@opencode/ai 0.0.0-dev-21803 → 0.0.0-dev-21812

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

@@ -155,6 +155,12 @@ export declare const protocol: Protocol<{
     readonly store?: boolean | undefined;
     readonly prompt_cache_key?: string | undefined;
     readonly reasoning_effort?: OpenResponsesOptions.ReasoningEffort | undefined;
+    readonly verbosity?: import("../schema/options.js").OpenString<"high" | "low" | "medium"> | undefined;
+    readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority"> | undefined;
+    readonly metadata?: {
+        readonly [x: string]: string;
+    } | undefined;
+    readonly safety_identifier?: string | undefined;
     readonly tool_stream?: boolean | undefined;
     readonly max_completion_tokens?: number | undefined;
     readonly max_tokens?: number | undefined;

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

@@ -301,6 +301,7 @@ export declare const protocol: Protocol<{
         readonly user_id?: string | null | undefined;
     } | undefined;
     readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "standard_only"> | undefined;
+    readonly speed?: import("../schema/options.js").OpenString<"fast" | "standard"> | undefined;
     readonly thinking?: {
         readonly type: string;
         readonly budget_tokens?: number | undefined;

package/dist/protocols/anthropic-messages.d.ts

@@ -102,6 +102,7 @@ declare const Options: Schema.Struct<{
     readonly effort: Schema.optional<Schema.String>;
     readonly service_tier: Schema.optional<Schema.declare<import("../schema/options.js").OpenString<"auto" | "standard_only">, import("../schema/options.js").OpenString<"auto" | "standard_only">>>;
     readonly serviceTier: Schema.optional<Schema.declare<import("../schema/options.js").OpenString<"auto" | "standard_only">, import("../schema/options.js").OpenString<"auto" | "standard_only">>>;
+    readonly speed: Schema.optional<Schema.declare<import("../schema/options.js").OpenString<"fast" | "standard">, import("../schema/options.js").OpenString<"fast" | "standard">>>;
     readonly metadata: Schema.optional<Schema.Struct<{
         readonly user_id: Schema.optional<Schema.NullOr<Schema.String>>;
     }>>;
@@ -434,6 +435,7 @@ export declare const AnthropicMessagesBody: Schema.Struct<{
         readonly user_id: Schema.optional<Schema.NullOr<Schema.String>>;
     }>>;
     service_tier: Schema.optional<Schema.declare<import("../schema/options.js").OpenString<"auto" | "standard_only">, import("../schema/options.js").OpenString<"auto" | "standard_only">>>;
+    speed: Schema.optional<Schema.declare<import("../schema/options.js").OpenString<"fast" | "standard">, import("../schema/options.js").OpenString<"fast" | "standard">>>;
 }>;
 export type AnthropicMessagesBody = Schema.Schema.Type<typeof AnthropicMessagesBody>;
 /**
@@ -747,6 +749,7 @@ export declare const protocol: Protocol<{
         readonly user_id?: string | null | undefined;
     } | undefined;
     readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "standard_only"> | undefined;
+    readonly speed?: import("../schema/options.js").OpenString<"fast" | "standard"> | undefined;
 }, string, {
     readonly [x: string]: unknown;
     readonly type: string;
@@ -829,7 +832,7 @@ export declare const protocol: Protocol<{
     reasoningSignatures: {};
     lifecycle: Lifecycle.State;
 }>;
-export declare const transport: <Body extends Pick<AnthropicMessagesBody, "messages" | "context_management" | "thinking">>() => {
+export declare const transport: <Body extends Pick<AnthropicMessagesBody, "messages" | "context_management" | "thinking" | "speed">>() => {
     with: (patch: Partial<HttpTransport.HttpJsonInput<Body, string>>) => HttpTransport.HttpJsonTransport<Body, string>;
     id: string;
     execute: (prepared: HttpTransport.HttpPrepared<string>, request: LLMRequest, runtime: import("../route/transport/index.js").TransportRuntime, options?: import("../route/transport/index.js").TransportExecuteOptions) => Effect.Effect<import("../route/transport/index.js").TransportExecution<string>, AIError, import("effect/Scope").Scope>;
@@ -1141,5 +1144,6 @@ export declare const route: Route<{
         readonly user_id?: string | null | undefined;
     } | undefined;
     readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "standard_only"> | undefined;
+    readonly speed?: import("../schema/options.js").OpenString<"fast" | "standard"> | undefined;
 }, HttpTransport.HttpPrepared<string>, import("../route/client.js").CompactionOperations | undefined>;
 export * as AnthropicMessages from "./anthropic-messages.js";

package/dist/protocols/anthropic-messages.js

@@ -53,6 +53,7 @@ const AnthropicCacheControl = Schema.Struct({
     ttl: Schema.optional(Schema.Literals(["5m", "1h"])),
 });
 const AnthropicServiceTier = knownString();
+const AnthropicSpeed = knownString();
 const AnthropicTextBlock = Schema.Struct({
     type: Schema.tag("text"),
     text: Schema.String,
@@ -287,6 +288,7 @@ const Options = Schema.Struct({
     effort: Schema.optional(Schema.String),
     service_tier: Schema.optional(AnthropicServiceTier),
     serviceTier: Schema.optional(AnthropicServiceTier),
+    speed: Schema.optional(AnthropicSpeed),
     metadata: Schema.optional(AnthropicMetadata),
     container: Schema.optional(AnthropicContainer),
     inference_geo: optionalNull(Schema.String),
@@ -325,6 +327,7 @@ const AnthropicBodyFields = {
     inference_geo: Schema.optional(Schema.NullOr(Schema.String)),
     metadata: Schema.optional(AnthropicMetadata),
     service_tier: Schema.optional(AnthropicServiceTier),
+    speed: Schema.optional(AnthropicSpeed),
 };
 export const AnthropicMessagesBody = Schema.Struct(AnthropicBodyFields);
 const AnthropicIterationUsage = Schema.StructWithRest(Schema.Struct({
@@ -964,6 +967,7 @@ const fromRequest = Effect.fn("AnthropicMessages.fromRequest")(function* (reques
         inference_geo: options.inference_geo ?? options.inferenceGeo ?? undefined,
         metadata: options.metadata,
         service_tier: options.service_tier ?? options.serviceTier,
+        speed: options.speed,
     };
     if (!management)
         return body;
@@ -1457,6 +1461,9 @@ function requiredBetaHeaders(body) {
         betas.push("compact-2026-01-12");
     if (body.messages.some((message) => message.role === "system" && message.output_config !== undefined))
         betas.push("mid-conversation-output-config-2026-07-01");
+    // Opus 5.5 accepts `speed` without this header; earlier fast-mode models reject `speed` without it.
+    if (body.speed !== undefined)
+        betas.push("fast-mode-2026-02-01");
     const thinking = body.thinking;
     if (thinking && !isThinkingOff(thinking) && thinking.block_binding)
         betas.push(THINKING_BINDING_BETA);

package/dist/protocols/meta-messages.d.ts

@@ -294,6 +294,7 @@ export declare const protocol: Protocol<{
         readonly user_id?: string | null | undefined;
     } | undefined;
     readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "standard_only"> | undefined;
+    readonly speed?: import("../schema/options.js").OpenString<"fast" | "standard"> | undefined;
     readonly tools?: readonly ({
         readonly type: "web_search";
         readonly name: "web_search";

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

@@ -274,6 +274,11 @@ export declare const bodyFields: {
     store: Schema.optional<Schema.Boolean>;
     prompt_cache_key: Schema.optional<Schema.String>;
     reasoning_effort: Schema.optional<Schema.declare<OpenResponsesOptions.ReasoningEffort, OpenResponsesOptions.ReasoningEffort>>;
+    verbosity: Schema.optional<Schema.declare<import("../schema/options.js").OpenString<"high" | "low" | "medium">, import("../schema/options.js").OpenString<"high" | "low" | "medium">>>;
+    service_tier: Schema.optional<Schema.declare<import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority">, import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority">>>;
+    parallel_tool_calls: Schema.optional<Schema.Boolean>;
+    metadata: Schema.optional<Schema.$Record<Schema.String, Schema.String>>;
+    safety_identifier: Schema.optional<Schema.String>;
     tool_stream: Schema.optional<Schema.Boolean>;
     max_completion_tokens: Schema.optional<Schema.Number>;
     max_tokens: Schema.optional<Schema.Number>;
@@ -403,6 +408,11 @@ declare const OpenAIChatBody: Schema.Struct<{
     store: Schema.optional<Schema.Boolean>;
     prompt_cache_key: Schema.optional<Schema.String>;
     reasoning_effort: Schema.optional<Schema.declare<OpenResponsesOptions.ReasoningEffort, OpenResponsesOptions.ReasoningEffort>>;
+    verbosity: Schema.optional<Schema.declare<import("../schema/options.js").OpenString<"high" | "low" | "medium">, import("../schema/options.js").OpenString<"high" | "low" | "medium">>>;
+    service_tier: Schema.optional<Schema.declare<import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority">, import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority">>>;
+    parallel_tool_calls: Schema.optional<Schema.Boolean>;
+    metadata: Schema.optional<Schema.$Record<Schema.String, Schema.String>>;
+    safety_identifier: Schema.optional<Schema.String>;
     tool_stream: Schema.optional<Schema.Boolean>;
     max_completion_tokens: Schema.optional<Schema.Number>;
     max_tokens: Schema.optional<Schema.Number>;
@@ -597,6 +607,13 @@ export declare const fromRequest: (request: LLMRequest<import("../schema/options
     store?: boolean | undefined;
     prompt_cache_key?: string | undefined;
     reasoning_effort?: OpenResponsesOptions.ReasoningEffort | undefined;
+    verbosity?: import("../schema/options.js").OpenString<"high" | "low" | "medium"> | undefined;
+    service_tier?: import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority"> | undefined;
+    parallel_tool_calls?: boolean | undefined;
+    metadata?: {
+        readonly [x: string]: string;
+    } | undefined;
+    safety_identifier?: string | undefined;
     model: string & import("effect/Brand").Brand<"AI.ModelID">;
     messages: ({
         readonly role: "system";
@@ -727,6 +744,13 @@ export declare const fromRequest: (request: LLMRequest<import("../schema/options
     store?: boolean | undefined;
     prompt_cache_key?: string | undefined;
     reasoning_effort?: OpenResponsesOptions.ReasoningEffort | undefined;
+    verbosity?: import("../schema/options.js").OpenString<"high" | "low" | "medium"> | undefined;
+    service_tier?: import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority"> | undefined;
+    parallel_tool_calls?: boolean | undefined;
+    metadata?: {
+        readonly [x: string]: string;
+    } | undefined;
+    safety_identifier?: string | undefined;
     model: string & import("effect/Brand").Brand<"AI.ModelID">;
     messages: ({
         readonly role: "system";
@@ -1198,6 +1222,13 @@ export declare const protocol: Protocol<{
     readonly store?: boolean | undefined;
     readonly prompt_cache_key?: string | undefined;
     readonly reasoning_effort?: OpenResponsesOptions.ReasoningEffort | undefined;
+    readonly verbosity?: import("../schema/options.js").OpenString<"high" | "low" | "medium"> | undefined;
+    readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority"> | undefined;
+    readonly parallel_tool_calls?: boolean | undefined;
+    readonly metadata?: {
+        readonly [x: string]: string;
+    } | undefined;
+    readonly safety_identifier?: string | undefined;
     readonly tool_stream?: boolean | undefined;
     readonly max_completion_tokens?: number | undefined;
     readonly max_tokens?: number | undefined;
@@ -1403,6 +1434,13 @@ export declare const httpTransport: HttpTransport.HttpJsonTransport<{
     readonly store?: boolean | undefined;
     readonly prompt_cache_key?: string | undefined;
     readonly reasoning_effort?: OpenResponsesOptions.ReasoningEffort | undefined;
+    readonly verbosity?: import("../schema/options.js").OpenString<"high" | "low" | "medium"> | undefined;
+    readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority"> | undefined;
+    readonly parallel_tool_calls?: boolean | undefined;
+    readonly metadata?: {
+        readonly [x: string]: string;
+    } | undefined;
+    readonly safety_identifier?: string | undefined;
     readonly tool_stream?: boolean | undefined;
     readonly max_completion_tokens?: number | undefined;
     readonly max_tokens?: number | undefined;
@@ -1535,6 +1573,13 @@ export declare const route: Route<{
     readonly store?: boolean | undefined;
     readonly prompt_cache_key?: string | undefined;
     readonly reasoning_effort?: OpenResponsesOptions.ReasoningEffort | undefined;
+    readonly verbosity?: import("../schema/options.js").OpenString<"high" | "low" | "medium"> | undefined;
+    readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority"> | undefined;
+    readonly parallel_tool_calls?: boolean | undefined;
+    readonly metadata?: {
+        readonly [x: string]: string;
+    } | undefined;
+    readonly safety_identifier?: string | undefined;
     readonly tool_stream?: boolean | undefined;
     readonly max_completion_tokens?: number | undefined;
     readonly max_tokens?: number | undefined;

package/dist/protocols/openai-chat.js

@@ -10,6 +10,7 @@ import { AIError, AIErrorReason, InvalidProviderOutputError, LLMEvent, ProviderI
 import { classifyProviderFailure } from "../provider-error.js";
 import { isRecord, JsonObject, optionalArray, optionalNull, ProviderShared } from "./shared.js";
 import { cacheControl } from "./utils/cache.js";
+import { resolveParallelToolCalls } from "./open-responses.js";
 import { OpenResponsesOptions } from "./utils/open-responses-options.js";
 import { Lifecycle } from "./utils/lifecycle.js";
 import { ToolStream } from "./utils/tool-stream.js";
@@ -140,6 +141,11 @@ export const bodyFields = {
     store: Schema.optional(Schema.Boolean),
     prompt_cache_key: Schema.optional(Schema.String),
     reasoning_effort: Schema.optional(OpenResponsesOptions.ReasoningEffort),
+    verbosity: Schema.optional(OpenResponsesOptions.TextVerbositySchema),
+    service_tier: Schema.optional(OpenResponsesOptions.ServiceTierSchema),
+    parallel_tool_calls: Schema.optional(Schema.Boolean),
+    metadata: Schema.optional(Schema.Record(Schema.String, Schema.String)),
+    safety_identifier: Schema.optional(Schema.String),
     tool_stream: Schema.optional(Schema.Boolean),
     max_completion_tokens: Schema.optional(Schema.Number),
     max_tokens: Schema.optional(Schema.Number),
@@ -593,8 +599,10 @@ const detectZaiToolStream = (provider, baseURL, modelID) => {
         return false;
     return true;
 };
-const lowerOptions = (request, supportsStore) => {
+const lowerOptions = (request, supportsStore, hasActiveTools) => {
     const options = OpenResponsesOptions.resolve(request);
+    // OpenAI rejects `parallel_tool_calls` unless tools are present.
+    const parallelToolCalls = hasActiveTools ? resolveParallelToolCalls(request, options) : undefined;
     // Default off: strict providers 400 on unknown body fields, so only send
     // the key where compatibility explicitly allows it. Header-based affinity
     // (x-session-affinity, x-grok-conv-id, ...) is unaffected.
@@ -603,6 +611,11 @@ const lowerOptions = (request, supportsStore) => {
         ...(supportsStore ? { store: options.store ?? false } : {}),
         ...(cacheKey ? { prompt_cache_key: cacheKey } : {}),
         ...(options.reasoningEffort ? { reasoning_effort: options.reasoningEffort } : {}),
+        ...(options.textVerbosity ? { verbosity: options.textVerbosity } : {}),
+        ...(options.serviceTier ? { service_tier: options.serviceTier } : {}),
+        ...(parallelToolCalls !== undefined ? { parallel_tool_calls: parallelToolCalls } : {}),
+        ...(options.metadata ? { metadata: options.metadata } : {}),
+        ...(options.safetyIdentifier ? { safety_identifier: options.safetyIdentifier } : {}),
     };
 };
 export const fromRequest = Effect.fn("OpenAIChat.fromRequest")(function* (request, options = {}) {
@@ -644,7 +657,7 @@ export const fromRequest = Effect.fn("OpenAIChat.fromRequest")(function* (reques
         presence_penalty: generation?.presencePenalty,
         seed: generation?.seed,
         stop: generation?.stop,
-        ...lowerOptions(request, supportsStore),
+        ...lowerOptions(request, supportsStore, hasActiveTools),
     };
 });
 // Provider wrappers that pass their own marker writer keep it; otherwise a declared cache format adds one.
@@ -841,16 +854,9 @@ const step = (state, event) => Effect.gen(function* () {
     let nextToolIndex = state.nextToolIndex;
     let lifecycle = state.lifecycle;
     const reasoning = reasoningDelta(delta, state.reasoningField);
-    const hasLateContent = Boolean(delta?.content) ||
-        Boolean(delta?.refusal) ||
-        reasoning !== undefined ||
-        (Array.isArray(delta?.reasoning_details) && delta.reasoning_details.length > 0) ||
-        toolDeltas.some((tool) => Boolean(tool.id) || Boolean(tool.function?.name) || Boolean(tool.function?.arguments));
-    if (state.finishReason !== undefined) {
-        if (hasLateContent)
-            return yield* ProviderShared.eventError(ADAPTER, "OpenAI Chat received content after the finish reason", ProviderShared.encodeJson(event));
-        return [{ ...state, usage }, events];
-    }
+    if (state.finishReason !== undefined &&
+        toolDeltas.some((tool) => Boolean(tool.id) || Boolean(tool.function?.name) || Boolean(tool.function?.arguments)))
+        return yield* ProviderShared.eventError(ADAPTER, "OpenAI Chat received content after the finish reason", ProviderShared.encodeJson(event));
     const reasoningField = state.reasoningField ?? reasoning?.field;
     const reasoningTextObserved = state.reasoningTextObserved || reasoning !== undefined;
     const detailDelta = Array.isArray(delta?.reasoning_details)
@@ -943,7 +949,7 @@ const step = (state, event) => Effect.gen(function* () {
             pendingTools,
             toolCallEvents: finished?.events ?? state.toolCallEvents,
             usage,
-            finishReason,
+            finishReason: state.finishReason ?? finishReason,
             lifecycle,
             reasoningField,
             reasoningTextObserved,

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

@@ -128,6 +128,13 @@ export declare const route: Route<{
     readonly store?: boolean | undefined;
     readonly prompt_cache_key?: string | undefined;
     readonly reasoning_effort?: import("../effect.js").ReasoningEffort | undefined;
+    readonly verbosity?: import("../effect.js").OpenString<"high" | "low" | "medium"> | undefined;
+    readonly service_tier?: import("../effect.js").OpenString<"auto" | "default" | "flex" | "priority"> | undefined;
+    readonly parallel_tool_calls?: boolean | undefined;
+    readonly metadata?: {
+        readonly [x: string]: string;
+    } | undefined;
+    readonly safety_identifier?: string | undefined;
     readonly tool_stream?: boolean | undefined;
     readonly max_completion_tokens?: number | undefined;
     readonly max_tokens?: number | undefined;

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

@@ -11,6 +11,10 @@ export declare const ContextManagement: Schema.$Array<Schema.Struct<{
     readonly compactThreshold: Schema.optional<Schema.Int>;
 }>>;
 export type ContextManagement = typeof ContextManagement.Type;
+export declare const ReasoningMode: Schema.declare<import("../schema/options.js").OpenString<"pro" | "standard">, import("../schema/options.js").OpenString<"pro" | "standard">>;
+export type ReasoningMode = typeof ReasoningMode.Type;
+export declare const ReasoningContext: Schema.declare<import("../schema/options.js").OpenString<"all_turns" | "auto" | "current_turn">, import("../schema/options.js").OpenString<"all_turns" | "auto" | "current_turn">>;
+export type ReasoningContext = typeof ReasoningContext.Type;
 declare const OpenAIResponsesBody: Schema.Struct<{
     readonly model: Schema.String;
     readonly input: Schema.$Array<Schema.Union<readonly [Schema.Struct<{
@@ -140,10 +144,6 @@ declare const OpenAIResponsesBody: Schema.Struct<{
     readonly service_tier: Schema.optional<Schema.declare<import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority">, import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority">>>;
     readonly prompt_cache_key: Schema.optional<Schema.String>;
     readonly include: Schema.optional<Schema.$Array<Schema.declare<import("../schema/options.js").OpenString<"code_interpreter_call.outputs" | "computer_call_output.output.image_url" | "file_search_call.results" | "message.input_image.image_url" | "message.output_text.logprobs" | "reasoning.encrypted_content" | "web_search_call.action.sources" | "web_search_call.results">, import("../schema/options.js").OpenString<"code_interpreter_call.outputs" | "computer_call_output.output.image_url" | "file_search_call.results" | "message.input_image.image_url" | "message.output_text.logprobs" | "reasoning.encrypted_content" | "web_search_call.action.sources" | "web_search_call.results">>>>;
-    readonly reasoning: Schema.optional<Schema.Struct<{
-        readonly effort: Schema.optional<Schema.declare<OpenResponsesOptions.ReasoningEffort, OpenResponsesOptions.ReasoningEffort>>;
-        readonly summary: Schema.optional<Schema.Literals<readonly ["auto", "concise", "detailed"]>>;
-    }>>;
     readonly text: Schema.optional<Schema.Struct<{
         readonly verbosity: Schema.optional<Schema.declare<import("../schema/options.js").OpenString<"high" | "low" | "medium">, import("../schema/options.js").OpenString<"high" | "low" | "medium">>>;
     }>>;
@@ -195,6 +195,12 @@ declare const OpenAIResponsesBody: Schema.Struct<{
     }>]>, Schema.Struct<{
         readonly type: Schema.tag<"image_generation">;
     }>]>>;
+    readonly reasoning: Schema.optional<Schema.Struct<{
+        readonly effort: Schema.optional<Schema.declare<OpenResponsesOptions.ReasoningEffort, OpenResponsesOptions.ReasoningEffort>>;
+        readonly summary: Schema.optional<Schema.Literals<readonly ["auto", "concise", "detailed"]>>;
+        readonly mode: Schema.optional<Schema.declare<import("../schema/options.js").OpenString<"pro" | "standard">, import("../schema/options.js").OpenString<"pro" | "standard">>>;
+        readonly context: Schema.optional<Schema.declare<import("../schema/options.js").OpenString<"all_turns" | "auto" | "current_turn">, import("../schema/options.js").OpenString<"all_turns" | "auto" | "current_turn">>>;
+    }>>;
     readonly context_management: Schema.optional<Schema.$Array<Schema.Struct<{
         readonly type: Schema.Literal<"compaction">;
         readonly compact_threshold: Schema.optional<Schema.Int>;
@@ -348,10 +354,6 @@ export declare const protocol: Protocol<{
     readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority"> | undefined;
     readonly prompt_cache_key?: string | undefined;
     readonly include?: readonly import("../schema/options.js").OpenString<"code_interpreter_call.outputs" | "computer_call_output.output.image_url" | "file_search_call.results" | "message.input_image.image_url" | "message.output_text.logprobs" | "reasoning.encrypted_content" | "web_search_call.action.sources" | "web_search_call.results">[] | undefined;
-    readonly reasoning?: {
-        readonly effort?: OpenResponsesOptions.ReasoningEffort | undefined;
-        readonly summary?: "auto" | "concise" | "detailed" | undefined;
-    } | undefined;
     readonly text?: {
         readonly verbosity?: import("../schema/options.js").OpenString<"high" | "low" | "medium"> | undefined;
     } | undefined;
@@ -407,6 +409,12 @@ export declare const protocol: Protocol<{
     } | {
         readonly type: "image_generation";
     } | undefined;
+    readonly reasoning?: {
+        readonly effort?: OpenResponsesOptions.ReasoningEffort | undefined;
+        readonly summary?: "auto" | "concise" | "detailed" | undefined;
+        readonly mode?: import("../schema/options.js").OpenString<"pro" | "standard"> | undefined;
+        readonly context?: import("../schema/options.js").OpenString<"all_turns" | "auto" | "current_turn"> | undefined;
+    } | undefined;
     readonly context_management?: readonly {
         readonly type: "compaction";
         readonly compact_threshold?: number | undefined;
@@ -613,10 +621,6 @@ export declare const httpTransport: HttpTransport.HttpJsonTransport<{
     readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority"> | undefined;
     readonly prompt_cache_key?: string | undefined;
     readonly include?: readonly import("../schema/options.js").OpenString<"code_interpreter_call.outputs" | "computer_call_output.output.image_url" | "file_search_call.results" | "message.input_image.image_url" | "message.output_text.logprobs" | "reasoning.encrypted_content" | "web_search_call.action.sources" | "web_search_call.results">[] | undefined;
-    readonly reasoning?: {
-        readonly effort?: OpenResponsesOptions.ReasoningEffort | undefined;
-        readonly summary?: "auto" | "concise" | "detailed" | undefined;
-    } | undefined;
     readonly text?: {
         readonly verbosity?: import("../schema/options.js").OpenString<"high" | "low" | "medium"> | undefined;
     } | undefined;
@@ -672,6 +676,12 @@ export declare const httpTransport: HttpTransport.HttpJsonTransport<{
     } | {
         readonly type: "image_generation";
     } | undefined;
+    readonly reasoning?: {
+        readonly effort?: OpenResponsesOptions.ReasoningEffort | undefined;
+        readonly summary?: "auto" | "concise" | "detailed" | undefined;
+        readonly mode?: import("../schema/options.js").OpenString<"pro" | "standard"> | undefined;
+        readonly context?: import("../schema/options.js").OpenString<"all_turns" | "auto" | "current_turn"> | undefined;
+    } | undefined;
     readonly context_management?: readonly {
         readonly type: "compaction";
         readonly compact_threshold?: number | undefined;
@@ -820,10 +830,6 @@ export declare const channelTransport: (options: import("./open-responses-channe
     readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority"> | undefined;
     readonly prompt_cache_key?: string | undefined;
     readonly include?: readonly import("../schema/options.js").OpenString<"code_interpreter_call.outputs" | "computer_call_output.output.image_url" | "file_search_call.results" | "message.input_image.image_url" | "message.output_text.logprobs" | "reasoning.encrypted_content" | "web_search_call.action.sources" | "web_search_call.results">[] | undefined;
-    readonly reasoning?: {
-        readonly effort?: OpenResponsesOptions.ReasoningEffort | undefined;
-        readonly summary?: "auto" | "concise" | "detailed" | undefined;
-    } | undefined;
     readonly text?: {
         readonly verbosity?: import("../schema/options.js").OpenString<"high" | "low" | "medium"> | undefined;
     } | undefined;
@@ -879,6 +885,12 @@ export declare const channelTransport: (options: import("./open-responses-channe
     } | {
         readonly type: "image_generation";
     } | undefined;
+    readonly reasoning?: {
+        readonly effort?: OpenResponsesOptions.ReasoningEffort | undefined;
+        readonly summary?: "auto" | "concise" | "detailed" | undefined;
+        readonly mode?: import("../schema/options.js").OpenString<"pro" | "standard"> | undefined;
+        readonly context?: import("../schema/options.js").OpenString<"all_turns" | "auto" | "current_turn"> | undefined;
+    } | undefined;
     readonly context_management?: readonly {
         readonly type: "compaction";
         readonly compact_threshold?: number | undefined;
@@ -1027,10 +1039,6 @@ export declare const transport: import("../route/transport/index.js").Transport<
     readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority"> | undefined;
     readonly prompt_cache_key?: string | undefined;
     readonly include?: readonly import("../schema/options.js").OpenString<"code_interpreter_call.outputs" | "computer_call_output.output.image_url" | "file_search_call.results" | "message.input_image.image_url" | "message.output_text.logprobs" | "reasoning.encrypted_content" | "web_search_call.action.sources" | "web_search_call.results">[] | undefined;
-    readonly reasoning?: {
-        readonly effort?: OpenResponsesOptions.ReasoningEffort | undefined;
-        readonly summary?: "auto" | "concise" | "detailed" | undefined;
-    } | undefined;
     readonly text?: {
         readonly verbosity?: import("../schema/options.js").OpenString<"high" | "low" | "medium"> | undefined;
     } | undefined;
@@ -1086,6 +1094,12 @@ export declare const transport: import("../route/transport/index.js").Transport<
     } | {
         readonly type: "image_generation";
     } | undefined;
+    readonly reasoning?: {
+        readonly effort?: OpenResponsesOptions.ReasoningEffort | undefined;
+        readonly summary?: "auto" | "concise" | "detailed" | undefined;
+        readonly mode?: import("../schema/options.js").OpenString<"pro" | "standard"> | undefined;
+        readonly context?: import("../schema/options.js").OpenString<"all_turns" | "auto" | "current_turn"> | undefined;
+    } | undefined;
     readonly context_management?: readonly {
         readonly type: "compaction";
         readonly compact_threshold?: number | undefined;
@@ -1234,10 +1248,6 @@ export declare const route: Route<{
     readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority"> | undefined;
     readonly prompt_cache_key?: string | undefined;
     readonly include?: readonly import("../schema/options.js").OpenString<"code_interpreter_call.outputs" | "computer_call_output.output.image_url" | "file_search_call.results" | "message.input_image.image_url" | "message.output_text.logprobs" | "reasoning.encrypted_content" | "web_search_call.action.sources" | "web_search_call.results">[] | undefined;
-    readonly reasoning?: {
-        readonly effort?: OpenResponsesOptions.ReasoningEffort | undefined;
-        readonly summary?: "auto" | "concise" | "detailed" | undefined;
-    } | undefined;
     readonly text?: {
         readonly verbosity?: import("../schema/options.js").OpenString<"high" | "low" | "medium"> | undefined;
     } | undefined;
@@ -1293,6 +1303,12 @@ export declare const route: Route<{
     } | {
         readonly type: "image_generation";
     } | undefined;
+    readonly reasoning?: {
+        readonly effort?: OpenResponsesOptions.ReasoningEffort | undefined;
+        readonly summary?: "auto" | "concise" | "detailed" | undefined;
+        readonly mode?: import("../schema/options.js").OpenString<"pro" | "standard"> | undefined;
+        readonly context?: import("../schema/options.js").OpenString<"all_turns" | "auto" | "current_turn"> | undefined;
+    } | undefined;
     readonly context_management?: readonly {
         readonly type: "compaction";
         readonly compact_threshold?: number | undefined;

package/dist/protocols/openai-responses.js

@@ -10,7 +10,7 @@ import { LLMRequest } from "../schema/index.js";
 import { resolveEffortUpdates } from "../effort-updates.js";
 import { OpenResponses } from "./open-responses.js";
 import { OpenResponsesOptions } from "./utils/open-responses-options.js";
-import { JsonObject, optionalArray, optionalNull, ProviderShared } from "./shared.js";
+import { JsonObject, lenient, optionalArray, optionalNull, ProviderShared } from "./shared.js";
 import { ResponsesHostedTools } from "./utils/responses-hosted-tools.js";
 import { OpenResponsesChannel } from "./open-responses-channel.js";
 import { ResponsesCompaction } from "./utils/responses-compaction.js";
@@ -25,6 +25,12 @@ export const ContextManagement = Schema.Array(Schema.Struct({
     type: Schema.Literal("compaction"),
     compactThreshold: Schema.optional(Schema.Int.check(Schema.isGreaterThan(0))),
 }));
+export const ReasoningMode = ProviderShared.knownString();
+export const ReasoningContext = ProviderShared.knownString();
+const ReasoningOptions = Schema.Struct({
+    reasoningMode: lenient(ReasoningMode),
+    reasoningContext: lenient(ReasoningContext),
+});
 const OpenAIResponsesImageGenerationTool = Schema.Struct({
     type: Schema.tag("image_generation"),
     action: Schema.optional(Schema.Literals(["auto", "generate", "edit"])),
@@ -86,6 +92,12 @@ const OpenAIResponsesCoreFields = {
     ...OpenResponses.coreFields,
     tools: optionalArray(OpenAIResponsesTools),
     tool_choice: Schema.optional(OpenAIResponsesToolChoice),
+    reasoning: Schema.optional(Schema.Struct({
+        effort: Schema.optional(OpenResponsesOptions.ReasoningEffort),
+        summary: Schema.optional(Schema.Literals(["auto", "concise", "detailed"])),
+        mode: Schema.optional(ReasoningMode),
+        context: Schema.optional(ReasoningContext),
+    })),
     context_management: Schema.optional(Schema.Array(Schema.Struct({
         type: Schema.Literal("compaction"),
         compact_threshold: Schema.optional(Schema.Int.check(Schema.isGreaterThan(0))),
@@ -111,6 +123,8 @@ const adapter = {
 const supportsEffortUpdates = (request) => {
     if (request.providerOptions?.contextManagement !== undefined)
         return false;
+    if (request.providerOptions?.reasoningMode === "pro")
+        return false;
     if (Schema.is(Schema.Struct({ mode: Schema.Literal("pro") }))(request.http?.body?.reasoning))
         return false;
     const override = request.model.compatibility?.supportsEffortUpdates;
@@ -162,10 +176,21 @@ const lowerToolChoice = (toolChoice, tools) => ProviderShared.matchToolChoice(NA
 const fromRequest = Effect.fn("OpenAIResponses.fromRequest")(function* (request) {
     const management = yield* ProviderShared.validateWith(Schema.decodeUnknownEffect(Schema.UndefinedOr(ContextManagement)))(request.providerOptions?.contextManagement);
     const options = OpenResponsesOptions.resolve(request);
+    const reasoning = Schema.decodeUnknownSync(ReasoningOptions)(request.providerOptions ?? {});
     const updates = resolveEffortUpdates(request, options.reasoningEffort);
+    const generation = OpenResponses.lowerGeneration(request, { ...options, reasoningEffort: updates.effort });
     return {
         ...(yield* OpenResponses.lowerConversation(updates.request, adapter)),
-        ...OpenResponses.lowerGeneration(request, { ...options, reasoningEffort: updates.effort }),
+        ...generation,
+        ...(generation.reasoning || reasoning.reasoningMode || reasoning.reasoningContext
+            ? {
+                reasoning: {
+                    ...generation.reasoning,
+                    mode: reasoning.reasoningMode,
+                    context: reasoning.reasoningContext,
+                },
+            }
+            : {}),
         context_management: management?.map((edit) => ({ type: edit.type, compact_threshold: edit.compactThreshold })),
         tools: request.tools.length === 0 ? undefined : yield* lowerTools(request),
         tool_choice: request.tools.length === 0

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

@@ -80,6 +80,12 @@ export declare const protocol: Protocol<{
     readonly store?: boolean | undefined;
     readonly prompt_cache_key?: string | undefined;
     readonly reasoning_effort?: OpenResponsesOptions.ReasoningEffort | undefined;
+    readonly verbosity?: import("../schema/options.js").OpenString<"high" | "low" | "medium"> | undefined;
+    readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority"> | undefined;
+    readonly metadata?: {
+        readonly [x: string]: string;
+    } | undefined;
+    readonly safety_identifier?: string | undefined;
     readonly tool_stream?: boolean | undefined;
     readonly max_completion_tokens?: number | undefined;
     readonly max_tokens?: number | undefined;

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

@@ -180,6 +180,7 @@ export declare const protocol: Protocol<{
     readonly presence_penalty?: number | undefined;
     readonly frequency_penalty?: number | undefined;
     readonly stream: true;
+    readonly top_k?: number | undefined;
 }, string, {
     readonly [x: string]: unknown;
     readonly type: string;

package/dist/protocols/xai-responses.js

@@ -21,6 +21,10 @@ const XAIResponsesHostedToolItem = Schema.Union([
         error: Schema.optional(Schema.Unknown),
     }), [JsonObject]),
 ]);
+const Body = Schema.Struct({
+    ...OpenResponses.OpenResponsesBody.fields,
+    top_k: Schema.optional(Schema.Number),
+});
 const adapter = {
     id: ADAPTER,
     name: NAME,
@@ -43,7 +47,14 @@ const fromRequest = Effect.fn("XAIResponses.fromRequest")(function* (request) {
             route: request.model.route.id,
             message: "xAI requires explicit compaction through LLMClient.compact; automatic context management is not supported",
         });
-    return yield* OpenResponses.fromRequestWithAdapter(request, adapter);
+    const body = yield* OpenResponses.fromRequestWithAdapter(request, adapter);
+    // xAI Responses does not support penalties (they are ignored), but does support top_k.
+    return {
+        ...body,
+        presence_penalty: undefined,
+        frequency_penalty: undefined,
+        top_k: request.generation?.topK,
+    };
 });
 const HOSTED_TOOLS = {
     web_search_call: { name: "web_search", input: (item) => item.action ?? {} },
@@ -70,7 +81,7 @@ const step = (state, input) => {
 export const protocol = Protocol.make({
     id: ADAPTER,
     body: {
-        schema: OpenResponses.OpenResponsesBody,
+        schema: Body,
         from: fromRequest,
     },
     stream: {

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

@@ -147,6 +147,13 @@ export declare const protocol: Protocol<{
     readonly store?: boolean | undefined;
     readonly prompt_cache_key?: string | undefined;
     readonly reasoning_effort?: OpenResponsesOptions.ReasoningEffort | undefined;
+    readonly verbosity?: import("../schema/options.js").OpenString<"high" | "low" | "medium"> | undefined;
+    readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "default" | "flex" | "priority"> | undefined;
+    readonly parallel_tool_calls?: boolean | undefined;
+    readonly metadata?: {
+        readonly [x: string]: string;
+    } | undefined;
+    readonly safety_identifier?: string | undefined;
     readonly tool_stream?: boolean | undefined;
     readonly max_completion_tokens?: number | undefined;
     readonly max_tokens?: number | undefined;

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

@@ -298,6 +298,7 @@ export declare const protocol: Protocol<{
         readonly user_id?: string | null | undefined;
     } | undefined;
     readonly service_tier?: import("../schema/options.js").OpenString<"auto" | "standard_only"> | undefined;
+    readonly speed?: import("../schema/options.js").OpenString<"fast" | "standard"> | undefined;
     readonly thinking?: {
         readonly type: import("../schema/options.js").OpenString<"adaptive" | "disabled" | "enabled">;
     } | undefined;

package/dist/provider-error.js

@@ -203,7 +203,7 @@ export function classifyProviderFailure(input) {
         (codes.includes("context_length_exceeded") ||
             codes.includes("model_context_window_exceeded") ||
             codes.includes("request_too_large") ||
-            isContextOverflow(text)))
+            (!codes.includes("throttlingexception") && isContextOverflow(text))))
         return new InvalidRequestError({ ...details, classification: "context-overflow" });
     if (input.status === 413 || isPayloadTooLarge(text))
         return new InvalidRequestError({ ...details, classification: "payload-too-large" });