@opencode/ai 2.0.21 → 2.0.22

package/dist/cache-policy.js

@@ -36,6 +36,7 @@ const resolve = (policy) => {
 // prefix caching, Gemini's implicit + out-of-band CachedContent). Skip the
 // whole policy pass for these — emitting hints would be harmless but pointless.
 const RESPECTS_INLINE_HINTS = new Set([
+    "alibaba-chat",
     "alibaba-messages",
     "anthropic-messages",
     "anthropic-compatible-messages",
@@ -47,17 +48,19 @@ const RESPECTS_INLINE_HINTS = new Set([
     "zai-coding-messages",
     "bedrock-converse",
     "openrouter",
+    "digitalocean",
 ]);
 // OpenRouter upstreams other than Anthropic and Alibaba Qwen cache without breakpoints. Gemini uses only the last
 // breakpoint, so a conversation-tail breakpoint writes a new cache every step and costs more than none. Qwen ignores
 // breakpoints on tool definitions and caches tools with the system prompt.
+const QWEN = { system: true, messages: { tail: 1 } };
 const openRouterPolicy = (modelID) => {
     // `~anthropic/claude-sonnet-latest` style IDs are OpenRouter aliases for the latest model in a family.
     const id = modelID.replace(/^~/, "");
     if (id.startsWith("anthropic/"))
         return AUTO;
     if (id.startsWith("qwen/"))
-        return { system: true, messages: { tail: 1 } };
+        return QWEN;
     return NONE;
 };
 const makeHint = (ttlSeconds) => ttlSeconds !== undefined ? new CacheHint({ type: "ephemeral", ttlSeconds }) : new CacheHint({ type: "ephemeral" });
@@ -142,7 +145,11 @@ export const applyCachePolicy = (request) => {
         return request;
     const policy = request.model.route.id === "openrouter" && (request.cache === undefined || request.cache === "auto")
         ? openRouterPolicy(request.model.id)
-        : resolve(request.cache);
+        : request.model.route.id === "alibaba-chat" && (request.cache === undefined || request.cache === "auto")
+            ? request.model.id.toLowerCase().startsWith("qwen")
+                ? QWEN
+                : NONE
+            : resolve(request.cache);
     if (!policy.tools && !policy.system && !policy.messages)
         return request;
     const hint = makeHint(policy.ttlSeconds);

package/dist/llm.d.ts

@@ -206,6 +206,8 @@ export declare class GenerateObjectResponse<T> {
         readonly reason: {
             readonly normalized: "length" | "stop" | "tool-calls" | "content-filter" | "error" | "unknown";
             readonly raw?: string | undefined;
+            readonly category?: string | undefined;
+            readonly explanation?: string | undefined;
         };
         readonly index: number;
         readonly providerMetadata?: {
@@ -219,6 +221,8 @@ export declare class GenerateObjectResponse<T> {
         readonly reason: {
             readonly normalized: "length" | "stop" | "tool-calls" | "content-filter" | "error" | "unknown";
             readonly raw?: string | undefined;
+            readonly category?: string | undefined;
+            readonly explanation?: string | undefined;
         };
         readonly providerMetadata?: {
             readonly [x: string]: {

package/dist/promise.d.ts

@@ -233,6 +233,8 @@ export declare const make: (options?: Options) => {
             readonly reason: {
                 readonly normalized: "length" | "stop" | "tool-calls" | "content-filter" | "error" | "unknown";
                 readonly raw?: string | undefined;
+                readonly category?: string | undefined;
+                readonly explanation?: string | undefined;
             };
             readonly index: number;
             readonly providerMetadata?: {
@@ -246,6 +248,8 @@ export declare const make: (options?: Options) => {
             readonly reason: {
                 readonly normalized: "length" | "stop" | "tool-calls" | "content-filter" | "error" | "unknown";
                 readonly raw?: string | undefined;
+                readonly category?: string | undefined;
+                readonly explanation?: string | undefined;
             };
             readonly providerMetadata?: {
                 readonly [x: string]: {
@@ -720,6 +724,8 @@ export declare const ai: {
             readonly reason: {
                 readonly normalized: "length" | "stop" | "tool-calls" | "content-filter" | "error" | "unknown";
                 readonly raw?: string | undefined;
+                readonly category?: string | undefined;
+                readonly explanation?: string | undefined;
             };
             readonly index: number;
             readonly providerMetadata?: {
@@ -733,6 +739,8 @@ export declare const ai: {
             readonly reason: {
                 readonly normalized: "length" | "stop" | "tool-calls" | "content-filter" | "error" | "unknown";
                 readonly raw?: string | undefined;
+                readonly category?: string | undefined;
+                readonly explanation?: string | undefined;
             };
             readonly providerMetadata?: {
                 readonly [x: string]: {

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

@@ -186,6 +186,7 @@ export declare const protocol: Protocol<{
     } | null | undefined;
     readonly usage?: {
         readonly [x: string]: unknown;
+        readonly cache_read_input_tokens?: number | null | undefined;
         readonly cached_tokens?: number | null | undefined;
         readonly prompt_tokens?: number | null | undefined;
         readonly completion_tokens?: number | null | undefined;
@@ -196,6 +197,7 @@ export declare const protocol: Protocol<{
             readonly cache_write_tokens?: number | null | undefined;
         } | null | undefined;
         readonly prompt_cache_hit_tokens?: number | null | undefined;
+        readonly cache_created_input_tokens?: number | null | undefined;
         readonly completion_tokens_details?: {
             readonly [x: string]: unknown;
             readonly reasoning_tokens?: number | null | undefined;
@@ -225,6 +227,7 @@ export declare const protocol: Protocol<{
         } | null | undefined;
         readonly usage?: {
             readonly [x: string]: unknown;
+            readonly cache_read_input_tokens?: number | null | undefined;
             readonly cached_tokens?: number | null | undefined;
             readonly prompt_tokens?: number | null | undefined;
             readonly completion_tokens?: number | null | undefined;
@@ -235,6 +238,7 @@ export declare const protocol: Protocol<{
                 readonly cache_write_tokens?: number | null | undefined;
             } | null | undefined;
             readonly prompt_cache_hit_tokens?: number | null | undefined;
+            readonly cache_created_input_tokens?: number | null | undefined;
             readonly completion_tokens_details?: {
                 readonly [x: string]: unknown;
                 readonly reasoning_tokens?: number | null | undefined;

package/dist/protocols/alibaba-chat.js

@@ -2,6 +2,7 @@ import { Effect, Schema } from "effect";
 import { Protocol } from "../route/protocol.js";
 import { OpenAIChat } from "./openai-chat.js";
 import { JsonObject, ProviderShared } from "./shared.js";
+import { cacheControl } from "./utils/cache.js";
 import { OpenResponsesOptions } from "./utils/open-responses-options.js";
 const Options = Schema.Struct({
     reasoningEffort: Schema.optional(OpenResponsesOptions.ReasoningEffort),
@@ -53,7 +54,7 @@ export const protocol = Protocol.make({
         from: Effect.fn("AlibabaChat.fromRequest")(function* (req) {
             const opts = yield* ProviderShared.validateWith(Schema.decodeUnknownEffect(Options))(req.providerOptions ?? {});
             return {
-                ...(yield* OpenAIChat.protocol.body.from(req)),
+                ...(yield* OpenAIChat.fromRequest(req, { cacheControl: cacheControl() })),
                 enable_thinking: opts.enableThinking,
                 // Alibaba also rejects an explicit budget that is not below `max_completion_tokens`.
                 thinking_budget: opts.thinkingBudget === undefined

package/dist/protocols/anthropic-messages.js

@@ -359,6 +359,7 @@ const AnthropicStreamDelta = Schema.Struct({
     signature: Schema.optional(Schema.String),
     stop_reason: optionalNull(Schema.String),
     stop_sequence: optionalNull(Schema.String),
+    stop_details: optionalNull(Schema.Struct({ category: optionalNull(Schema.String), explanation: optionalNull(Schema.String) })),
 });
 const decodeAnthropicStreamDelta = Schema.decodeUnknownOption(AnthropicStreamDelta);
 const AnthropicEvent = Schema.Struct({
@@ -677,16 +678,14 @@ const requireThinkingSignature = (request) => {
 // Mid-conversation system messages became available with Opus 4.8 and version
 // 5 of the other supported Claude families. Treat later family versions as
 // compatible without assuming that every Anthropic Messages model is Claude.
+// Opus 4.8 and every Claude 5 model accept mid-conversation system messages; later versions inherit support.
 const supportsNativeSystemUpdates = (request) => {
-    const match = /(?:^|[./])claude-(fable|haiku|mythos|opus|sonnet)-(\d+)(?:[.-](\d+))?/.exec(String(request.model.id).toLowerCase());
-    if (!match)
+    const version = claudeVersion(String(request.model.id));
+    if (version === undefined)
         return false;
-    const major = Number(match[2]);
-    if (match[1] !== "opus")
-        return major >= 5;
-    if (major !== 4)
-        return major >= 5;
-    return match[3] !== undefined && match[3].length <= 2 && Number(match[3]) >= 8;
+    if (version.family === "opus" && version.major === 4)
+        return version.minor >= 8;
+    return version.major >= 5;
 };
 const endsInServerToolUse = (message) => {
     const last = message.content.at(-1);
@@ -851,6 +850,7 @@ const lowerMessages = Effect.fn("AnthropicMessages.lowerMessages")(function* (re
     }
     return messages;
 });
+// Per-turn effort started with Claude Opus 5 and every Claude 5.1 model; later versions of any family inherit it.
 const supportsEffortUpdates = (model) => {
     const override = model.compatibility?.supportsEffortUpdates;
     if (override !== undefined)
@@ -858,10 +858,8 @@ const supportsEffortUpdates = (model) => {
     const version = claudeVersion(model.id);
     if (version === undefined)
         return false;
-    if (version.family === "opus")
-        return version.major >= 5;
-    if (version.family !== "fable" && version.family !== "mythos")
-        return false;
+    if (version.family === "opus" && version.major >= 5)
+        return true;
     return version.major > 5 || (version.major === 5 && version.minor >= 1);
 };
 const applyThinkingBindingDefault = (model, thinking) => {
@@ -1227,10 +1225,14 @@ const onMessageDelta = (state, event) => {
         const finishMetadata = stopSequence === null || stopSequence === undefined
             ? state.pendingFinish?.providerMetadata
             : providerMetadata(state.providerMetadataKey, { stopSequence });
+        const category = event.delta?.stop_details?.category;
+        const explanation = event.delta?.stop_details?.explanation;
         return {
             reason: {
                 normalized: mapFinishReason(stopReason),
                 raw: stopReason,
+                ...(category ? { category } : {}),
+                ...(explanation ? { explanation } : {}),
             },
             providerMetadata: finishMetadata,
         };

package/dist/protocols/bedrock-converse.js

@@ -236,13 +236,28 @@ const lowerToolResult = Effect.fn("BedrockConverse.lowerToolResult")(function* (
         },
     };
 });
+// Keep Claude and Nova tool-result images inline; put other models' images beside the result.
+const keepToolImagesInline = (id) => id.includes("anthropic.claude-") || id.includes("amazon.nova-");
 const lowerMessages = Effect.fn("BedrockConverse.lowerMessages")(function* (request, breakpoints) {
     const messages = [];
     const documentNames = new Set();
     // Mistral can reject replay IDs even when they satisfy Converse's broader ID syntax.
     const normalizeID = request.model.id.includes("mistral.") ? MistralToolID.normalizer(request) : (id) => id;
     const providerMetadataKey = request.model.route.providerMetadataKey ?? String(request.model.provider);
+    const hoistImages = !keepToolImagesInline(request.model.id);
+    // Bedrock expects parallel tool results before any images hoisted beside them.
+    const pendingImages = [];
+    const flushImages = () => {
+        if (pendingImages.length === 0)
+            return;
+        const previous = messages.at(-1);
+        if (previous?.role === "user")
+            messages[messages.length - 1] = { role: "user", content: [...previous.content, ...pendingImages] };
+        pendingImages.length = 0;
+    };
     for (const message of request.messages) {
+        if (message.role !== "tool")
+            flushImages();
         if (message.role === "system") {
             const part = yield* ProviderShared.wrappedSystemUpdate("Bedrock Converse", message);
             const content = textWithCache(breakpoints, part.text, part.cache);
@@ -317,7 +332,20 @@ const lowerMessages = Effect.fn("BedrockConverse.lowerMessages")(function* (requ
         for (const part of message.content) {
             if (!ProviderShared.supportsContent(part, ["tool-result"]))
                 return yield* ProviderShared.unsupportedContent("Bedrock Converse", "tool", ["tool-result"]);
-            content.push(yield* lowerToolResult(part, documentNames, normalizeID));
+            const result = yield* lowerToolResult(part, documentNames, normalizeID);
+            const images = hoistImages
+                ? result.toolResult.content.filter((item) => "image" in item)
+                : [];
+            const nonImageContent = result.toolResult.content.filter((item) => !("image" in item));
+            content.push(images.length === 0
+                ? result
+                : {
+                    toolResult: {
+                        ...result.toolResult,
+                        content: nonImageContent.length > 0 ? nonImageContent : [{ text: "See attached image." }],
+                    },
+                });
+            pendingImages.push(...images);
             const cachePoint = BedrockCache.block(breakpoints, part.cache);
             if (cachePoint)
                 content.push(cachePoint);
@@ -328,6 +356,7 @@ const lowerMessages = Effect.fn("BedrockConverse.lowerMessages")(function* (requ
         else
             messages.push({ role: "user", content });
     }
+    flushImages();
     return messages;
 });
 // System prompts share the cache-point convention: emit the text block, then

package/dist/protocols/open-responses.d.ts

@@ -1191,6 +1191,8 @@ export declare const step: (state: ParserState, event: NormalizedEvent) => AIErr
     readonly reason: {
         readonly normalized: "length" | "stop" | "tool-calls" | "content-filter" | "error" | "unknown";
         readonly raw?: string | undefined;
+        readonly category?: string | undefined;
+        readonly explanation?: string | undefined;
     };
     readonly index: number;
     readonly providerMetadata?: {
@@ -1204,6 +1206,8 @@ export declare const step: (state: ParserState, event: NormalizedEvent) => AIErr
     readonly reason: {
         readonly normalized: "length" | "stop" | "tool-calls" | "content-filter" | "error" | "unknown";
         readonly raw?: string | undefined;
+        readonly category?: string | undefined;
+        readonly explanation?: string | undefined;
     };
     readonly providerMetadata?: {
         readonly [x: string]: {

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

@@ -315,6 +315,8 @@ export declare const OpenAIChatEvent: Schema.StructWithRest<Schema.Struct<{
             readonly total_tokens: Schema.optional<Schema.NullOr<Schema.Number>>;
             readonly cached_tokens: Schema.optional<Schema.NullOr<Schema.Number>>;
             readonly prompt_cache_hit_tokens: Schema.optional<Schema.NullOr<Schema.Number>>;
+            readonly cache_read_input_tokens: Schema.optional<Schema.NullOr<Schema.Number>>;
+            readonly cache_created_input_tokens: Schema.optional<Schema.NullOr<Schema.Number>>;
             readonly prompt_tokens_details: Schema.optional<Schema.NullOr<Schema.StructWithRest<Schema.Struct<{
                 readonly cached_tokens: Schema.optional<Schema.NullOr<Schema.Number>>;
                 readonly cache_write_tokens: Schema.optional<Schema.NullOr<Schema.Number>>;
@@ -332,6 +334,8 @@ export declare const OpenAIChatEvent: Schema.StructWithRest<Schema.Struct<{
         readonly total_tokens: Schema.optional<Schema.NullOr<Schema.Number>>;
         readonly cached_tokens: Schema.optional<Schema.NullOr<Schema.Number>>;
         readonly prompt_cache_hit_tokens: Schema.optional<Schema.NullOr<Schema.Number>>;
+        readonly cache_read_input_tokens: Schema.optional<Schema.NullOr<Schema.Number>>;
+        readonly cache_created_input_tokens: Schema.optional<Schema.NullOr<Schema.Number>>;
         readonly prompt_tokens_details: Schema.optional<Schema.NullOr<Schema.StructWithRest<Schema.Struct<{
             readonly cached_tokens: Schema.optional<Schema.NullOr<Schema.Number>>;
             readonly cache_write_tokens: Schema.optional<Schema.NullOr<Schema.Number>>;
@@ -755,6 +759,7 @@ export declare const protocol: Protocol<{
     } | null | undefined;
     readonly usage?: {
         readonly [x: string]: unknown;
+        readonly cache_read_input_tokens?: number | null | undefined;
         readonly cached_tokens?: number | null | undefined;
         readonly prompt_tokens?: number | null | undefined;
         readonly completion_tokens?: number | null | undefined;
@@ -765,6 +770,7 @@ export declare const protocol: Protocol<{
             readonly cache_write_tokens?: number | null | undefined;
         } | null | undefined;
         readonly prompt_cache_hit_tokens?: number | null | undefined;
+        readonly cache_created_input_tokens?: number | null | undefined;
         readonly completion_tokens_details?: {
             readonly [x: string]: unknown;
             readonly reasoning_tokens?: number | null | undefined;
@@ -794,6 +800,7 @@ export declare const protocol: Protocol<{
         } | null | undefined;
         readonly usage?: {
             readonly [x: string]: unknown;
+            readonly cache_read_input_tokens?: number | null | undefined;
             readonly cached_tokens?: number | null | undefined;
             readonly prompt_tokens?: number | null | undefined;
             readonly completion_tokens?: number | null | undefined;
@@ -804,6 +811,7 @@ export declare const protocol: Protocol<{
                 readonly cache_write_tokens?: number | null | undefined;
             } | null | undefined;
             readonly prompt_cache_hit_tokens?: number | null | undefined;
+            readonly cache_created_input_tokens?: number | null | undefined;
             readonly completion_tokens_details?: {
                 readonly [x: string]: unknown;
                 readonly reasoning_tokens?: number | null | undefined;

package/dist/protocols/openai-chat.js

@@ -161,9 +161,11 @@ const OpenAIChatUsage = Schema.StructWithRest(Schema.Struct({
     prompt_tokens: optionalNull(Schema.Number),
     completion_tokens: optionalNull(Schema.Number),
     total_tokens: optionalNull(Schema.Number),
-    // Zai reports cache hits as top-level `cached_tokens`; DeepSeek uses `prompt_cache_hit_tokens`.
+    // Provider-specific cache accounting fields.
     cached_tokens: optionalNull(Schema.Number),
     prompt_cache_hit_tokens: optionalNull(Schema.Number),
+    cache_read_input_tokens: optionalNull(Schema.Number),
+    cache_created_input_tokens: optionalNull(Schema.Number),
     prompt_tokens_details: optionalNull(Schema.StructWithRest(Schema.Struct({
         cached_tokens: optionalNull(Schema.Number),
         cache_write_tokens: optionalNull(Schema.Number),
@@ -719,17 +721,19 @@ const mapFinishReason = Effect.fn("OpenAIChat.mapFinishReason")(function* (event
 // satisfied on both sides.
 // Providers differ on cache-hit location: OpenAI uses
 // `prompt_tokens_details.cached_tokens`, DeepSeek uses
-// `prompt_cache_hit_tokens`, and Zai uses top-level `cached_tokens`.
+// `prompt_cache_hit_tokens`, Zai uses top-level `cached_tokens`, and
+// DigitalOcean uses top-level `cache_read_input_tokens` / `cache_created_input_tokens`.
 const mapUsage = (usage, providerMetadataKey) => {
     if (!usage)
         return undefined;
     const input = usage.prompt_tokens ?? undefined;
     const output = usage.completion_tokens ?? undefined;
-    const cached = (usage.prompt_tokens_details?.cached_tokens ??
+    const cached = usage.prompt_tokens_details?.cached_tokens ??
         usage.prompt_cache_hit_tokens ??
         usage.cached_tokens ??
-        undefined);
-    const cacheWrite = usage.prompt_tokens_details?.cache_write_tokens ?? undefined;
+        usage.cache_read_input_tokens ??
+        undefined;
+    const cacheWrite = usage.prompt_tokens_details?.cache_write_tokens ?? usage.cache_created_input_tokens ?? undefined;
     const reasoning = usage.completion_tokens_details?.reasoning_tokens ?? undefined;
     const nonCached = ProviderShared.subtractTokens(input, ProviderShared.sumTokens(cached, cacheWrite));
     return new Usage({

package/dist/protocols/openai-responses.js

@@ -111,8 +111,8 @@ const adapter = {
     name: NAME,
     restoreHostedToolItem: (item) => (Schema.is(OpenAIResponsesHostedToolItem)(item) ? item : undefined),
 };
-// GPT-6 Astra, Sol, and Luna accept `configuration_update` only in standard mode (not `reasoning.mode: "pro"` or
-// `-pro` slugs), and never alongside automatic `context_management` compaction.
+// GPT-6 and later default to `configuration_update` support, except in `reasoning.mode: "pro"`
+// or alongside automatic `context_management` compaction.
 const supportsEffortUpdates = (request) => {
     if (request.providerOptions?.contextManagement !== undefined)
         return false;
@@ -121,7 +121,8 @@ const supportsEffortUpdates = (request) => {
     const override = request.model.compatibility?.supportsEffortUpdates;
     if (override !== undefined)
         return override;
-    return /(?:^|\/)gpt-6-(?:astra|sol|luna)$/i.test(request.model.id);
+    const match = /(?:^|\/)gpt-(\d+)(?:\.\d+)?(?:-|$)/i.exec(request.model.id);
+    return match !== null && Number(match[1]) >= 6;
 };
 const nativeImageToolInput = (tool) => {
     const native = tool.native?.openai;

package/dist/protocols/utils/cache.d.ts

@@ -1,6 +1,11 @@
+import type { CacheHint } from "../../schema/index.js";
 export interface Breakpoints {
     remaining: number;
     dropped: number;
 }
 export declare const newBreakpoints: (cap: number) => Breakpoints;
 export declare const ttlBucket: (ttlSeconds: number | undefined) => "1h" | undefined;
+export declare const cacheControl: () => (cache: CacheHint | undefined) => {
+    type: "ephemeral";
+    ttl: "1h" | undefined;
+} | undefined;

package/dist/protocols/utils/cache.js

@@ -1,5 +1,13 @@
-// Shared counter and TTL mapping for provider cache-marker lowering.
 export const newBreakpoints = (cap) => ({ remaining: cap, dropped: 0 });
 // Requests of at least one hour use the explicit `"1h"` bucket; shorter
 // requests omit the wire TTL and use the provider default.
 export const ttlBucket = (ttlSeconds) => ttlSeconds !== undefined && ttlSeconds >= 3600 ? "1h" : undefined;
+export const cacheControl = () => {
+    const breakpoints = newBreakpoints(4);
+    return (cache) => {
+        if (cache === undefined || breakpoints.remaining === 0)
+            return undefined;
+        breakpoints.remaining -= 1;
+        return { type: "ephemeral", ttl: ttlBucket(cache.ttlSeconds) };
+    };
+};

package/dist/protocols/utils/tool-stream.d.ts

@@ -258,6 +258,8 @@ export declare const finish: <K extends StreamKey>(route: string, tools: State<K
         readonly reason: {
             readonly normalized: "length" | "stop" | "tool-calls" | "content-filter" | "error" | "unknown";
             readonly raw?: string | undefined;
+            readonly category?: string | undefined;
+            readonly explanation?: string | undefined;
         };
         readonly index: number;
         readonly providerMetadata?: {
@@ -271,6 +273,8 @@ export declare const finish: <K extends StreamKey>(route: string, tools: State<K
         readonly reason: {
             readonly normalized: "length" | "stop" | "tool-calls" | "content-filter" | "error" | "unknown";
             readonly raw?: string | undefined;
+            readonly category?: string | undefined;
+            readonly explanation?: string | undefined;
         };
         readonly providerMetadata?: {
             readonly [x: string]: {
@@ -481,6 +485,8 @@ export declare const finishWithInput: <K extends StreamKey>(route: string, tools
         readonly reason: {
             readonly normalized: "length" | "stop" | "tool-calls" | "content-filter" | "error" | "unknown";
             readonly raw?: string | undefined;
+            readonly category?: string | undefined;
+            readonly explanation?: string | undefined;
         };
         readonly index: number;
         readonly providerMetadata?: {
@@ -494,6 +500,8 @@ export declare const finishWithInput: <K extends StreamKey>(route: string, tools
         readonly reason: {
             readonly normalized: "length" | "stop" | "tool-calls" | "content-filter" | "error" | "unknown";
             readonly raw?: string | undefined;
+            readonly category?: string | undefined;
+            readonly explanation?: string | undefined;
         };
         readonly providerMetadata?: {
             readonly [x: string]: {
@@ -701,6 +709,8 @@ export declare const finishAll: <K extends StreamKey>(route: string, tools: Stat
         readonly reason: {
             readonly normalized: "length" | "stop" | "tool-calls" | "content-filter" | "error" | "unknown";
             readonly raw?: string | undefined;
+            readonly category?: string | undefined;
+            readonly explanation?: string | undefined;
         };
         readonly index: number;
         readonly providerMetadata?: {
@@ -714,6 +724,8 @@ export declare const finishAll: <K extends StreamKey>(route: string, tools: Stat
         readonly reason: {
             readonly normalized: "length" | "stop" | "tool-calls" | "content-filter" | "error" | "unknown";
             readonly raw?: string | undefined;
+            readonly category?: string | undefined;
+            readonly explanation?: string | undefined;
         };
         readonly providerMetadata?: {
             readonly [x: string]: {

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

@@ -163,6 +163,7 @@ export declare const protocol: Protocol<{
     } | null | undefined;
     readonly usage?: {
         readonly [x: string]: unknown;
+        readonly cache_read_input_tokens?: number | null | undefined;
         readonly cached_tokens?: number | null | undefined;
         readonly prompt_tokens?: number | null | undefined;
         readonly completion_tokens?: number | null | undefined;
@@ -173,6 +174,7 @@ export declare const protocol: Protocol<{
             readonly cache_write_tokens?: number | null | undefined;
         } | null | undefined;
         readonly prompt_cache_hit_tokens?: number | null | undefined;
+        readonly cache_created_input_tokens?: number | null | undefined;
         readonly completion_tokens_details?: {
             readonly [x: string]: unknown;
             readonly reasoning_tokens?: number | null | undefined;
@@ -202,6 +204,7 @@ export declare const protocol: Protocol<{
         } | null | undefined;
         readonly usage?: {
             readonly [x: string]: unknown;
+            readonly cache_read_input_tokens?: number | null | undefined;
             readonly cached_tokens?: number | null | undefined;
             readonly prompt_tokens?: number | null | undefined;
             readonly completion_tokens?: number | null | undefined;
@@ -212,6 +215,7 @@ export declare const protocol: Protocol<{
                 readonly cache_write_tokens?: number | null | undefined;
             } | null | undefined;
             readonly prompt_cache_hit_tokens?: number | null | undefined;
+            readonly cache_created_input_tokens?: number | null | undefined;
             readonly completion_tokens_details?: {
                 readonly [x: string]: unknown;
                 readonly reasoning_tokens?: number | null | undefined;

package/dist/protocols/zai-chat.js

@@ -19,15 +19,18 @@ const Body = Schema.Struct({
     request_id: Options.fields.requestID,
     user_id: Options.fields.userID,
 });
+// Tool streaming was introduced in GLM-4.6; later versions inherit support.
+const supportsToolStreaming = (modelID) => {
+    const match = /(?:^|\/)glm-(\d+)(?:\.(\d+))?(?:-|$)/i.exec(modelID);
+    return match !== null && (Number(match[1]) > 4 || (Number(match[1]) === 4 && Number(match[2] ?? 0) >= 6));
+};
 const fromRequest = Effect.fn("ZAIChat.fromRequest")(function* (request) {
     const options = yield* ProviderShared.validateWith(Schema.decodeUnknownEffect(Options))(request.providerOptions ?? {});
     const body = yield* OpenAIChat.protocol.body.from(request);
     return {
         ...body,
         thinking: options.thinking,
-        // Tool streaming was introduced in GLM-4.6; older models must not receive the opt-in.
-        tool_stream: options.toolStream ??
-            (body.tools?.length && /^glm-(?:4\.[67]|5(?:[.-]|$))/i.test(request.model.id) ? true : undefined),
+        tool_stream: options.toolStream ?? (body.tools?.length && supportsToolStreaming(request.model.id) ? true : undefined),
         do_sample: options.doSample,
         response_format: options.responseFormat,
         request_id: options.requestID,

package/dist/provider-error.js

@@ -13,18 +13,24 @@ const patterns = [
     /tokens in request more than max tokens allowed/i,
     /maximum prompt length is \d+/i,
     /reduce the length of the messages/i,
+    // DeepInfra
+    /requested input length \d+ exceeds maximum input length/i,
     /maximum context length is \d+ tokens/i,
     /exceeds (?:the )?maximum allowed input length of [\d,]+ tokens?/i,
-    /input \(\d+ tokens\) is longer than the model'?s context length \(\d+ tokens\)/i,
+    // Novita omits the token counts.
+    /input(?: \(\d+ tokens\))? is longer than the model'?s context length/i,
     /exceeds the limit of \d+/i,
     /exceeds the available context size/i,
     /greater than the context length/i,
+    // Hugging Face Text Generation Inference, e.g. Together
+    /`inputs` tokens \+ `max_new_tokens` must be <= \d+/i,
     /context window exceeds limit/i,
     /exceeded model token limit/i,
     /context[_ ]length[_ ]exceeded/i,
     /context length is only \d+ tokens/i,
     /input length.*exceeds.*context length/i,
-    /prompt too long; exceeded (?:max )?context length/i,
+    // Z.ai code 1261 arrives as `Prompt too long` or `Prompt 超长`.
+    /prompt (?:too long|超长)/i,
     /too large for model with \d+ maximum context length/i,
     /prompt has [\d,]+ tokens?, but the configured context size is [\d,]+ tokens?/i,
     /model_context_window_exceeded/i,
@@ -92,7 +98,8 @@ const QUOTA_CODES = new Set([
     "creditlimitexceeded",
 ]);
 // Google reports an invalid API key as HTTP 400 INVALID_ARGUMENT with this `details[].reason`.
-const AUTH_CODES = new Set(["authentication_error", "permission_error", "api_key_invalid"]);
+// Z.ai's Responses API reports account and plan rejections mid-stream as `permission_denied`.
+const AUTH_CODES = new Set(["authentication_error", "permission_error", "permission_denied", "api_key_invalid"]);
 const SERVER_CODES = new Set([
     "api_error",
     "internal_error",
@@ -106,6 +113,7 @@ const SERVER_CODES = new Set([
 ]);
 // `invalid_request` is the Vercel AI Gateway's code for an upstream request rejection.
 const INVALID_REQUEST_CODES = new Set([
+    "model_not_found",
     "invalid_prompt",
     "invalid_request",
     "invalid_request_error",
@@ -125,14 +133,17 @@ const CONTENT_POLICY_CODES = new Set([
 // OpenCode Zen replaces upstream codes outside its allow-list but keeps the original
 // as a `[code]` label at the start of the rewritten message.
 const GATEWAY_CODE_LABEL = /^[^:\n]+: \[([A-Za-z0-9_.-]+)\]/;
+// xAI reports an invalid API key as HTTP 400 with the generic `invalid-argument` code.
+const AUTH_TEXT = /incorrect api key provided/i;
 const RATE_LIMIT_TEXT = /rate increased too quickly|rate[-_\s]?limit|too[_\s]?many[_\s]?requests/i;
 // Only consulted on 429, where throttles and account caps share a status.
-const QUOTA_TEXT = /insufficient[-_\s]?quota|quota[-_\s]?exceeded|budget exceeded|usage limit/i;
+// Z.ai reports balance, plan expiry, plan limits, and plan model access on 429.
+const QUOTA_TEXT = /insufficient[-_\s]?(?:quota|balance)|quota[-_\s]?exceeded|budget exceeded|usage limit|limit exhausted|package has expired|plan does not yet include/i;
 // Policy rejections without a dedicated code, matched against the provider's own
 // explanation only. OpenAI reuses `invalid_prompt` for usage-policy rejections while
 // Bedrock Mantle reuses it for schema validation; Anthropic reports blocked output
 // under `invalid_request_error`.
-const CONTENT_POLICY_TEXT = /violating our usage policy|blocked by content filtering policy|content[-_\s]?policy|rejected as a result of our safety system/i;
+const CONTENT_POLICY_TEXT = /violating our usage policy|blocked by content filtering policy|content[-_\s]?policy|rejected as a result of our safety system|detected potentially unsafe or sensitive content/i;
 const SERVER_ERROR_TEXT = /\b(?:try again|(?:please |you can )?retry (?:the |this |your )?request|try (?:the |this |your )?request again|(?:currently |temporarily )?at capacity|overloaded|temporarily unavailable|service[-_\s]?unavailable|(?:server|internal)[-_\s]?error|server (?:is )?busy|provider returned (?:an )?error|resource[-_\s]?exhausted|upstream (?:connect|connection|request)|request buffer limit while retrying upstream)\b/i;
 const Message = Schema.String.check(Schema.isPattern(/\S/));
 const messageAt = (fields, message) => Schema.Struct(fields).pipe(Schema.decodeTo(Schema.String, {
@@ -185,7 +196,10 @@ export function classifyProviderFailure(input) {
         codes.some((code) => QUOTA_CODES.has(code)) ||
         (input.status === 429 && QUOTA_TEXT.test(text)))
         return new QuotaExceededError(details);
-    if (input.status === 401 || input.status === 403 || codes.some((code) => AUTH_CODES.has(code)))
+    if (input.status === 401 ||
+        input.status === 403 ||
+        codes.some((code) => AUTH_CODES.has(code)) ||
+        (input.status === 400 && AUTH_TEXT.test(text)))
         return new AuthenticationError(details);
     if (input.status === 429 ||
         codes.some((code) => code.includes("rate_limit") || code === "too_many_requests" || code === "throttlingexception") ||