@oh-my-pi/pi-catalog 15.13.1 → 15.13.3

package/CHANGELOG.md

@@ -2,6 +2,47 @@
 
 ## [Unreleased]
 
+## [15.13.3] - 2026-06-15
+
+### Added
+
+- Added Azure OpenAI as a catalog provider (`azure`, default model `gpt-5.5`, env var `AZURE_OPENAI_API_KEY`), bundling the OpenAI-family models Azure serves over the Responses API (GPT-4/4.1/4o, GPT-5 family, o-series, Codex). Like Amazon Bedrock it is catalog-only — models ship in the bundle and become selectable once the env key is set, with the deployment base URL resolved at runtime from `AZURE_OPENAI_BASE_URL`/`AZURE_OPENAI_RESOURCE_NAME`.
+- Added models.dev-backed bundled catalogs for providers that previously shipped no offline models: Hugging Face, Kilo, Moonshot, NanoGPT, Synthetic, Venice, Ollama Cloud, and the Xiaomi Token Plan regions (ams/cn/sgp). They still discover live when credentialed; the bundle is now a non-empty baseline.
+
+### Changed
+
+- Updated stale provider default models to their latest bundled versions: OpenAI-family providers (`azure`, `github-copilot`, `aimlapi`) → GPT-5.5; Gemini providers (`google`, `google-gemini-cli`, `google-vertex`) → `gemini-3.1-pro-preview`; GLM providers (`zai`, `zhipu-coding-plan`) → `glm-5.2`, `cerebras` → `zai-glm-4.7`; Kimi providers (`fireworks`, `opencode-go`, `moonshot`) → `kimi-k2.7-code`, `kimi-code` → `kimi-for-coding`, `together` → `moonshotai/Kimi-K2.7-Code`; `alibaba-coding-plan` → `qwen3.7-plus`; and Claude-Sonnet defaults (`cloudflare-ai-gateway`, `cursor`, `gitlab-duo`, `kilo`, `opencode-zen`, `vercel-ai-gateway`) → Claude Opus 4.x.
+- Restricted models.dev Azure discovery to OpenAI-family IDs (`gpt-`, `o1`, `o3`, `o4`, `codex`, `chatgpt`), excluding Foundry-hosted third parties (Claude/DeepSeek/Llama/Mistral/Phi) that Azure serves through non-Responses APIs.
+- Detected the Azure OpenAI Responses compat surface (developer role, strict tool mode, strict tool-result pairing) by provider id as well as base URL, so bundled `azure` models whose deployment host is only known at runtime still get the right wire behavior.
+- Renamed the `Qwen3-ASR-Flash` model label to `Qwen3 ASR Flash`
+
+### Fixed
+
+- Fixed tool syntax selection for Gemini-family and Gemma model IDs by routing them to dedicated `gemini` and `gemma` formats instead of generic XML
+- Fixed `zhipu-coding-plan` and `together` shipping no bundled models: their descriptors referenced non-existent models.dev keys (`zhipu-coding-plan`, `together`); pointed them at the real keys (`zhipuai-coding-plan`, `togetherai`) so they bundle their GLM and full catalogs respectively.
+- Folded the `azure-openai-responses` API into the OpenAI Responses thinking-inference branches so Azure reasoning models (o-series, GPT-5, Codex) resolve the discrete effort vocabulary (including `xhigh`) and effort-control mode instead of falling through to generic defaults.
+- Fixed `ollama-cloud` discovery inheriting an unsafe cross-provider `contextWindow`/`maxTokens` when `/api/show` returns no size metadata; it now falls back to the safe 128K context / 8K output caps.
+- Dropped internal Fireworks control-plane resource ids (`accounts/fireworks/{models,routers}/…`) from the bundle; only the public request ids ship.
+
+## [15.13.2] - 2026-06-15
+
+### Added
+
+- Added the `ToolCallSyntax` union and `FALLBACK_TOOL_SYNTAX` constant to `@oh-my-pi/pi-catalog/identity` (re-exported from `@oh-my-pi/pi-ai/grammar`).
+- Added `preferredToolSyntax(modelId)` to `@oh-my-pi/pi-catalog/identity`, resolving a model's native tool-call syntax affinity from its family token (Claude→`anthropic`, GLM→`glm`, Kimi→`kimi`, Qwen→`qwen3`, DeepSeek→`deepseek`, OpenAI/gpt-oss→`harmony`, else the `xml` fallback).
+- Added `flux-1-schnell-fp8` to the Fireworks serverless model catalog
+- Added `gpt-oss-20b` to the Fireworks model catalog
+- Added `qwen3-embedding-8b` to the Fireworks model catalog
+- Added `qwen3-reranker-8b` to the Fireworks model catalog
+- Added `Gemma 4 E2B IT` and `Gemma 4 E4B IT` to the Google model catalog
+- Added `qwen/qwen3-asr-flash` to the Zenmux model catalog
+- Added sparse `supportsTools` model metadata so providers can mark models that require in-band tool-call formatting.
+
+### Changed
+
+- Kept non-tool-capable Fireworks serverless models in discovery results and marked them with `supportsTools: false` for fallback-aware handling
+- Extended `modelFamilyToken(modelId)` to classify Claude/OpenAI ids the structured parser misses (older dated forms such as `claude-3-5-sonnet-20241022` and `gpt-4o`), returning `anthropic`/`openai` instead of an empty token.
+
 ## [15.13.1] - 2026-06-15
 
 ### Added

package/dist/types/compat/openai.d.ts

@@ -14,10 +14,12 @@ interface OpenAIResponsesSpecLike {
  * Build the resolved Responses-API compat record. The Responses flavor
  * deliberately differs from chat-completions: GitHub Copilot's responses
  * endpoint accepts the `developer` role, while strict tool mode is scoped to
- * first-party OpenAI/Azure/Copilot providers. Developer-role and prompt-cache
- * detection are URL-only on purpose — the historical call sites never
- * consulted the provider id for them. The GPT-5 juice-zero hack keys on the
- * model name, matching the historical request-time check.
+ * first-party OpenAI/Azure/Copilot providers. Azure is detected by provider id
+ * as well as URL — bundled `azure` models carry no baseUrl (the deployment host
+ * is per-resource, resolved at runtime) — while OpenAI/Copilot developer-role
+ * and prompt-cache detection stay URL-keyed, as the historical call sites were.
+ * The GPT-5 juice-zero hack keys on the model name, matching the historical
+ * request-time check.
  */
 export declare function buildOpenAIResponsesCompat(spec: OpenAIResponsesSpecLike): ResolvedOpenAIResponsesCompat;
 export {};

package/dist/types/identity/family.d.ts

@@ -16,6 +16,8 @@ export declare function isClaudeModelId(modelId: string): boolean;
 export declare function isAnthropicNamespacedModelId(modelId: string): boolean;
 /** Qwen family ids (substring match — Qwen SKUs have no stable prefix shape). */
 export declare function isQwenModelId(modelId: string): boolean;
+/** Gemma open-weights family (`gemma-3-27b-it`, `google/gemma-4-E2B-it`, `gemma2-9b`). */
+export declare function isGemmaModelId(modelId: string): boolean;
 /** DeepSeek family by id or display name (proxies often rename the id but keep the name). */
 export declare function isDeepseekModelIdOrName(value: string): boolean;
 /** Xiaomi MiMo family by id or display name. */
@@ -37,6 +39,8 @@ export declare function isMinimaxM2FamilyModelId(modelId: string): boolean;
  * and `none`.
  */
 export declare function isOpenAIGptOssModelId(modelId: string): boolean;
+/** OpenAI model ids (gpt-*, o1-*, o3-*, o4-*, or prefixed with openai/). */
+export declare function isOpenAIModelId(modelId: string): boolean;
 /**
  * Reasoning-capable GLM coding SKUs: glm-4.5 and up on the base / `-air` /
  * `-turbo` lines. Excludes the vision (`…v`) shape, the non-reasoning

package/dist/types/identity/index.d.ts

@@ -7,3 +7,4 @@ export * from "./markers";
 export * from "./priority";
 export * from "./reference";
 export * from "./selection";
+export * from "./tool-syntax";

package/dist/types/identity/tool-syntax.d.ts

@@ -0,0 +1,3 @@
+export type ToolCallSyntax = "glm" | "hermes" | "kimi" | "xml" | "anthropic" | "deepseek" | "harmony" | "pi" | "qwen3" | "gemini" | "gemma";
+export declare const FALLBACK_TOOL_SYNTAX: ToolCallSyntax;
+export declare function preferredToolSyntax(modelId: string): ToolCallSyntax;

package/dist/types/provider-models/descriptors.d.ts

@@ -8,7 +8,7 @@
 import type { ModelManagerConfig, ProviderCatalogEntry, ProviderDescriptor } from "./descriptor-types";
 export declare const CATALOG_PROVIDERS: readonly [{
     readonly id: "aimlapi";
-    readonly defaultModel: "gpt-4o";
+    readonly defaultModel: "gpt-5.5-2026-04-23";
     readonly envVars: readonly ["AIMLAPI_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"openai-completions", unknown>;
     readonly dynamicModelsAuthoritative: true;
@@ -17,7 +17,7 @@ export declare const CATALOG_PROVIDERS: readonly [{
     };
 }, {
     readonly id: "alibaba-coding-plan";
-    readonly defaultModel: "qwen3.5-plus";
+    readonly defaultModel: "qwen3.7-plus";
     readonly envVars: readonly ["ALIBABA_CODING_PLAN_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"openai-completions", unknown>;
     readonly catalogDiscovery: {
@@ -30,9 +30,13 @@ export declare const CATALOG_PROVIDERS: readonly [{
     readonly id: "anthropic";
     readonly defaultModel: "claude-opus-4-8";
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"anthropic-messages", unknown>;
+}, {
+    readonly id: "azure";
+    readonly defaultModel: "gpt-5.5";
+    readonly envVars: readonly ["AZURE_OPENAI_API_KEY"];
 }, {
     readonly id: "cerebras";
-    readonly defaultModel: "zai-glm-4.6";
+    readonly defaultModel: "zai-glm-4.7";
     readonly envVars: readonly ["CEREBRAS_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"openai-completions", unknown>;
     readonly catalogDiscovery: {
@@ -40,7 +44,7 @@ export declare const CATALOG_PROVIDERS: readonly [{
     };
 }, {
     readonly id: "cloudflare-ai-gateway";
-    readonly defaultModel: "claude-sonnet-4-5";
+    readonly defaultModel: "anthropic/claude-opus-4-8";
     readonly envVars: readonly ["CLOUDFLARE_AI_GATEWAY_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"anthropic-messages", unknown>;
     readonly catalogDiscovery: {
@@ -48,7 +52,7 @@ export declare const CATALOG_PROVIDERS: readonly [{
     };
 }, {
     readonly id: "cursor";
-    readonly defaultModel: "claude-sonnet-4-6";
+    readonly defaultModel: "claude-4.6-opus-high";
     readonly envVars: readonly ["CURSOR_ACCESS_TOKEN"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"cursor-agent", unknown>;
     readonly catalogDiscovery: {
@@ -71,7 +75,7 @@ export declare const CATALOG_PROVIDERS: readonly [{
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"openai-completions", unknown>;
 }, {
     readonly id: "fireworks";
-    readonly defaultModel: "kimi-k2.6";
+    readonly defaultModel: "kimi-k2.7-code";
     readonly envVars: readonly ["FIREWORKS_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"openai-completions", unknown>;
     readonly catalogDiscovery: {
@@ -79,16 +83,16 @@ export declare const CATALOG_PROVIDERS: readonly [{
     };
 }, {
     readonly id: "github-copilot";
-    readonly defaultModel: "gpt-4o";
+    readonly defaultModel: "gpt-5.5";
     readonly envVars: readonly ["COPILOT_GITHUB_TOKEN"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<import("..").Api, unknown>;
 }, {
     readonly id: "gitlab-duo";
-    readonly defaultModel: "duo-chat-sonnet-4-5";
+    readonly defaultModel: "duo-chat-opus-4-6";
     readonly envVars: readonly ["GITLAB_TOKEN"];
 }, {
     readonly id: "google";
-    readonly defaultModel: "gemini-2.5-pro";
+    readonly defaultModel: "gemini-3.1-pro-preview";
     readonly envVars: readonly ["GEMINI_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"google-generative-ai", unknown>;
 }, {
@@ -97,11 +101,11 @@ export declare const CATALOG_PROVIDERS: readonly [{
     readonly specialModelManager: true;
 }, {
     readonly id: "google-gemini-cli";
-    readonly defaultModel: "gemini-2.5-pro";
+    readonly defaultModel: "gemini-3.1-pro-preview";
     readonly specialModelManager: true;
 }, {
     readonly id: "google-vertex";
-    readonly defaultModel: "gemini-3-pro-preview";
+    readonly defaultModel: "gemini-3.1-pro-preview";
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<import("..").Api, unknown>;
     readonly allowUnauthenticated: true;
 }, {
@@ -119,7 +123,7 @@ export declare const CATALOG_PROVIDERS: readonly [{
     };
 }, {
     readonly id: "kilo";
-    readonly defaultModel: "anthropic/claude-sonnet-4.5";
+    readonly defaultModel: "anthropic/claude-opus-4.8";
     readonly envVars: readonly ["KILO_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"openai-completions", unknown>;
     readonly catalogDiscovery: {
@@ -128,7 +132,7 @@ export declare const CATALOG_PROVIDERS: readonly [{
     };
 }, {
     readonly id: "kimi-code";
-    readonly defaultModel: "kimi-k2.5";
+    readonly defaultModel: "kimi-for-coding";
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"openai-completions", unknown>;
     readonly catalogDiscovery: {
         readonly label: "Kimi Code";
@@ -168,7 +172,7 @@ export declare const CATALOG_PROVIDERS: readonly [{
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"openai-completions", unknown>;
 }, {
     readonly id: "moonshot";
-    readonly defaultModel: "kimi-k2.5";
+    readonly defaultModel: "kimi-k2.7-code";
     readonly envVars: readonly ["MOONSHOT_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"openai-completions", unknown>;
     readonly catalogDiscovery: {
@@ -217,12 +221,12 @@ export declare const CATALOG_PROVIDERS: readonly [{
     readonly specialModelManager: true;
 }, {
     readonly id: "opencode-go";
-    readonly defaultModel: "kimi-k2.5";
+    readonly defaultModel: "kimi-k2.7-code";
     readonly envVars: readonly ["OPENCODE_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<import("..").Api, unknown>;
 }, {
     readonly id: "opencode-zen";
-    readonly defaultModel: "claude-sonnet-4-6";
+    readonly defaultModel: "claude-opus-4-8";
     readonly envVars: readonly ["OPENCODE_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<import("..").Api, unknown>;
 }, {
@@ -262,7 +266,7 @@ export declare const CATALOG_PROVIDERS: readonly [{
     };
 }, {
     readonly id: "together";
-    readonly defaultModel: "moonshotai/Kimi-K2.5";
+    readonly defaultModel: "moonshotai/Kimi-K2.7-Code";
     readonly envVars: readonly ["TOGETHER_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"openai-completions", unknown>;
     readonly catalogDiscovery: {
@@ -279,7 +283,7 @@ export declare const CATALOG_PROVIDERS: readonly [{
     };
 }, {
     readonly id: "vercel-ai-gateway";
-    readonly defaultModel: "anthropic/claude-sonnet-4-6";
+    readonly defaultModel: "anthropic/claude-opus-4.8";
     readonly envVars: readonly ["AI_GATEWAY_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"anthropic-messages", unknown>;
     readonly catalogDiscovery: {
@@ -353,7 +357,7 @@ export declare const CATALOG_PROVIDERS: readonly [{
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"openai-completions", unknown>;
 }, {
     readonly id: "zai";
-    readonly defaultModel: "glm-5.1";
+    readonly defaultModel: "glm-5.2";
     readonly envVars: readonly ["ZAI_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"anthropic-messages", unknown>;
     readonly catalogDiscovery: {
@@ -369,7 +373,7 @@ export declare const CATALOG_PROVIDERS: readonly [{
     };
 }, {
     readonly id: "zhipu-coding-plan";
-    readonly defaultModel: "glm-5.1";
+    readonly defaultModel: "glm-5.2";
     readonly envVars: readonly ["ZHIPU_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<"openai-completions", unknown>;
     readonly catalogDiscovery: {

package/dist/types/types.d.ts

@@ -371,6 +371,13 @@ export interface Model<TApi extends Api = Api> {
     baseUrl: string;
     reasoning: boolean;
     input: ("text" | "image")[];
+    /**
+     * Native provider tool-call support. `false` is the only unsupported signal:
+     * `true` and `undefined` both mean callers may use native tools. Catalog and
+     * discovery sources should set this sparsely when an upstream explicitly
+     * reports that native tool calling is unsupported.
+     */
+    supportsTools?: boolean;
     cost: {
         input: number;
         output: number;

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "15.13.1",
+	"version": "15.13.3",
 	"description": "Model catalog for omp: bundled model database, provider discovery descriptors, model identity, classification, and equivalence",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
@@ -34,11 +34,11 @@
 	},
 	"dependencies": {
 		"@bufbuild/protobuf": "^2.12.0",
-		"@oh-my-pi/pi-utils": "15.13.1",
+		"@oh-my-pi/pi-utils": "15.13.3",
 		"zod": "^4"
 	},
 	"devDependencies": {
-		"@oh-my-pi/pi-ai": "15.13.1",
+		"@oh-my-pi/pi-ai": "15.13.3",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/compat/openai.ts

@@ -301,29 +301,28 @@ interface OpenAIResponsesSpecLike {
  * Build the resolved Responses-API compat record. The Responses flavor
  * deliberately differs from chat-completions: GitHub Copilot's responses
  * endpoint accepts the `developer` role, while strict tool mode is scoped to
- * first-party OpenAI/Azure/Copilot providers. Developer-role and prompt-cache
- * detection are URL-only on purpose — the historical call sites never
- * consulted the provider id for them. The GPT-5 juice-zero hack keys on the
- * model name, matching the historical request-time check.
+ * first-party OpenAI/Azure/Copilot providers. Azure is detected by provider id
+ * as well as URL — bundled `azure` models carry no baseUrl (the deployment host
+ * is per-resource, resolved at runtime) — while OpenAI/Copilot developer-role
+ * and prompt-cache detection stay URL-keyed, as the historical call sites were.
+ * The GPT-5 juice-zero hack keys on the model name, matching the historical
+ * request-time check.
  */
 export function buildOpenAIResponsesCompat(spec: OpenAIResponsesSpecLike): ResolvedOpenAIResponsesCompat {
 	const baseUrl = spec.baseUrl ?? "";
+	const isAzure = modelMatchesHost({ provider: spec.provider, baseUrl }, "azureOpenAI");
 	const compat: ResolvedOpenAIResponsesCompat = {
-		supportsDeveloperRole:
-			hostMatchesUrl(baseUrl, "openai") ||
-			hostMatchesUrl(baseUrl, "azureOpenAI") ||
-			hostMatchesUrl(baseUrl, "githubCopilot"),
+		supportsDeveloperRole: isAzure || hostMatchesUrl(baseUrl, "openai") || hostMatchesUrl(baseUrl, "githubCopilot"),
 		supportsStrictMode:
 			spec.provider === "openai" ||
-			spec.provider === "azure" ||
+			isAzure ||
 			spec.provider === "github-copilot" ||
-			hostMatchesUrl(baseUrl, "openai") ||
-			hostMatchesUrl(baseUrl, "azureOpenAI"),
+			hostMatchesUrl(baseUrl, "openai"),
 		supportsReasoningEffort: true,
 		supportsLongPromptCacheRetention: hostMatchesUrl(baseUrl, "openai"),
 		// Azure OpenAI and GitHub Copilot Responses paths require tool results
 		// to strictly match prior tool calls when building Responses inputs.
-		strictResponsesPairing: hostMatchesUrl(baseUrl, "azureOpenAI") || spec.provider === "github-copilot",
+		strictResponsesPairing: isAzure || spec.provider === "github-copilot",
 		requiresJuiceZeroHack: spec.name.toLowerCase().startsWith("gpt-5"),
 		reasoningEffortMap: {},
 	};

package/src/identity/family.ts

@@ -41,6 +41,11 @@ export function isQwenModelId(modelId: string): boolean {
 	return modelId.toLowerCase().includes("qwen");
 }
 
+/** Gemma open-weights family (`gemma-3-27b-it`, `google/gemma-4-E2B-it`, `gemma2-9b`). */
+export function isGemmaModelId(modelId: string): boolean {
+	return /(^|\/)gemma[-.]?\d/i.test(modelId);
+}
+
 /** DeepSeek family by id or display name (proxies often rename the id but keep the name). */
 export function isDeepseekModelIdOrName(value: string): boolean {
 	return value.toLowerCase().includes("deepseek");
@@ -78,6 +83,11 @@ export function isOpenAIGptOssModelId(modelId: string): boolean {
 	return /(^|\/)gpt-oss[-:]/i.test(modelId);
 }
 
+/** OpenAI model ids (gpt-*, o1-*, o3-*, o4-*, or prefixed with openai/). */
+export function isOpenAIModelId(modelId: string): boolean {
+	return /(^|\/)(gpt|o1|o3|o4)[-.]/i.test(modelId) || modelId.toLowerCase().includes("openai/");
+}
+
 /**
  * Reasoning-capable GLM coding SKUs: glm-4.5 and up on the base / `-air` /
  * `-turbo` lines. Excludes the vision (`…v`) shape, the non-reasoning
@@ -114,12 +124,15 @@ export function isGlmVisionModelId(modelId: string): boolean {
 export function modelFamilyToken(modelId: string): string {
 	const parsed = parseKnownModel(modelId);
 	if (parsed.family !== "unknown") return parsed.family;
+	if (isClaudeModelId(modelId) || isAnthropicNamespacedModelId(modelId)) return "anthropic";
+	if (isOpenAIModelId(modelId)) return "openai";
 	if (isKimiModelId(modelId)) return "kimi";
 	if (isQwenModelId(modelId)) return "qwen";
 	if (isMinimaxM2FamilyModelId(modelId)) return "minimax";
 	if (isOpenAIGptOssModelId(modelId)) return "gpt-oss";
 	if (isDeepseekModelIdOrName(modelId)) return "deepseek";
 	if (isMimoModelIdOrName(modelId)) return "mimo";
+	if (isGemmaModelId(modelId)) return "gemma";
 	if (parseGlmModel(bareModelId(modelId))) return "glm";
 	return "";
 }

package/src/identity/index.ts

@@ -7,3 +7,4 @@ export * from "./markers";
 export * from "./priority";
 export * from "./reference";
 export * from "./selection";
+export * from "./tool-syntax";

package/src/identity/tool-syntax.ts

@@ -0,0 +1,40 @@
+import { modelFamilyToken } from "./family";
+
+export type ToolCallSyntax =
+	| "glm"
+	| "hermes"
+	| "kimi"
+	| "xml"
+	| "anthropic"
+	| "deepseek"
+	| "harmony"
+	| "pi"
+	| "qwen3"
+	| "gemini"
+	| "gemma";
+
+export const FALLBACK_TOOL_SYNTAX: ToolCallSyntax = "xml";
+
+export function preferredToolSyntax(modelId: string): ToolCallSyntax {
+	switch (modelFamilyToken(modelId)) {
+		case "anthropic":
+			return "anthropic";
+		case "glm":
+			return "glm";
+		case "gemini":
+			return "gemini";
+		case "gemma":
+			return "gemma";
+		case "kimi":
+			return "kimi";
+		case "qwen":
+			return "qwen3";
+		case "deepseek":
+			return "deepseek";
+		case "openai":
+		case "gpt-oss":
+			return "harmony";
+		default:
+			return FALLBACK_TOOL_SYNTAX;
+	}
+}

package/src/model-thinking.ts

@@ -219,7 +219,7 @@ export function deriveThinking<TApi extends Api>(spec: ModelSpec<TApi>, compat:
  * through other request fields.
  */
 function omitsWireReasoningEffort(api: Api, compat: CompatOf<Api>): boolean {
-	if (api !== "openai-responses" && api !== "openai-codex-responses") {
+	if (api !== "openai-responses" && api !== "openai-codex-responses" && api !== "azure-openai-responses") {
 		return false;
 	}
 	return (compat as ResolvedOpenAIResponsesCompat | undefined)?.supportsReasoningEffort === false;
@@ -426,7 +426,11 @@ function inferFallbackEfforts<TApi extends Api>(spec: ModelSpec<TApi>, compat: C
 		return DEFAULT_REASONING_EFFORTS;
 	}
 	// OpenAI Responses APIs encode discrete effort levels, including xhigh.
-	if (spec.api === "openai-responses" || spec.api === "openai-codex-responses") {
+	if (
+		spec.api === "openai-responses" ||
+		spec.api === "openai-codex-responses" ||
+		spec.api === "azure-openai-responses"
+	) {
 		return DEFAULT_REASONING_EFFORTS_WITH_XHIGH;
 	}
 	return DEFAULT_REASONING_EFFORTS;