@opencode/ai 0.0.0-dev-21255 → 0.0.0-dev-21266

package/dist/protocols/alibaba-chat.d.ts

@@ -29,7 +29,6 @@ declare const Options: Schema.Struct<{
 export type OptionsInput = typeof Options.Type;
 export declare const compatibility: {
     maxTokensField: "max_completion_tokens";
-    supportsStore: false;
     supportsStrictMode: false;
     reasoningField: string;
     zaiToolStream: false;

package/dist/protocols/alibaba-chat.js

@@ -29,7 +29,6 @@ const Options = Schema.Struct({
 });
 export const compatibility = {
     maxTokensField: "max_completion_tokens",
-    supportsStore: false,
     supportsStrictMode: false,
     reasoningField: "reasoning_content",
     zaiToolStream: false,

package/dist/protocols/openai-chat.js

@@ -549,45 +549,19 @@ const detectMaxTokensField = (provider, baseURL) => {
         return "max_tokens";
     return "max_completion_tokens";
 };
+// `store` defaults to false on OpenAI and Azure, but org-level logging can still
+// store requests that omit it, so send an explicit `store: false` there. Strict
+// OpenAI-style APIs reject the field with a 400, so omit it everywhere else.
 const detectSupportsStore = (provider, baseURL) => {
     const p = provider.toLowerCase();
     const url = (baseURL ?? "").toLowerCase();
-    const isNvidia = p === "nvidia" || url.includes("integrate.api.nvidia.com");
-    const isMoonshot = p === "moonshotai" || p === "moonshotai-cn" || url.includes("api.moonshot.");
-    const isTogether = p === "togetherai" || p === "together" || url.includes("api.together.");
-    const isZai = p === "zai" ||
-        p === "zai-coding-plan" ||
-        p === "zhipuai" ||
-        p === "zhipuai-coding-plan" ||
-        url.includes("api.z.ai") ||
-        url.includes("open.bigmodel.cn");
-    const isDeepSeek = p === "deepseek" || url.includes("deepseek.com");
-    const isCerebras = p === "cerebras" || url.includes("cerebras.ai");
-    const isXai = p === "xai" || url.includes("api.x.ai");
-    const isChutes = p === "chutes" || url.includes("chutes.ai");
-    const isCloudflareWorkersAI = p === "cloudflare-workers-ai" || url.includes("api.cloudflare.com");
-    const isCloudflareAiGateway = p === "cloudflare-ai-gateway" || url.includes("gateway.ai.cloudflare.com");
-    const isVercelAiGateway = p === "vercel-ai-gateway" || url.includes("ai-gateway.vercel.sh") || url.includes("vercel.sh");
-    const isAntLing = p === "ant-ling" || url.includes("api.ant-ling.com");
-    const isOpencode = p === "opencode" || url.includes("opencode.ai");
-    const isGemini = url.includes("generativelanguage.googleapis.com");
-    const isNonStandard = isGemini ||
-        isNvidia ||
-        isCerebras ||
-        isXai ||
-        isTogether ||
-        isChutes ||
-        isDeepSeek ||
-        isZai ||
-        isMoonshot ||
-        isOpencode ||
-        isCloudflareWorkersAI ||
-        isCloudflareAiGateway ||
-        isVercelAiGateway ||
-        isAntLing;
-    return !isNonStandard;
+    return (p === "openai" ||
+        p === "azure" ||
+        url.includes("api.openai.com") ||
+        url.includes(".openai.azure.com") ||
+        url.includes(".cognitiveservices.azure.com") ||
+        url.includes(".services.ai.azure.com"));
 };
-const detectSupportsUsageInStreaming = () => true;
 const detectSupportsStrictMode = (provider, baseURL) => {
     const p = provider.toLowerCase();
     const url = (baseURL ?? "").toLowerCase();
@@ -620,11 +594,7 @@ const lowerOptions = (request, supportsStore) => {
     // (x-session-affinity, x-grok-conv-id, ...) is unaffected.
     const cacheKey = (request.model.compatibility?.supportsPromptCacheKey ?? false) ? ProviderShared.promptCacheKey(request) : undefined;
     return {
-        ...(supportsStore && options.store !== undefined ? { store: options.store } : {}),
-        // For providers that support `store`, ensure stateless `store:false` is sent
-        // even when no explicit `providerOptions.store` was supplied, mirroring the
-        // native OpenAI Chat default. Non-standard providers omit `store` entirely.
-        ...(supportsStore && options.store === undefined ? { store: false } : {}),
+        ...(supportsStore ? { store: options.store ?? false } : {}),
         ...(cacheKey ? { prompt_cache_key: cacheKey } : {}),
         ...(options.reasoningEffort ? { reasoning_effort: options.reasoningEffort } : {}),
     };
@@ -642,7 +612,6 @@ export const fromRequest = Effect.fn("OpenAIChat.fromRequest")(function* (reques
     const detectedMaxTokensField = detectMaxTokensField(provider, baseURL);
     const maxTokensField = request.model.compatibility?.maxTokensField ?? detectedMaxTokensField;
     const supportsStore = request.model.compatibility?.supportsStore ?? detectSupportsStore(provider, baseURL);
-    const supportsUsageInStreaming = request.model.compatibility?.supportsUsageInStreaming ?? detectSupportsUsageInStreaming();
     const supportsStrictMode = request.model.compatibility?.supportsStrictMode ?? detectSupportsStrictMode(provider, baseURL);
     const zaiToolStream = request.model.compatibility?.zaiToolStream ?? detectZaiToolStream(provider, baseURL, request.model.id);
     const hasHistory = hasToolHistory(request.messages);
@@ -658,7 +627,7 @@ export const fromRequest = Effect.fn("OpenAIChat.fromRequest")(function* (reques
             : flattened.tools.map((tool) => lowerTool(tool, lowering, supportsStrictMode)),
         tool_choice: hasActiveTools && request.toolChoice ? yield* lowerToolChoice(request.toolChoice) : undefined,
         stream: true,
-        ...(supportsUsageInStreaming ? { stream_options: { include_usage: true } } : {}),
+        ...((request.model.compatibility?.streamUsage ?? true) ? { stream_options: { include_usage: true } } : {}),
         ...(zaiToolStream && hasActiveTools ? { tool_stream: true } : {}),
         ...(maxTokensField === "max_completion_tokens"
             ? { max_completion_tokens: generation?.maxTokens }

package/dist/protocols/venice-chat.d.ts

@@ -49,7 +49,6 @@ interface State {
 }
 export declare const compatibility: {
     maxTokensField: "max_completion_tokens";
-    supportsStore: false;
     supportsPromptCacheKey: true;
 };
 export declare const protocol: Protocol<{

package/dist/protocols/venice-chat.js

@@ -260,7 +260,6 @@ const onHalt = Effect.fn("VeniceChat.onHalt")(function* (state) {
 });
 export const compatibility = {
     maxTokensField: "max_completion_tokens",
-    supportsStore: false,
     supportsPromptCacheKey: true,
 };
 export const protocol = Protocol.make({

package/dist/protocols/zai-chat.d.ts

@@ -18,7 +18,6 @@ export type OptionsInput = {
 };
 export declare const compatibility: {
     maxTokensField: "max_tokens";
-    supportsStore: false;
     supportsStrictMode: false;
     reasoningField: string;
     zaiToolStream: false;

package/dist/protocols/zai-chat.js

@@ -39,7 +39,6 @@ const fromRequest = Effect.fn("ZAIChat.fromRequest")(function* (request) {
 });
 export const compatibility = {
     maxTokensField: "max_tokens",
-    supportsStore: false,
     supportsStrictMode: false,
     reasoningField: "reasoning_content",
     zaiToolStream: false,

package/dist/providers/cerebras.js

@@ -28,7 +28,6 @@ export const configure = (input = {}) => {
             compatibility: {
                 maxTokensField: "max_tokens",
                 reasoningField: "reasoning",
-                supportsStore: false,
                 supportsPromptCacheKey: true,
             },
         }),

package/dist/providers/cohere.js

@@ -28,8 +28,6 @@ export const configure = (input = {}) => {
             id: modelID,
             compatibility: {
                 maxTokensField: "max_tokens",
-                supportsStore: false,
-                supportsUsageInStreaming: true,
                 reasoningField: "reasoning_content",
                 supportsStrictMode: false,
             },

package/dist/providers/deepinfra.js

@@ -31,7 +31,6 @@ export const configure = (input = {}) => {
             compatibility: {
                 maxTokensField: "max_tokens",
                 reasoningField: "reasoning_content",
-                supportsStore: false,
                 supportsPromptCacheKey: true,
             },
         }),

package/dist/providers/deepseek.js

@@ -25,7 +25,7 @@ export const configure = (input = {}) => {
         id,
         model: (modelID) => configured.model({
             id: modelID,
-            compatibility: { maxTokensField: "max_tokens", supportsStore: false },
+            compatibility: { maxTokensField: "max_tokens" },
         }),
         configure,
     };

package/dist/providers/groq.js

@@ -63,7 +63,6 @@ export const configure = (input = {}) => {
                 maxTokensField: "max_completion_tokens",
                 reasoningField: "reasoning",
                 requireReasoning: false,
-                supportsStore: false,
                 supportsStrictMode: false,
             },
         }),

package/dist/providers/meta.js

@@ -79,7 +79,7 @@ export const configure = (input = {}) => {
     const responses = (modelID) => configuredResponses.model({ id: modelID });
     const chat = (modelID) => configuredChat.model({
         id: modelID,
-        compatibility: { maxTokensField: "max_completion_tokens", supportsStore: false, supportsPromptCacheKey: true },
+        compatibility: { maxTokensField: "max_completion_tokens", supportsPromptCacheKey: true },
     });
     const messages = (modelID) => configuredMessages.model({
         id: modelID,

package/dist/providers/minimax.js

@@ -70,7 +70,7 @@ export const configure = (input = {}) => {
     const messages = (modelID) => messagesRoute.with(defaults).model({ id: modelID });
     const chat = (modelID) => chatRoute.with(defaults).model({
         id: modelID,
-        compatibility: { supportsStore: false, supportsStrictMode: false },
+        compatibility: { supportsStrictMode: false },
     });
     const responses = (modelID) => responsesRoute.with(defaults).model({ id: modelID });
     return { id, model: messages, messages, chat, responses, configure };

package/dist/providers/moonshot.js

@@ -61,7 +61,6 @@ export const configure = (input = {}) => {
         id: modelID,
         compatibility: {
             maxTokensField: "max_tokens",
-            supportsStore: false,
             supportsStrictMode: false,
             supportsPromptCacheKey: true,
             sanitizer: "moonshot",

package/dist/providers/togetherai.js

@@ -25,7 +25,7 @@ export const configure = (input = {}) => {
         id,
         model: (modelID) => configured.model({
             id: modelID,
-            compatibility: { maxTokensField: "max_tokens", supportsStore: false, supportsStrictMode: false },
+            compatibility: { maxTokensField: "max_tokens", supportsStrictMode: false },
         }),
         configure,
     };

package/dist/schema/options.d.ts

@@ -110,7 +110,8 @@ declare const LanguageModelCompatibility_base: Schema.Class<LanguageModelCompati
     readonly requireFinishReason: Schema.optional<Schema.Boolean>;
     readonly requireAssistantAfterTool: Schema.optional<Schema.Boolean>;
     readonly supportsStore: Schema.optional<Schema.Boolean>;
-    readonly supportsUsageInStreaming: Schema.optional<Schema.Boolean>;
+    /** Request usage in the stream with `stream_options.include_usage`. Defaults to true. */
+    readonly streamUsage: Schema.optional<Schema.Boolean>;
     readonly supportsStrictMode: Schema.optional<Schema.Boolean>;
     readonly supportsPromptCacheKey: Schema.optional<Schema.Boolean>;
     /** Prompt-cache marker syntax the endpoint accepts. Overrides the route default; both send `cache_control`, and `alibaba` skips tool definitions. */

package/dist/schema/options.js

@@ -145,7 +145,8 @@ export class LanguageModelCompatibility extends Schema.Class("LLM.LanguageModelC
     requireFinishReason: Schema.optional(Schema.Boolean),
     requireAssistantAfterTool: Schema.optional(Schema.Boolean),
     supportsStore: Schema.optional(Schema.Boolean),
-    supportsUsageInStreaming: Schema.optional(Schema.Boolean),
+    /** Request usage in the stream with `stream_options.include_usage`. Defaults to true. */
+    streamUsage: Schema.optional(Schema.Boolean),
     supportsStrictMode: Schema.optional(Schema.Boolean),
     // Accepts `prompt_cache_key` in the request body. Chat omits the key unless
     // this is true; other protocols omit it only when this is false.

package/package.json

@@ -1,6 +1,6 @@
 {
   "$schema": "https://json.schemastore.org/package.json",
-  "version": "0.0.0-dev-21255",
+  "version": "0.0.0-dev-21266",
   "name": "@opencode/ai",
   "type": "module",
   "license": "MIT",
@@ -34,7 +34,7 @@
   "devDependencies": {
     "@clack/prompts": "1.0.0-alpha.1",
     "@effect/platform-node": "4.0.1",
-    "@opencode/http-recorder": "0.0.0-dev-21255",
+    "@opencode/http-recorder": "0.0.0-dev-21266",
     "@tsconfig/bun": "1.0.9",
     "@types/bun": "1.4.0",
     "@typescript/native": "npm:typescript@7.0.2",
@@ -44,7 +44,7 @@
     "@aws-sdk/credential-providers": "3.1057.0",
     "@smithy/eventstream-codec": "4.2.14",
     "@smithy/util-utf8": "4.2.2",
-    "@opencode/schema": "0.0.0-dev-21255",
+    "@opencode/schema": "0.0.0-dev-21266",
     "aws4fetch": "1.0.20",
     "effect": "4.0.1",
     "google-auth-library": "10.5.0"