai-org/GLM-5.2",
-			"name": "GLM-5.2",
+			"name": "zai-org/GLM-5.2",
 			"api": "openai-completions",
 			"provider": "synthetic",
 			"baseUrl": "https://api.synthetic.new/openai/v1",
@@ -83219,14 +83267,6 @@
 			},
 			"contextWindow": 2000000,
 			"maxTokens": 2000000,
-			"compat": {
-				"reasoningEffortMap": {
-					"minimal": "low"
-				},
-				"includeEncryptedReasoning": false,
-				"filterReasoningHistory": true,
-				"omitReasoningEffort": false
-			},
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -83239,6 +83279,14 @@
 				"effortMap": {
 					"minimal": "low"
 				}
+			},
+			"compat": {
+				"reasoningEffortMap": {
+					"minimal": "low"
+				},
+				"includeEncryptedReasoning": false,
+				"filterReasoningHistory": true,
+				"omitReasoningEffort": false
 			}
 		},
 		"grok-4.3": {
@@ -83260,14 +83308,6 @@
 			},
 			"contextWindow": 1000000,
 			"maxTokens": 1000000,
-			"compat": {
-				"reasoningEffortMap": {
-					"minimal": "low"
-				},
-				"includeEncryptedReasoning": false,
-				"filterReasoningHistory": true,
-				"omitReasoningEffort": false
-			},
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -83280,6 +83320,14 @@
 				"effortMap": {
 					"minimal": "low"
 				}
+			},
+			"compat": {
+				"reasoningEffortMap": {
+					"minimal": "low"
+				},
+				"includeEncryptedReasoning": false,
+				"filterReasoningHistory": true,
+				"omitReasoningEffort": false
 			}
 		},
 		"grok-build": {
@@ -88423,6 +88471,35 @@
 				"requiresEffort": true
 			}
 		},
+		"tencent/hy3": {
+			"id": "tencent/hy3",
+			"name": "Hy3",
+			"api": "openai-completions",
+			"provider": "zenmux",
+			"baseUrl": "https://zenmux.ai/api/v1",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.134561595,
+				"output": 0.539161765,
+				"cacheRead": 0.033869245,
+				"cacheWrite": 0
+			},
+			"contextWindow": 262144,
+			"maxTokens": null,
+			"thinking": {
+				"mode": "effort",
+				"efforts": [
+					"minimal",
+					"low",
+					"medium",
+					"high",
+					"xhigh"
+				]
+			}
+		},
 		"tencent/hy3-preview": {
 			"id": "tencent/hy3-preview",
 			"name": "Hy3 preview",

package/src/provider-models/openai-compat.ts

@@ -3032,6 +3032,15 @@ export interface FetchLiteLLMRichModelsOptions<TApi extends Api> {
 }
 
 type LiteLLMRichModelEntry = Record<string, unknown>;
+type LiteLLMRichEndpointModel<TApi extends Api> = {
+	model: ModelSpec<TApi>;
+	supportsVision: unknown;
+	supportsReasoning: unknown;
+	hasContextWindow: boolean;
+	hasMaxTokens: boolean;
+	hasToolMetadata: boolean;
+	hasSupportedOpenAIParams: boolean;
+};
 
 const LITELLM_RICH_ENDPOINTS = ["/model_group/info", "/v2/model/info", "/model/info", "/v1/model/info"] as const;
 export const OPENAI_COMPAT_DISCOVERY_DEFAULT_CONTEXT_WINDOW = 128_000;
@@ -3264,7 +3273,7 @@ async function fetchLiteLLMRichEndpoint<TApi extends Api>(
 	managementBaseUrl: string,
 	runtimeBaseUrl: string,
 	signal?: AbortSignal,
-): Promise<ModelSpec<TApi>[] | null> {
+): Promise<{ models: LiteLLMRichEndpointModel<TApi>[]; incompleteVisionMetadata: boolean } | null> {
 	const fetchImpl = discoveryFetch(options.fetch);
 	const requestHeaders: Record<string, string> = {
 		Accept: "application/json",
@@ -3296,17 +3305,39 @@ async function fetchLiteLLMRichEndpoint<TApi extends Api>(
 	if (!entries || entries.length === 0) {
 		return null;
 	}
-	const deduped = new Map<string, ModelSpec<TApi>>();
+	const deduped = new Map<string, LiteLLMRichEndpointModel<TApi>>();
+	let incompleteVisionMetadata = false;
 	for (const entry of entries) {
 		const model = mapLiteLLMRichEntry(entry, options, runtimeBaseUrl);
 		if (model) {
-			deduped.set(model.id, model);
+			const supportsVision = getLiteLLMMetadataValue(entry, "supports_vision");
+			const supportsReasoning = getLiteLLMMetadataValue(entry, "supports_reasoning");
+			const supportsFunctionCalling = getLiteLLMMetadataValue(entry, "supports_function_calling");
+			const supportedOpenAIParams = getSupportedOpenAIParams(entry);
+			if (supportsVision !== true && supportsVision !== false) {
+				incompleteVisionMetadata = true;
+			}
+			deduped.set(model.id, {
+				model,
+				supportsVision,
+				supportsReasoning,
+				hasContextWindow: toPositiveNumber(getLiteLLMMetadataValue(entry, "max_input_tokens"), null) !== null,
+				hasMaxTokens: toPositiveNumber(getLiteLLMMetadataValue(entry, "max_output_tokens"), null) !== null,
+				hasToolMetadata:
+					supportsFunctionCalling === true ||
+					supportsFunctionCalling === false ||
+					supportedOpenAIParams !== undefined,
+				hasSupportedOpenAIParams: supportedOpenAIParams !== undefined,
+			});
 		}
 	}
 	if (deduped.size === 0) {
 		return null;
 	}
-	return Array.from(deduped.values()).sort((left, right) => left.id.localeCompare(right.id));
+	return {
+		models: Array.from(deduped.values()).sort((left, right) => left.model.id.localeCompare(right.model.id)),
+		incompleteVisionMetadata,
+	};
 }
 
 export async function fetchLiteLLMRichModels<TApi extends Api>(
@@ -3318,13 +3349,55 @@ export async function fetchLiteLLMRichModels<TApi extends Api>(
 		return null;
 	}
 	const fetchModels = async (signal?: AbortSignal): Promise<ModelSpec<TApi>[] | null> => {
+		const deduped = new Map<string, LiteLLMRichEndpointModel<TApi>>();
 		for (const endpoint of LITELLM_RICH_ENDPOINTS) {
-			const models = await fetchLiteLLMRichEndpoint(endpoint, options, managementBaseUrl, runtimeBaseUrl, signal);
-			if (models) {
-				return models;
+			const result = await fetchLiteLLMRichEndpoint(endpoint, options, managementBaseUrl, runtimeBaseUrl, signal);
+			if (!result) {
+				continue;
+			}
+			const hadPriorModels = deduped.size > 0;
+			for (const next of result.models) {
+				const existing = deduped.get(next.model.id);
+				if (!existing) {
+					if (!hadPriorModels) {
+						deduped.set(next.model.id, next);
+					}
+					continue;
+				}
+				const model: ModelSpec<TApi> = {
+					...existing.model,
+					name: next.model.name === next.model.id ? existing.model.name : next.model.name,
+					contextWindow: next.hasContextWindow ? next.model.contextWindow : existing.model.contextWindow,
+					maxTokens: next.hasMaxTokens ? next.model.maxTokens : existing.model.maxTokens,
+					input:
+						next.supportsVision === true || next.supportsVision === false
+							? next.model.input
+							: existing.model.input,
+					reasoning: typeof next.supportsReasoning === "boolean" ? next.model.reasoning : existing.model.reasoning,
+					compat: next.hasSupportedOpenAIParams ? next.model.compat : existing.model.compat,
+				};
+				if (next.hasToolMetadata) {
+					model.supportsTools = next.model.supportsTools;
+				}
+				deduped.set(next.model.id, { ...next, model });
+			}
+			let hasIncompleteVisionMetadata = false;
+			for (const entry of deduped.values()) {
+				if (entry.supportsVision !== true && entry.supportsVision !== false) {
+					hasIncompleteVisionMetadata = true;
+					break;
+				}
+			}
+			if (!hasIncompleteVisionMetadata) {
+				break;
 			}
 		}
-		return null;
+		if (deduped.size === 0) {
+			return null;
+		}
+		return Array.from(deduped.values())
+			.map(entry => entry.model)
+			.sort((left, right) => left.id.localeCompare(right.id));
 	};
 	if (options.signal !== undefined) {
 		return fetchModels(options.signal);
@@ -3339,18 +3412,20 @@ export function litellmModelManagerOptions(
 	const baseUrl = config?.baseUrl ?? Bun.env.LITELLM_BASE_URL ?? "http://localhost:4000/v1";
 	return {
 		providerId: "litellm",
-		// rich-v3 invalidates rows cached before reseller usage-suffix stripping
-		// and placeholder-only `all-team-models` filtering; bump the version whenever
-		// the mappers below change, or warm authoritative caches keep serving
-		// pre-change rows for the full TTL.
-		cacheProviderId: `litellm:rich-v3:${Bun.hash(baseUrl).toString(36)}`,
+		// rich-v4 invalidates rows cached before LiteLLM ids gained bundled
+		// reference fallback and before discovery continued past `/model_group/info`
+		// when that endpoint omitted vision metadata. Earlier versions handled
+		// reseller usage-suffix stripping and placeholder-only `all-team-models`
+		// filtering; bump the version whenever the mappers below change, or warm
+		// authoritative caches keep serving pre-change rows for the full TTL.
+		cacheProviderId: `litellm:rich-v4:${Bun.hash(baseUrl).toString(36)}`,
 		// litellm is a local-only proxy and is never bundled in models.json (that
 		// would leak the machine's localhost catalog). Prefer the proxy's richer
-		// management metadata, then fall back to /v1/models and enrich bare ids
-		// against models.dev like the gateway providers (fireworks et al.) do.
+		// management metadata, then enrich ids against models.dev with the bundled
+		// catalog as a fallback before using /v1/models.
 		fetchDynamicModels: async () => {
 			const modelsDevReferences = await loadModelsDevReferences<"openai-completions">(config?.fetch);
-			const resolveReference = (id: string) => modelsDevReferences.get(id);
+			const resolveReference = createReferenceResolver(modelsDevReferences);
 			const richModels = await fetchLiteLLMRichModels({
 				api: "openai-completions",
 				provider: "litellm",