@openclaw/ai 2026.8.1-beta.1 → 2026.8.1-beta.3

package/dist/internal/anthropic.d.mts

@@ -1,11 +1,19 @@
-import { a as requiresClaudeMandatoryAdaptiveThinking, c as resolveClaudeMythos5ModelIdentity, d as resolveClaudeSonnet5ModelIdentity, g as supportsClaudeNativeXhighEffort, h as supportsClaudeNativeMaxEffort, i as requiresClaudeDefaultSampling, l as resolveClaudeNativeThinkingLevelMap, o as resolveClaudeFable5ModelIdentity, p as supportsClaudeAdaptiveThinking, s as resolveClaudeModelIdentity, u as resolveClaudeOpus5ModelIdentity } from "../anthropic-SrGtwsJu.mjs";
+import { a as requiresClaudeMandatoryAdaptiveThinking, c as resolveClaudeMythos5ModelIdentity, d as resolveClaudeSonnet5ModelIdentity, g as supportsClaudeNativeXhighEffort, h as supportsClaudeNativeMaxEffort, i as requiresClaudeDefaultSampling, l as resolveClaudeNativeThinkingLevelMap, o as resolveClaudeFable5ModelIdentity, p as supportsClaudeAdaptiveThinking, s as resolveClaudeModelIdentity, u as resolveClaudeOpus5ModelIdentity } from "../anthropic-BECQCNdF.mjs";
 import { t as AssistantMessageDiagnostic } from "../diagnostics-BaTA9eVl.mjs";
-import { E as Model, F as SimpleStreamOptions, I as StopReason, R as StreamFunction, X as Usage, u as Context } from "../types-CH7ReIcU.mjs";
-import { n as AnthropicOptions, r as AnthropicThinkingDisplay, t as AnthropicEffort } from "../provider-options-C5kYML7i.mjs";
+import { B as StopReason, D as Model, H as StreamFunction, d as Context, et as Usage, z as SimpleStreamOptions } from "../types-DTAP49HF.mjs";
+import { n as AnthropicOptions, r as AnthropicThinkingDisplay, t as AnthropicEffort } from "../provider-options-Cc5ZQ6Aa.mjs";
+import { o as resolveAnthropicServerCompactionPlan } from "../anthropic-payload-policy-CSx7MoKH.mjs";
+
 //#region packages/ai/src/providers/anthropic.d.ts
-declare const streamAnthropic: StreamFunction<"anthropic-messages", AnthropicOptions>;
+type AnthropicCompactionOptions = AnthropicOptions & {
+  anthropicServerCompaction?: boolean;
+  authProfileId?: string;
+};
+declare const streamAnthropic: StreamFunction<"anthropic-messages", AnthropicCompactionOptions>;
 type AnthropicSimpleStreamOptions = SimpleStreamOptions & {
-  toolChoice?: AnthropicOptions["toolChoice"];
+  anthropicServerCompaction?: boolean;
+  authProfileId?: string;
+  toolChoice?: AnthropicCompactionOptions["toolChoice"];
 };
 declare const streamSimpleAnthropic: StreamFunction<"anthropic-messages", AnthropicSimpleStreamOptions>;
 //#endregion
@@ -56,7 +64,13 @@ declare function resolveAnthropicThinkingEffort(model: Model<"anthropic-messages
 declare function mapAnthropicStopReason(reason: string | undefined): StopReason;
 /** Remove unsupported assistant prefills while preserving completed tool-use turns. */
 declare function prepareClaudeNoPrefillRequestContext(model: Model, context: Context): Context;
-declare function applyClaudeRequestContract(params: Record<string, unknown>, model: {
+type ClaudeSamplingRequestParams = {
+  temperature?: unknown;
+  top_p?: unknown;
+  top_k?: unknown;
+  service_tier?: unknown;
+};
+declare function applyClaudeRequestContract(params: ClaudeSamplingRequestParams, model: {
   id?: string;
   params?: Record<string, unknown>;
   api?: string;
@@ -207,4 +221,4 @@ declare function applyAnthropicMessageStartUsage(target: Usage, payload: Anthrop
 /** Keep cumulative billing separate from the final server-side iteration context. */
 declare function applyAnthropicMessageDeltaUsage(target: Usage, payload: AnthropicUsagePayload | undefined, messageStartPromptUsage: AnthropicPromptUsageSnapshot | undefined): void;
 //#endregion
-export { ANTHROPIC_CLAUDE_CODE_BILLING_SYSTEM_BLOCK, ANTHROPIC_CLAUDE_CODE_VERSION, ANTHROPIC_OMITTED_REASONING_TEXT, ANTHROPIC_SERVER_SIDE_FALLBACKS, ANTHROPIC_SERVER_SIDE_FALLBACK_BETA, AnthropicCacheWriteUsage, type AnthropicEffort, AnthropicFallbackBoundary, AnthropicIterationUsageResult, AnthropicIterationUsageSnapshot, type AnthropicOptions, AnthropicProjectedToolChoice, AnthropicPromptUsageSnapshot, type AnthropicThinkingDisplay, AnthropicToolProjection, CLAUDE_OPUS_FALLBACK_MODEL_COST, applyAnthropicFallbackBoundary, applyAnthropicMessageDeltaUsage, applyAnthropicMessageStartUsage, applyAnthropicRefusal, applyClaudeRequestContract, defaultsClaudeAdaptiveThinking, findActiveAnthropicToolTurnAssistantIndex, mapAnthropicStopReason, normalizeAnthropicToolCallId, normalizeAnthropicToolChoice, omitFoundryBearerCredentialHeaders, prepareClaudeNoPrefillRequestContext, projectAnthropicTools, readAnthropicCacheWriteUsage, readAnthropicFallbackBoundary, readAnthropicPromptUsageSnapshot, readAnthropicUsageTokenCount, readLastAnthropicIterationUsage, reconcileAnthropicToolChoice, requiresClaudeAdaptiveThinking, requiresClaudeDefaultSampling, requiresClaudeMandatoryAdaptiveThinking, resolveAnthropicFallbackServingModelCost, resolveAnthropicThinkingEffort, resolveClaudeFable5ModelIdentity, resolveClaudeModelIdentity, resolveClaudeMythos5ModelIdentity, resolveClaudeNativeThinkingLevelMap, resolveClaudeOpus5ModelIdentity, resolveClaudeSonnet5ModelIdentity, resolveModelBoundThinkingReplayMode, resolveOriginalAnthropicToolName, streamAnthropic, streamSimpleAnthropic, supportsClaudeAdaptiveThinking, supportsClaudeNativeMaxEffort, supportsClaudeNativeXhighEffort, toClaudeCodeToolName, usesClaudeFable5MessagesContract, usesClaudeStreamingRefusalContract, usesFoundryBearerAuth };
+export { ANTHROPIC_CLAUDE_CODE_BILLING_SYSTEM_BLOCK, ANTHROPIC_CLAUDE_CODE_VERSION, ANTHROPIC_OMITTED_REASONING_TEXT, ANTHROPIC_SERVER_SIDE_FALLBACKS, ANTHROPIC_SERVER_SIDE_FALLBACK_BETA, AnthropicCacheWriteUsage, type AnthropicEffort, AnthropicFallbackBoundary, AnthropicIterationUsageResult, AnthropicIterationUsageSnapshot, type AnthropicOptions, AnthropicProjectedToolChoice, AnthropicPromptUsageSnapshot, type AnthropicThinkingDisplay, AnthropicToolProjection, CLAUDE_OPUS_FALLBACK_MODEL_COST, applyAnthropicFallbackBoundary, applyAnthropicMessageDeltaUsage, applyAnthropicMessageStartUsage, applyAnthropicRefusal, applyClaudeRequestContract, defaultsClaudeAdaptiveThinking, findActiveAnthropicToolTurnAssistantIndex, mapAnthropicStopReason, normalizeAnthropicToolCallId, normalizeAnthropicToolChoice, omitFoundryBearerCredentialHeaders, prepareClaudeNoPrefillRequestContext, projectAnthropicTools, readAnthropicCacheWriteUsage, readAnthropicFallbackBoundary, readAnthropicPromptUsageSnapshot, readAnthropicUsageTokenCount, readLastAnthropicIterationUsage, reconcileAnthropicToolChoice, requiresClaudeAdaptiveThinking, requiresClaudeDefaultSampling, requiresClaudeMandatoryAdaptiveThinking, resolveAnthropicFallbackServingModelCost, resolveAnthropicServerCompactionPlan, resolveAnthropicThinkingEffort, resolveClaudeFable5ModelIdentity, resolveClaudeModelIdentity, resolveClaudeMythos5ModelIdentity, resolveClaudeNativeThinkingLevelMap, resolveClaudeOpus5ModelIdentity, resolveClaudeSonnet5ModelIdentity, resolveModelBoundThinkingReplayMode, resolveOriginalAnthropicToolName, streamAnthropic, streamSimpleAnthropic, supportsClaudeAdaptiveThinking, supportsClaudeNativeMaxEffort, supportsClaudeNativeXhighEffort, toClaudeCodeToolName, usesClaudeFable5MessagesContract, usesClaudeStreamingRefusalContract, usesFoundryBearerAuth };

package/dist/internal/anthropic.mjs

@@ -1,5 +1,5 @@
-import { a as requiresClaudeMandatoryAdaptiveThinking, c as resolveClaudeMythos5ModelIdentity, d as resolveClaudeSonnet5ModelIdentity, g as supportsClaudeNativeXhighEffort, h as supportsClaudeNativeMaxEffort, i as requiresClaudeDefaultSampling, l as resolveClaudeNativeThinkingLevelMap, o as resolveClaudeFable5ModelIdentity, p as supportsClaudeAdaptiveThinking, s as resolveClaudeModelIdentity, u as resolveClaudeOpus5ModelIdentity } from "../src-QkygScBs.mjs";
-import { a as ANTHROPIC_CLAUDE_CODE_BILLING_SYSTEM_BLOCK, c as defaultsClaudeAdaptiveThinking, d as requiresClaudeAdaptiveThinking, f as resolveAnthropicThinkingEffort, h as usesClaudeStreamingRefusalContract, l as mapAnthropicStopReason, m as usesClaudeFable5MessagesContract, o as ANTHROPIC_CLAUDE_CODE_VERSION, p as resolveModelBoundThinkingReplayMode, s as applyClaudeRequestContract, u as prepareClaudeNoPrefillRequestContext } from "../host-Bl7Kgddo.mjs";
-import { C as usesFoundryBearerAuth, S as omitFoundryBearerCredentialHeaders, _ as CLAUDE_OPUS_FALLBACK_MODEL_COST, a as readAnthropicUsageTokenCount, b as resolveAnthropicFallbackServingModelCost, c as normalizeAnthropicToolChoice, d as resolveOriginalAnthropicToolName, f as toClaudeCodeToolName, g as ANTHROPIC_SERVER_SIDE_FALLBACK_BETA, h as ANTHROPIC_SERVER_SIDE_FALLBACKS, i as readAnthropicPromptUsageSnapshot, l as projectAnthropicTools, m as findActiveAnthropicToolTurnAssistantIndex, n as applyAnthropicMessageStartUsage, o as readLastAnthropicIterationUsage, p as ANTHROPIC_OMITTED_REASONING_TEXT, r as readAnthropicCacheWriteUsage, s as normalizeAnthropicToolCallId, t as applyAnthropicMessageDeltaUsage, u as reconcileAnthropicToolChoice, v as applyAnthropicFallbackBoundary, x as applyAnthropicRefusal, y as readAnthropicFallbackBoundary } from "../anthropic-usage-Ma4iX6uG.mjs";
-import { n as streamAnthropic, r as streamSimpleAnthropic } from "../anthropic-CE-7SkF6.mjs";
-export { ANTHROPIC_CLAUDE_CODE_BILLING_SYSTEM_BLOCK, ANTHROPIC_CLAUDE_CODE_VERSION, ANTHROPIC_OMITTED_REASONING_TEXT, ANTHROPIC_SERVER_SIDE_FALLBACKS, ANTHROPIC_SERVER_SIDE_FALLBACK_BETA, CLAUDE_OPUS_FALLBACK_MODEL_COST, applyAnthropicFallbackBoundary, applyAnthropicMessageDeltaUsage, applyAnthropicMessageStartUsage, applyAnthropicRefusal, applyClaudeRequestContract, defaultsClaudeAdaptiveThinking, findActiveAnthropicToolTurnAssistantIndex, mapAnthropicStopReason, normalizeAnthropicToolCallId, normalizeAnthropicToolChoice, omitFoundryBearerCredentialHeaders, prepareClaudeNoPrefillRequestContext, projectAnthropicTools, readAnthropicCacheWriteUsage, readAnthropicFallbackBoundary, readAnthropicPromptUsageSnapshot, readAnthropicUsageTokenCount, readLastAnthropicIterationUsage, reconcileAnthropicToolChoice, requiresClaudeAdaptiveThinking, requiresClaudeDefaultSampling, requiresClaudeMandatoryAdaptiveThinking, resolveAnthropicFallbackServingModelCost, resolveAnthropicThinkingEffort, resolveClaudeFable5ModelIdentity, resolveClaudeModelIdentity, resolveClaudeMythos5ModelIdentity, resolveClaudeNativeThinkingLevelMap, resolveClaudeOpus5ModelIdentity, resolveClaudeSonnet5ModelIdentity, resolveModelBoundThinkingReplayMode, resolveOriginalAnthropicToolName, streamAnthropic, streamSimpleAnthropic, supportsClaudeAdaptiveThinking, supportsClaudeNativeMaxEffort, supportsClaudeNativeXhighEffort, toClaudeCodeToolName, usesClaudeFable5MessagesContract, usesClaudeStreamingRefusalContract, usesFoundryBearerAuth };
+import { a as requiresClaudeMandatoryAdaptiveThinking, c as resolveClaudeMythos5ModelIdentity, d as resolveClaudeSonnet5ModelIdentity, g as supportsClaudeNativeXhighEffort, h as supportsClaudeNativeMaxEffort, i as requiresClaudeDefaultSampling, l as resolveClaudeNativeThinkingLevelMap, o as resolveClaudeFable5ModelIdentity, p as supportsClaudeAdaptiveThinking, s as resolveClaudeModelIdentity, u as resolveClaudeOpus5ModelIdentity } from "../anthropic-BQXQudmA.mjs";
+import { S as usesClaudeStreamingRefusalContract, _ as prepareClaudeNoPrefillRequestContext, b as resolveModelBoundThinkingReplayMode, f as ANTHROPIC_CLAUDE_CODE_BILLING_SYSTEM_BLOCK, g as mapAnthropicStopReason, h as defaultsClaudeAdaptiveThinking, m as applyClaudeRequestContract, p as ANTHROPIC_CLAUDE_CODE_VERSION, v as requiresClaudeAdaptiveThinking, x as usesClaudeFable5MessagesContract, y as resolveAnthropicThinkingEffort } from "../host-DF_yxOsA.mjs";
+import { C as readAnthropicFallbackBoundary, D as usesFoundryBearerAuth, E as omitFoundryBearerCredentialHeaders, I as resolveAnthropicServerCompactionPlan, S as applyAnthropicFallbackBoundary, T as applyAnthropicRefusal, _ as ANTHROPIC_OMITTED_REASONING_TEXT, a as applyAnthropicMessageDeltaUsage, b as ANTHROPIC_SERVER_SIDE_FALLBACK_BETA, c as readAnthropicPromptUsageSnapshot, d as normalizeAnthropicToolCallId, f as normalizeAnthropicToolChoice, g as toClaudeCodeToolName, h as resolveOriginalAnthropicToolName, l as readAnthropicUsageTokenCount, m as reconcileAnthropicToolChoice, o as applyAnthropicMessageStartUsage, p as projectAnthropicTools, s as readAnthropicCacheWriteUsage, u as readLastAnthropicIterationUsage, v as findActiveAnthropicToolTurnAssistantIndex, w as resolveAnthropicFallbackServingModelCost, x as CLAUDE_OPUS_FALLBACK_MODEL_COST, y as ANTHROPIC_SERVER_SIDE_FALLBACKS } from "../anthropic-compaction-replay-4Zx7-aCr.mjs";
+import { n as streamAnthropic, r as streamSimpleAnthropic } from "../anthropic-B8KwTtlb.mjs";
+export { ANTHROPIC_CLAUDE_CODE_BILLING_SYSTEM_BLOCK, ANTHROPIC_CLAUDE_CODE_VERSION, ANTHROPIC_OMITTED_REASONING_TEXT, ANTHROPIC_SERVER_SIDE_FALLBACKS, ANTHROPIC_SERVER_SIDE_FALLBACK_BETA, CLAUDE_OPUS_FALLBACK_MODEL_COST, applyAnthropicFallbackBoundary, applyAnthropicMessageDeltaUsage, applyAnthropicMessageStartUsage, applyAnthropicRefusal, applyClaudeRequestContract, defaultsClaudeAdaptiveThinking, findActiveAnthropicToolTurnAssistantIndex, mapAnthropicStopReason, normalizeAnthropicToolCallId, normalizeAnthropicToolChoice, omitFoundryBearerCredentialHeaders, prepareClaudeNoPrefillRequestContext, projectAnthropicTools, readAnthropicCacheWriteUsage, readAnthropicFallbackBoundary, readAnthropicPromptUsageSnapshot, readAnthropicUsageTokenCount, readLastAnthropicIterationUsage, reconcileAnthropicToolChoice, requiresClaudeAdaptiveThinking, requiresClaudeDefaultSampling, requiresClaudeMandatoryAdaptiveThinking, resolveAnthropicFallbackServingModelCost, resolveAnthropicServerCompactionPlan, resolveAnthropicThinkingEffort, resolveClaudeFable5ModelIdentity, resolveClaudeModelIdentity, resolveClaudeMythos5ModelIdentity, resolveClaudeNativeThinkingLevelMap, resolveClaudeOpus5ModelIdentity, resolveClaudeSonnet5ModelIdentity, resolveModelBoundThinkingReplayMode, resolveOriginalAnthropicToolName, streamAnthropic, streamSimpleAnthropic, supportsClaudeAdaptiveThinking, supportsClaudeNativeMaxEffort, supportsClaudeNativeXhighEffort, toClaudeCodeToolName, usesClaudeFable5MessagesContract, usesClaudeStreamingRefusalContract, usesFoundryBearerAuth };

package/dist/internal/openai-responses-payload-policy.d.mts

@@ -0,0 +1,3 @@
+import { t as OPENAI_RESPONSES_APIS } from "../openai-responses-contracts-CD75pQ0m.mjs";
+import { i as resolveOpenAIResponsesServerCompactionPlan } from "../openai-responses-payload-policy-CNGFT9zr.mjs";
+export { OPENAI_RESPONSES_APIS, resolveOpenAIResponsesServerCompactionPlan };

package/dist/internal/openai-responses-payload-policy.mjs

@@ -0,0 +1,3 @@
+import { i as OPENAI_RESPONSES_APIS } from "../openai-responses-contracts-DvILoqBk.mjs";
+import { i as resolveOpenAIResponsesServerCompactionPlan } from "../openai-responses-payload-policy-BDo8CQsx.mjs";
+export { OPENAI_RESPONSES_APIS, resolveOpenAIResponsesServerCompactionPlan };

package/dist/internal/openai.d.mts

@@ -1,6 +1,7 @@
-import { E as Model, F as SimpleStreamOptions, I as StopReason, R as StreamFunction, X as Usage, u as Context, z as StreamOptions } from "../types-CH7ReIcU.mjs";
-import { _ as resolveOpenAIReasoningEffortForModel, a as OpenAICompletionsOptions, b as supportsOpenAITemperature, c as projectOpenAITools, d as OpenAIApiReasoningEffort, f as OpenAIReasoningEffort, g as normalizeOpenAIReasoningEffort, h as isOpenAIGpt56Model, l as reconcileOpenAICompletionsToolChoice, m as isOpenAIGpt55Model, o as OpenAICompletionsToolChoice, p as isOpenAIGpt54MiniModel, s as OpenAIToolProjection, u as reconcileOpenAIResponsesToolChoice, v as resolveOpenAISupportedReasoningEfforts, y as supportsOpenAIReasoningEffort } from "../provider-options-C5kYML7i.mjs";
-import { c as ResolvedOpenAICompletionsCompat, o as ResponsesPromptObservation, s as responsesPromptObserver } from "../openai-responses-contracts-BBKBfAqR.mjs";
+import { B as StopReason, D as Model, H as StreamFunction, d as Context, et as Usage, z as SimpleStreamOptions } from "../types-DTAP49HF.mjs";
+import { C as supportsOpenAITemperature, S as supportsOpenAIReasoningEffort, _ as isOpenAIGpt55Model, a as CodeModeToolSurfaceObservation, b as resolveOpenAIReasoningEffortForModel, c as codeModeToolSurfaceObserver, d as projectOpenAITools, f as reconcileOpenAICompletionsToolChoice, g as isOpenAIGpt54MiniModel, h as OpenAIReasoningEffort, i as BaseOpenAIStreamOptions, l as OpenAICompletionsToolChoice, m as OpenAIApiReasoningEffort, o as OpenAICompletionsOptions, p as reconcileOpenAIResponsesToolChoice, u as OpenAIToolProjection, v as isOpenAIGpt56Model, x as resolveOpenAISupportedReasoningEfforts, y as normalizeOpenAIReasoningEffort } from "../provider-options-Cc5ZQ6Aa.mjs";
+import { o as ResponsesPromptObservation, s as responsesPromptObserver } from "../openai-responses-contracts-CD75pQ0m.mjs";
+import { t as ResolvedOpenAICompletionsCompat } from "../openai-completions-compat-B0KTleXp.mjs";
 import { n as clampOpenAIPromptCacheKey, t as OPENAI_PROMPT_CACHE_KEY_MAX_LENGTH } from "../openai-prompt-cache-2uo_1OR1.mjs";
 import OpenAI from "openai";
 import { TSchema } from "typebox";
@@ -81,7 +82,7 @@ declare const streamOpenAICompletions: StreamFunction<"openai-completions", Open
 declare const streamSimpleOpenAICompletions: StreamFunction<"openai-completions", SimpleStreamOptions>;
 //#endregion
 //#region packages/ai/src/providers/openai-responses.d.ts
-interface OpenAIResponsesOptions extends StreamOptions {
+interface OpenAIResponsesOptions extends BaseOpenAIStreamOptions {
   reasoningEffort?: "minimal" | "low" | "medium" | "high" | "xhigh" | "max";
   reasoningSummary?: "auto" | "detailed" | "concise" | null;
   replayResponsesItemIds?: boolean;
@@ -223,7 +224,6 @@ type ToolSchemaCompatInput = {
   unsupportedToolSchemaKeywords?: unknown;
   omitEmptyArrayItems?: unknown;
 };
-declare function clearOpenAIToolSchemaCacheForTest(): void;
 /** Normalizes a tool parameter schema into the OpenAI strict JSON-schema subset. */
 declare function normalizeStrictOpenAIJsonSchema(schema: unknown, modelCompat?: ToolSchemaCompatInput | null): unknown;
 /** Normalizes tool parameters using strict OpenAI rules only when strict mode is active. */
@@ -257,4 +257,4 @@ type RuntimeToolInputSchemaProjection = {
 /** Projects one runtime tool input schema to JSON and reports runtime incompatibilities. */
 declare function projectRuntimeToolInputSchema(schema: unknown, path?: string): RuntimeToolInputSchemaProjection;
 //#endregion
-export { AZURE_RESPONSES_TEXT_CONTENT_PART_TYPE, AZURE_RESPONSES_TEXT_DELTA_EVENT_TYPE, AzureResponsesTextContentPart, AzureResponsesTextDeltaEvent, GEMINI_UNSUPPORTED_SCHEMA_KEYWORDS, LLAMACPP_GBNF_MAX_REPETITION_THRESHOLD, OPENAI_PROMPT_CACHE_KEY_MAX_LENGTH, OPENAI_RESPONSES_OUTPUT_TEXT_CONTENT_PART_TYPE, OPENAI_RESPONSES_OUTPUT_TEXT_DELTA_EVENT_TYPE, OpenAIApiReasoningEffort, type OpenAICompletionsOptions, OpenAICompletionsToolChoice, OpenAIReasoningEffort, OpenAIResponsesOptions, OpenAIStopReasonResult, OpenAIToolProjection, ResponsesMessageSnapshotCollapse, type ResponsesPromptObservation, ResponsesTerminalUsagePayload, ResponsesTextContentPartType, ResponsesTextDeltaEventType, ResponsesToolCallIdentity, ResponsesToolCallState, RuntimeToolInputSchemaJson, RuntimeToolInputSchemaProjection, ToolParameterSchemaOptions, ToolSchemaModelCompat, clampOpenAIPromptCacheKey, cleanSchemaForGemini, cleanSchemaForLlamacppGbnf, clearOpenAIToolSchemaCacheForTest, convertMessages, createResponsesToolCallTracker, extractToolSchemaModelCompat, findLlamacppGbnfSchemaViolations, findOpenAIStrictSchemaViolations, findOpenAIStrictToolProjectionDiagnostics, isAzureResponsesTextDeltaEvent, isAzureResponsesTextDeltaEventType, isOpenAICompatibleAzureResponsesBaseUrl, isOpenAIGpt54MiniModel, isOpenAIGpt55Model, isOpenAIGpt56Model, isResponsesTextContentPartType, isResponsesTextDeltaEventType, isStrictOpenAIJsonSchemaCompatible, isTraditionalAzureOpenAIHost, mapOpenAIStopReason, mapResponsesTerminalUsage, normalizeOpenAIReasoningEffort, normalizeOpenAIStrictCompatSchema, normalizeOpenAIStrictToolParameters, normalizeStrictOpenAIJsonSchema, normalizeToolParameterSchema, parseAzureDeploymentNameMap, projectOpenAITools, projectRuntimeToolInputSchema, readResponsesReasoningTokens, readResponsesToolCallItemIdentity, reconcileOpenAICompletionsToolChoice, reconcileOpenAIResponsesToolChoice, resolveAzureDeploymentNameFromMap, resolveOpenAIProjectedToolsStrictToolFlag, resolveOpenAIReasoningEffortForModel, resolveOpenAISupportedReasoningEfforts, resolveResponsesMessageSnapshotCollapse, resolveResponsesTerminalStopReason, resolveUnsupportedToolSchemaKeywords, responsesPromptObserver, shouldOmitEmptyArrayItems, streamOpenAICompletions, streamOpenAIResponses, streamSimpleOpenAICompletions, streamSimpleOpenAIResponses, stripUnsupportedSchemaKeywords, supportsOpenAIReasoningEffort, supportsOpenAITemperature };
+export { AZURE_RESPONSES_TEXT_CONTENT_PART_TYPE, AZURE_RESPONSES_TEXT_DELTA_EVENT_TYPE, AzureResponsesTextContentPart, AzureResponsesTextDeltaEvent, type CodeModeToolSurfaceObservation, GEMINI_UNSUPPORTED_SCHEMA_KEYWORDS, LLAMACPP_GBNF_MAX_REPETITION_THRESHOLD, OPENAI_PROMPT_CACHE_KEY_MAX_LENGTH, OPENAI_RESPONSES_OUTPUT_TEXT_CONTENT_PART_TYPE, OPENAI_RESPONSES_OUTPUT_TEXT_DELTA_EVENT_TYPE, OpenAIApiReasoningEffort, type OpenAICompletionsOptions, OpenAICompletionsToolChoice, OpenAIReasoningEffort, OpenAIResponsesOptions, OpenAIStopReasonResult, OpenAIToolProjection, ResponsesMessageSnapshotCollapse, type ResponsesPromptObservation, ResponsesTerminalUsagePayload, ResponsesTextContentPartType, ResponsesTextDeltaEventType, ResponsesToolCallIdentity, ResponsesToolCallState, RuntimeToolInputSchemaJson, RuntimeToolInputSchemaProjection, ToolParameterSchemaOptions, ToolSchemaModelCompat, clampOpenAIPromptCacheKey, cleanSchemaForGemini, cleanSchemaForLlamacppGbnf, codeModeToolSurfaceObserver, convertMessages, createResponsesToolCallTracker, extractToolSchemaModelCompat, findLlamacppGbnfSchemaViolations, findOpenAIStrictSchemaViolations, findOpenAIStrictToolProjectionDiagnostics, isAzureResponsesTextDeltaEvent, isAzureResponsesTextDeltaEventType, isOpenAICompatibleAzureResponsesBaseUrl, isOpenAIGpt54MiniModel, isOpenAIGpt55Model, isOpenAIGpt56Model, isResponsesTextContentPartType, isResponsesTextDeltaEventType, isStrictOpenAIJsonSchemaCompatible, isTraditionalAzureOpenAIHost, mapOpenAIStopReason, mapResponsesTerminalUsage, normalizeOpenAIReasoningEffort, normalizeOpenAIStrictCompatSchema, normalizeOpenAIStrictToolParameters, normalizeStrictOpenAIJsonSchema, normalizeToolParameterSchema, parseAzureDeploymentNameMap, projectOpenAITools, projectRuntimeToolInputSchema, readResponsesReasoningTokens, readResponsesToolCallItemIdentity, reconcileOpenAICompletionsToolChoice, reconcileOpenAIResponsesToolChoice, resolveAzureDeploymentNameFromMap, resolveOpenAIProjectedToolsStrictToolFlag, resolveOpenAIReasoningEffortForModel, resolveOpenAISupportedReasoningEfforts, resolveResponsesMessageSnapshotCollapse, resolveResponsesTerminalStopReason, resolveUnsupportedToolSchemaKeywords, responsesPromptObserver, shouldOmitEmptyArrayItems, streamOpenAICompletions, streamOpenAIResponses, streamSimpleOpenAICompletions, streamSimpleOpenAIResponses, stripUnsupportedSchemaKeywords, supportsOpenAIReasoningEffort, supportsOpenAITemperature };

package/dist/internal/openai.mjs

@@ -1,9 +1,11 @@
+import { S as supportsOpenAITemperature, _ as isOpenAIGpt56Model, b as resolveOpenAISupportedReasoningEfforts, g as isOpenAIGpt55Model, h as isOpenAIGpt54MiniModel, m as responsesPromptObserver, v as normalizeOpenAIReasoningEffort, x as supportsOpenAIReasoningEffort, y as resolveOpenAIReasoningEffortForModel } from "../openai-responses-contracts-DvILoqBk.mjs";
 import { n as clampOpenAIPromptCacheKey, t as OPENAI_PROMPT_CACHE_KEY_MAX_LENGTH } from "../openai-prompt-cache-mZTCdRPo.mjs";
-import { t as projectRuntimeToolInputSchema } from "../tool-schema-json-projection-B1b-XCn5.mjs";
-import { a as convertMessages, n as mapOpenAIStopReason } from "../openai-reasoning-compat-DKebnIBL.mjs";
-import { $ as shouldOmitEmptyArrayItems, B as responsesPromptObserver, G as normalizeOpenAIStrictToolParameters, H as clearOpenAIToolSchemaCacheForTest, J as findOpenAIStrictSchemaViolations, K as normalizeStrictOpenAIJsonSchema, Q as resolveUnsupportedToolSchemaKeywords, U as findOpenAIStrictToolProjectionDiagnostics, W as isStrictOpenAIJsonSchemaCompatible, X as extractToolSchemaModelCompat, Y as normalizeOpenAIStrictCompatSchema, Z as normalizeToolParameterSchema, _ as resolveResponsesMessageSnapshotCollapse, a as resolveResponsesTerminalStopReason, at as cleanSchemaForGemini, c as readResponsesToolCallItemIdentity, ct as isOpenAIGpt56Model, d as OPENAI_RESPONSES_OUTPUT_TEXT_CONTENT_PART_TYPE, dt as resolveOpenAISupportedReasoningEfforts, et as stripUnsupportedSchemaKeywords, f as OPENAI_RESPONSES_OUTPUT_TEXT_DELTA_EVENT_TYPE, ft as supportsOpenAIReasoningEffort, g as isResponsesTextDeltaEventType, h as isResponsesTextContentPartType, i as readResponsesReasoningTokens, it as GEMINI_UNSUPPORTED_SCHEMA_KEYWORDS, l as AZURE_RESPONSES_TEXT_CONTENT_PART_TYPE, lt as normalizeOpenAIReasoningEffort, m as isAzureResponsesTextDeltaEventType, nt as cleanSchemaForLlamacppGbnf, ot as isOpenAIGpt54MiniModel, p as isAzureResponsesTextDeltaEvent, pt as supportsOpenAITemperature, q as resolveOpenAIProjectedToolsStrictToolFlag, r as mapResponsesTerminalUsage, rt as findLlamacppGbnfSchemaViolations, s as createResponsesToolCallTracker, st as isOpenAIGpt55Model, tt as LLAMACPP_GBNF_MAX_REPETITION_THRESHOLD, u as AZURE_RESPONSES_TEXT_DELTA_EVENT_TYPE, ut as resolveOpenAIReasoningEffortForModel } from "../openai-responses-stream-internal-Bl8YyVtF.mjs";
-import { d as reconcileOpenAIResponsesToolChoice, l as projectOpenAITools, u as reconcileOpenAICompletionsToolChoice } from "../openai-transport-shared-Cipt7egQ.mjs";
+import { t as projectRuntimeToolInputSchema } from "../tool-schema-json-projection-FEr-jOoZ.mjs";
+import { r as convertMessages, t as mapOpenAIStopReason } from "../openai-stop-reason-9OONKyAm.mjs";
+import { $ as normalizeOpenAIStrictCompatSchema, A as isAzureResponsesTextDeltaEvent, D as AZURE_RESPONSES_TEXT_DELTA_EVENT_TYPE, E as AZURE_RESPONSES_TEXT_CONTENT_PART_TYPE, J as isStrictOpenAIJsonSchemaCompatible, M as isResponsesTextContentPartType, N as isResponsesTextDeltaEventType, O as OPENAI_RESPONSES_OUTPUT_TEXT_CONTENT_PART_TYPE, P as resolveResponsesMessageSnapshotCollapse, Q as findOpenAIStrictSchemaViolations, T as readResponsesToolCallItemIdentity, X as normalizeStrictOpenAIJsonSchema, Y as normalizeOpenAIStrictToolParameters, Z as resolveOpenAIProjectedToolsStrictToolFlag, at as LLAMACPP_GBNF_MAX_REPETITION_THRESHOLD, ct as GEMINI_UNSUPPORTED_SCHEMA_KEYWORDS, d as resolveResponsesTerminalStopReason, et as extractToolSchemaModelCompat, it as stripUnsupportedSchemaKeywords, j as isAzureResponsesTextDeltaEventType, k as OPENAI_RESPONSES_OUTPUT_TEXT_DELTA_EVENT_TYPE, l as mapResponsesTerminalUsage, lt as cleanSchemaForGemini, nt as resolveUnsupportedToolSchemaKeywords, ot as cleanSchemaForLlamacppGbnf, q as findOpenAIStrictToolProjectionDiagnostics, rt as shouldOmitEmptyArrayItems, st as findLlamacppGbnfSchemaViolations, tt as normalizeToolParameterSchema, u as readResponsesReasoningTokens, w as createResponsesToolCallTracker } from "../openai-responses-shared-Cvt5MgEm.mjs";
+import { t as codeModeToolSurfaceObserver } from "../provider-options-AvldWZt8.mjs";
+import { n as reconcileOpenAICompletionsToolChoice, r as reconcileOpenAIResponsesToolChoice, t as projectOpenAITools } from "../openai-tool-projection-CHGLt03m.mjs";
 import { i as resolveAzureDeploymentNameFromMap, n as isTraditionalAzureOpenAIHost, r as parseAzureDeploymentNameMap, t as isOpenAICompatibleAzureResponsesBaseUrl } from "../azure-openai-responses-client-compat-C7K7QfUE.mjs";
-import { n as streamOpenAICompletions, r as streamSimpleOpenAICompletions } from "../openai-completions-BaIpl93j.mjs";
-import { n as streamOpenAIResponses, r as streamSimpleOpenAIResponses } from "../openai-responses-Bdx3V_Yu.mjs";
-export { AZURE_RESPONSES_TEXT_CONTENT_PART_TYPE, AZURE_RESPONSES_TEXT_DELTA_EVENT_TYPE, GEMINI_UNSUPPORTED_SCHEMA_KEYWORDS, LLAMACPP_GBNF_MAX_REPETITION_THRESHOLD, OPENAI_PROMPT_CACHE_KEY_MAX_LENGTH, OPENAI_RESPONSES_OUTPUT_TEXT_CONTENT_PART_TYPE, OPENAI_RESPONSES_OUTPUT_TEXT_DELTA_EVENT_TYPE, clampOpenAIPromptCacheKey, cleanSchemaForGemini, cleanSchemaForLlamacppGbnf, clearOpenAIToolSchemaCacheForTest, convertMessages, createResponsesToolCallTracker, extractToolSchemaModelCompat, findLlamacppGbnfSchemaViolations, findOpenAIStrictSchemaViolations, findOpenAIStrictToolProjectionDiagnostics, isAzureResponsesTextDeltaEvent, isAzureResponsesTextDeltaEventType, isOpenAICompatibleAzureResponsesBaseUrl, isOpenAIGpt54MiniModel, isOpenAIGpt55Model, isOpenAIGpt56Model, isResponsesTextContentPartType, isResponsesTextDeltaEventType, isStrictOpenAIJsonSchemaCompatible, isTraditionalAzureOpenAIHost, mapOpenAIStopReason, mapResponsesTerminalUsage, normalizeOpenAIReasoningEffort, normalizeOpenAIStrictCompatSchema, normalizeOpenAIStrictToolParameters, normalizeStrictOpenAIJsonSchema, normalizeToolParameterSchema, parseAzureDeploymentNameMap, projectOpenAITools, projectRuntimeToolInputSchema, readResponsesReasoningTokens, readResponsesToolCallItemIdentity, reconcileOpenAICompletionsToolChoice, reconcileOpenAIResponsesToolChoice, resolveAzureDeploymentNameFromMap, resolveOpenAIProjectedToolsStrictToolFlag, resolveOpenAIReasoningEffortForModel, resolveOpenAISupportedReasoningEfforts, resolveResponsesMessageSnapshotCollapse, resolveResponsesTerminalStopReason, resolveUnsupportedToolSchemaKeywords, responsesPromptObserver, shouldOmitEmptyArrayItems, streamOpenAICompletions, streamOpenAIResponses, streamSimpleOpenAICompletions, streamSimpleOpenAIResponses, stripUnsupportedSchemaKeywords, supportsOpenAIReasoningEffort, supportsOpenAITemperature };
+import { n as streamOpenAICompletions, r as streamSimpleOpenAICompletions } from "../openai-completions-D88orSJ9.mjs";
+import { n as streamOpenAIResponses, r as streamSimpleOpenAIResponses } from "../openai-responses-BXEcr-Tp.mjs";
+export { AZURE_RESPONSES_TEXT_CONTENT_PART_TYPE, AZURE_RESPONSES_TEXT_DELTA_EVENT_TYPE, GEMINI_UNSUPPORTED_SCHEMA_KEYWORDS, LLAMACPP_GBNF_MAX_REPETITION_THRESHOLD, OPENAI_PROMPT_CACHE_KEY_MAX_LENGTH, OPENAI_RESPONSES_OUTPUT_TEXT_CONTENT_PART_TYPE, OPENAI_RESPONSES_OUTPUT_TEXT_DELTA_EVENT_TYPE, clampOpenAIPromptCacheKey, cleanSchemaForGemini, cleanSchemaForLlamacppGbnf, codeModeToolSurfaceObserver, convertMessages, createResponsesToolCallTracker, extractToolSchemaModelCompat, findLlamacppGbnfSchemaViolations, findOpenAIStrictSchemaViolations, findOpenAIStrictToolProjectionDiagnostics, isAzureResponsesTextDeltaEvent, isAzureResponsesTextDeltaEventType, isOpenAICompatibleAzureResponsesBaseUrl, isOpenAIGpt54MiniModel, isOpenAIGpt55Model, isOpenAIGpt56Model, isResponsesTextContentPartType, isResponsesTextDeltaEventType, isStrictOpenAIJsonSchemaCompatible, isTraditionalAzureOpenAIHost, mapOpenAIStopReason, mapResponsesTerminalUsage, normalizeOpenAIReasoningEffort, normalizeOpenAIStrictCompatSchema, normalizeOpenAIStrictToolParameters, normalizeStrictOpenAIJsonSchema, normalizeToolParameterSchema, parseAzureDeploymentNameMap, projectOpenAITools, projectRuntimeToolInputSchema, readResponsesReasoningTokens, readResponsesToolCallItemIdentity, reconcileOpenAICompletionsToolChoice, reconcileOpenAIResponsesToolChoice, resolveAzureDeploymentNameFromMap, resolveOpenAIProjectedToolsStrictToolFlag, resolveOpenAIReasoningEffortForModel, resolveOpenAISupportedReasoningEfforts, resolveResponsesMessageSnapshotCollapse, resolveResponsesTerminalStopReason, resolveUnsupportedToolSchemaKeywords, responsesPromptObserver, shouldOmitEmptyArrayItems, streamOpenAICompletions, streamOpenAIResponses, streamSimpleOpenAICompletions, streamSimpleOpenAIResponses, stripUnsupportedSchemaKeywords, supportsOpenAIReasoningEffort, supportsOpenAITemperature };

package/dist/internal/runtime.d.mts

@@ -1,5 +1,5 @@
-import { D as ModelThinkingLevel, E as Model, F as SimpleStreamOptions, P as ProviderStreamOptions, X as Usage, i as AssistantMessage, n as Api, o as AssistantMessageEventStreamContract, u as Context, z as StreamOptions } from "../types-CH7ReIcU.mjs";
-import { a as RegisteredApiProvider, t as ApiProvider } from "../api-registry-EpJoVwM1.mjs";
+import { D as Model, O as ModelThinkingLevel, R as ProviderStreamOptions, U as StreamOptions, a as AssistantMessage, d as Context, et as Usage, n as Api, s as AssistantMessageEventStreamContract, z as SimpleStreamOptions } from "../types-DTAP49HF.mjs";
+import { a as RegisteredApiProvider, t as ApiProvider } from "../api-registry-DFxSnFqG.mjs";
 import { t as sanitizeSurrogates } from "../sanitize-unicode-BZiVbGwK.mjs";
 import { a as createFirstStreamEventTimeoutError, c as withFirstStreamEventTimeout, i as createFirstStreamEventAbortController, n as FirstStreamEventInternalOptions, o as getFirstStreamEventTimeoutHandler, r as FirstStreamEventTimeoutContext, s as getFirstStreamEventTimeoutMs, t as FirstStreamEventAbortController } from "../stream-first-event-timeout-DvDeSucC.mjs";
 
@@ -117,6 +117,18 @@ declare function resolveOpenAICodexAccountId(token: string): string | null;
 //#region packages/ai/src/utils/overflow.d.ts
 /** Detects DS4-style raw token-count context overflow errors. */
 declare function isConfiguredContextSizeOverflowError(errorMessage: string): boolean;
+declare const CONTEXT_OVERFLOW_PATTERN_SCOPES: {
+  readonly "assistant-error": RegExp[];
+  readonly "failover-explicit": RegExp[];
+  readonly "provider-fallback": RegExp[];
+  readonly "failover-hint": readonly [RegExp];
+  readonly "context-window-too-small": readonly [RegExp];
+  readonly "tpm-rate-limit-hint": readonly [RegExp];
+  readonly "rate-limit-hint": readonly [RegExp];
+};
+type ContextOverflowMessageScope = keyof typeof CONTEXT_OVERFLOW_PATTERN_SCOPES;
+/** Match one canonical context-overflow wording scope without applying caller policy. */
+declare function matchesContextOverflowMessage(errorMessage: string, scope: ContextOverflowMessageScope): boolean;
 /**
  * Check if an assistant message represents a context overflow error.
  *
@@ -134,7 +146,7 @@ declare function isConfiguredContextSizeOverflowError(errorMessage: string): boo
  * - Google Gemini: "input token count exceeds the maximum"
  * - xAI (Grok): "maximum prompt length is X but request contains Y"
  * - Groq: "reduce the length of the messages"
- * - Cerebras: 400/413 status code (no body)
+ * - Cerebras: 413 status code (no body)
  * - Mistral: "Prompt contains X tokens ... too large for model with Y maximum context length"
  * - OpenRouter (all backends): "maximum context length is X tokens"
  * - Together AI: "The input (X tokens) is longer than the model's context length (Y tokens)."
@@ -161,7 +173,7 @@ declare function isConfiguredContextSizeOverflowError(errorMessage: string): boo
  * 1. Send a request that exceeds the model's context window
  * 2. Check the errorMessage in the response
  * 3. Create a regex pattern that matches the error
- * 4. The pattern should be added to OVERFLOW_PATTERNS in this file, or
+ * 4. The pattern should be added to the appropriate canonical scope in this file, or
  *    check the errorMessage yourself before calling this function
  *
  * @param message - The assistant message to check
@@ -186,6 +198,8 @@ interface ReasoningTagTextPartitioner {
   pushVisible(chunk: string): ReasoningTagTextDelta[];
   flush(): ReasoningTagTextDelta[];
   hasPending(): boolean;
+  /** Whether more input can change buffered Markdown or reasoning-tag ownership. */
+  hasPendingSyntax(): boolean;
   isInsideReasoning(): boolean;
 }
 /** Creates a block-incremental parser that emits only Markdown-stable text. */
@@ -221,4 +235,4 @@ type SseByteGuard = {
 };
 declare function createSseByteGuard(reader: ReadableStreamDefaultReader<Uint8Array>, opts: ReadSseStreamWithLimitOptions): SseByteGuard;
 //#endregion
-export { FirstStreamEventAbortController, FirstStreamEventInternalOptions, FirstStreamEventTimeoutContext, OpenAICodexJwtPayload, ReadSseStreamWithLimitOptions, type ReasoningTagTextDelta, type ReasoningTagTextPartitioner, SessionResourceCleanup, SseByteGuard, SseStreamOverflow, applyProviderReportedUsageCost, calculateCost, clampThinkingLevel, cleanupSessionResources, clearApiProviders, complete, completeSimple, createDeferredEventBuffer, createFirstStreamEventAbortController, createFirstStreamEventTimeoutError, createReasoningTagTextPartitioner, createSseByteGuard, decodeOpenAICodexJwtPayload, defaultApiRegistry, defaultLlmRuntime, findEnvKeys, getApiProvider, getApiProviders, getEnvApiKey, getFirstStreamEventTimeoutHandler, getFirstStreamEventTimeoutMs, getSupportedThinkingLevels, headersToRecord, isConfiguredContextSizeOverflowError, isContextOverflow, modelsAreEqual, notifyLlmRequestActivity, onLlmRequestActivity, parseJsonWithRepair, parseStreamingJson, registerSessionResourceCleanup, repairJson, resolveOpenAICodexAccountId, sanitizeSurrogates, shortHash, stream, streamSimple, withFirstStreamEventTimeout };
+export { ContextOverflowMessageScope, FirstStreamEventAbortController, FirstStreamEventInternalOptions, FirstStreamEventTimeoutContext, OpenAICodexJwtPayload, ReadSseStreamWithLimitOptions, type ReasoningTagTextDelta, type ReasoningTagTextPartitioner, SessionResourceCleanup, SseByteGuard, SseStreamOverflow, applyProviderReportedUsageCost, calculateCost, clampThinkingLevel, cleanupSessionResources, clearApiProviders, complete, completeSimple, createDeferredEventBuffer, createFirstStreamEventAbortController, createFirstStreamEventTimeoutError, createReasoningTagTextPartitioner, createSseByteGuard, decodeOpenAICodexJwtPayload, defaultApiRegistry, defaultLlmRuntime, findEnvKeys, getApiProvider, getApiProviders, getEnvApiKey, getFirstStreamEventTimeoutHandler, getFirstStreamEventTimeoutMs, getSupportedThinkingLevels, headersToRecord, isConfiguredContextSizeOverflowError, isContextOverflow, matchesContextOverflowMessage, modelsAreEqual, notifyLlmRequestActivity, onLlmRequestActivity, parseJsonWithRepair, parseStreamingJson, registerSessionResourceCleanup, repairJson, resolveOpenAICodexAccountId, sanitizeSurrogates, shortHash, stream, streamSimple, withFirstStreamEventTimeout };

package/dist/internal/runtime.mjs

@@ -1,15 +1,15 @@
 import { n as getEnvApiKey, t as findEnvKeys } from "../env-api-keys-DrgeBuva.mjs";
 import { n as createApiRegistry, t as createLlmRuntime } from "../stream-CREqxHgU.mjs";
-import { a as modelsAreEqual, i as getSupportedThinkingLevels, n as calculateCost, r as clampThinkingLevel, t as applyProviderReportedUsageCost } from "../model-utils-Dau5dlgm.mjs";
-import { t as sanitizeSurrogates } from "../sanitize-unicode-DT5o51ur.mjs";
+import { a as getSupportedThinkingLevels, i as clampThinkingLevel, n as applyProviderReportedUsageCost, o as modelsAreEqual, r as calculateCost, t as sanitizeSurrogates } from "../sanitize-unicode-BP7HLLZn.mjs";
+import { t as headersToRecord } from "../headers-B_e4-1J0.mjs";
 import { t as createDeferredEventBuffer } from "../deferred-event-buffer-DAvyP7qA.mjs";
-import { n as parseStreamingJson, r as repairJson, t as parseJsonWithRepair } from "../json-parse-BvXNt1-7.mjs";
+import { n as parseStreamingJson, r as repairJson, t as parseJsonWithRepair } from "../json-parse-CDnesDM_.mjs";
 import { n as onLlmRequestActivity, t as notifyLlmRequestActivity } from "../llm-request-activity-BjtkplhG.mjs";
-import { t as createReasoningTagTextPartitioner } from "../reasoning-tag-text-partitioner-CGDyLWUR.mjs";
-import { a as withFirstStreamEventTimeout, i as getFirstStreamEventTimeoutMs, n as createFirstStreamEventTimeoutError, r as getFirstStreamEventTimeoutHandler, t as createFirstStreamEventAbortController } from "../stream-first-event-timeout-BIBomOGq.mjs";
-import { t as headersToRecord } from "../headers-B_e4-1J0.mjs";
 import { t as shortHash } from "../hash-CHgqbJmD.mjs";
-import { i as registerSessionResourceCleanup, n as resolveOpenAICodexAccountId, r as cleanupSessionResources, t as decodeOpenAICodexJwtPayload } from "../openai-chatgpt-jwt-DhAAzLkj.mjs";
+import { a as withFirstStreamEventTimeout, i as getFirstStreamEventTimeoutMs, n as createFirstStreamEventTimeoutError, r as getFirstStreamEventTimeoutHandler, t as createFirstStreamEventAbortController } from "../stream-first-event-timeout-MK28puvq.mjs";
+import { t as createReasoningTagTextPartitioner } from "../reasoning-tag-text-partitioner-5ygO2rZc.mjs";
+import { n as registerSessionResourceCleanup, t as cleanupSessionResources } from "../session-resources-CkR4WWy1.mjs";
+import { n as resolveOpenAICodexAccountId, t as decodeOpenAICodexJwtPayload } from "../openai-chatgpt-jwt-KWcgd0d_.mjs";
 import { t as createSseByteGuard } from "../streaming-byte-guard-BrbkbwUu.mjs";
 //#region packages/ai/src/internal/default-runtime.ts
 const DEFAULT_RUNTIME_KEY = Symbol.for("openclaw.ai.defaultRuntime");
@@ -39,69 +39,82 @@ const CONFIGURED_CONTEXT_SIZE_OVERFLOW_RE = /prompt has [\d,]+ tokens?, but the
 function isConfiguredContextSizeOverflowError(errorMessage) {
 	return CONFIGURED_CONTEXT_SIZE_OVERFLOW_RE.test(errorMessage);
 }
-/**
-* Regex patterns to detect context overflow errors from different providers.
-*
-* These patterns match error messages returned when the input exceeds
-* the model's context window.
-*
-* Provider-specific patterns (with example error messages):
-*
-* - Anthropic: "prompt is too long: 213462 tokens > 200000 maximum"
-* - Anthropic: "413 {\"error\":{\"type\":\"request_too_large\",\"message\":\"Request exceeds the maximum size\"}}"
-* - OpenAI: "Your input exceeds the context window of this model"
-* - OpenAI/LiteLLM: "Requested token count exceeds the model's maximum context length of 131072 tokens"
-* - Google: "The input token count (1196265) exceeds the maximum number of tokens allowed (1048575)"
-* - xAI: "This model's maximum prompt length is 131072 but the request contains 537812 tokens"
-* - Groq: "Please reduce the length of the messages or completion"
-* - OpenRouter: "This endpoint's maximum context length is X tokens. However, you requested about Y tokens"
-* - Together AI: "The input (X tokens) is longer than the model's context length (Y tokens)."
-* - llama.cpp: "the request exceeds the available context size, try increasing it"
-* - LM Studio: "tokens to keep from the initial prompt is greater than the context length"
-* - GitHub Copilot: "prompt token count of X exceeds the limit of Y"
-* - MiniMax: "invalid params, context window exceeds limit"
-* - Kimi For Coding: "Your request exceeded model token limit: X (requested: Y)"
-* - Cerebras: "400/413 status code (no body)"
-* - Mistral: "Prompt contains X tokens ... too large for model with Y maximum context length"
-* - z.ai: May return "tokens in request more than max tokens allowed" (code 1210),
-*   "Prompt exceeds max length" (code 1261), or accept overflow silently; handled via the
-*   error patterns or usage.input > contextWindow
-* - Xiaomi MiMo: Truncates input to fill contextWindow exactly, then returns finish_reason "length"
-*   with output=0 (no room left to generate). Detected via stopReason "length" + zero output +
-*   input filling the context window.
-* - Ollama: Some deployments truncate silently, others return errors like "prompt too long; exceeded max context length by X tokens"
-*/
-const OVERFLOW_PATTERNS = [
-	/prompt is too long/i,
-	/request_too_large/i,
-	/input is too long for requested model/i,
-	/exceeds the context window/i,
-	/exceeds (?:the )?(?:model'?s )?maximum context length(?: of [\d,]+ tokens?|\s*\([\d,]+\))/i,
-	/input token count.*exceeds the maximum/i,
-	/maximum prompt length is \d+/i,
-	/reduce the length of the messages/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,
-	/exceeds the limit of \d+/i,
-	/exceeds the available context size/i,
-	/greater than the context length/i,
-	/context window exceeds limit/i,
-	/exceeded model token limit/i,
-	/tokens? in request more than max tokens? allowed/i,
-	/prompt exceeds max(?:imum)? length/i,
-	/too large for model with \d+ maximum context length/i,
-	CONFIGURED_CONTEXT_SIZE_OVERFLOW_RE,
-	/model_context_window_exceeded/i,
-	/prompt too long; exceeded (?:max )?context length/i,
-	/context[_ ]length[_ ]exceeded/i,
-	/too many tokens/i,
-	/token limit exceeded/i,
-	/^4(?:00|13)\s*(?:status code)?\s*\(no body\)/i
-];
+const CONTEXT_OVERFLOW_PATTERN_SCOPES = {
+	"assistant-error": [
+		/prompt is too long/i,
+		/request_too_large/i,
+		/input length and `?max_tokens`? exceed context limit: [\d,]+ \+ [\d,]+ > [\d,]+/i,
+		/input is too long for requested model/i,
+		/exceeds the context window/i,
+		/exceeds (?:the )?(?:model'?s )?maximum context length(?: of [\d,]+ tokens?|\s*\([\d,]+\))/i,
+		/input token count.*exceeds the maximum/i,
+		/maximum prompt length is \d+/i,
+		/reduce the length of the messages/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,
+		/exceeds the limit of \d+/i,
+		/exceeds the available context size/i,
+		/greater than the context length/i,
+		/context window exceeds limit/i,
+		/exceeded model token limit/i,
+		/tokens? in request more than max tokens? allowed/i,
+		/prompt exceeds max(?:imum)? length/i,
+		/too large for model with \d+ maximum context length/i,
+		CONFIGURED_CONTEXT_SIZE_OVERFLOW_RE,
+		/model_context_window_exceeded/i,
+		/prompt too long; exceeded (?:max )?context length/i,
+		/context[_ ]length[_ ]exceeded/i,
+		/too many tokens/i,
+		/token limit exceeded/i,
+		/^413\s*(?:status code)?\s*\(no body\)/i
+	],
+	"failover-explicit": [
+		/request_too_large/i,
+		/context_overflow/i,
+		CONFIGURED_CONTEXT_SIZE_OVERFLOW_RE,
+		/invalid_argument[\s\S]*maximum number of tokens/i,
+		/request exceeds the maximum size/i,
+		/context length exceeded/i,
+		/maximum context length/i,
+		/prompt is too long/i,
+		/prompt too long/i,
+		/exceeds model context window/i,
+		/model token limit/i,
+		/input exceeds[\s\S]*maximum number of tokens/i,
+		/^(?=[\s\S]*context window)(?=[\s\S]*ran out of (?:room|space))/i,
+		/request size exceeds[\s\S]*context window/i,
+		/context overflow:/i,
+		/exceed context limit/i,
+		/exceeds the model'?s maximum context/i,
+		/max_tokens[\s\S]*exceed[\s\S]*context/i,
+		/input length[\s\S]*exceed[\s\S]*context/i,
+		/413[\s\S]*too large/i,
+		/context_window_exceeded/i,
+		/input length [\d,]+\s+tokens? exceeds the model limit/i,
+		/上下文过长|上下文超出|上下文长度超|超出最大上下文|请压缩上下文/
+	],
+	"provider-fallback": [
+		/\binput token count exceeds the maximum number of input tokens\b/i,
+		/\binput is too long for this model\b/i,
+		/\binput exceeds the maximum number of tokens\b/i,
+		/\bollama error:\s*context length exceeded(?:,\s*too many tokens)?\b/i,
+		/\btotal tokens?.*exceeds? (?:the )?(?:model(?:'s)? )?(?:max|maximum|limit)/i,
+		/\b(?:request|prompt) \(\d[\d,]*\s*tokens?\) exceeds (?:the )?available context size\b/i,
+		/\binput (?:is )?too long for (?:the )?model\b/i
+	],
+	"failover-hint": [/context.*overflow|context window.*(too (?:large|long)|exceed|over|limit|max(?:imum)?|requested|sent|tokens)|prompt.*(too (?:large|long)|exceed|over|limit|max(?:imum)?)|(?:request|input).*(?:context|window|length|token).*(too (?:large|long)|exceed|over|limit|max(?:imum)?)/i],
+	"context-window-too-small": [/context window.*(too small|minimum is)/i],
+	"tpm-rate-limit-hint": [/\btpm\b|tokens per minute/i],
+	"rate-limit-hint": [/rate limit|too many requests|requests per (?:minute|hour|day)|quota|throttl|429\b|tokens per day/i]
+};
+/** Match one canonical context-overflow wording scope without applying caller policy. */
+function matchesContextOverflowMessage(errorMessage, scope) {
+	return CONTEXT_OVERFLOW_PATTERN_SCOPES[scope].some((pattern) => pattern.test(errorMessage));
+}
 /**
 * Patterns that indicate non-overflow errors (e.g. rate limiting, server errors).
-* Error messages matching unknown of these are excluded from overflow detection
+* Error messages matching any of these are excluded from overflow detection
 * even if they also match an OVERFLOW_PATTERN.
 *
 * Example: Bedrock formats throttling errors as "ThrottlingException: Too many tokens,
@@ -135,7 +148,7 @@ function resolveContextInputTokens(message) {
 * - Google Gemini: "input token count exceeds the maximum"
 * - xAI (Grok): "maximum prompt length is X but request contains Y"
 * - Groq: "reduce the length of the messages"
-* - Cerebras: 400/413 status code (no body)
+* - Cerebras: 413 status code (no body)
 * - Mistral: "Prompt contains X tokens ... too large for model with Y maximum context length"
 * - OpenRouter (all backends): "maximum context length is X tokens"
 * - Together AI: "The input (X tokens) is longer than the model's context length (Y tokens)."
@@ -162,7 +175,7 @@ function resolveContextInputTokens(message) {
 * 1. Send a request that exceeds the model's context window
 * 2. Check the errorMessage in the response
 * 3. Create a regex pattern that matches the error
-* 4. The pattern should be added to OVERFLOW_PATTERNS in this file, or
+* 4. The pattern should be added to the appropriate canonical scope in this file, or
 *    check the errorMessage yourself before calling this function
 *
 * @param message - The assistant message to check
@@ -172,7 +185,7 @@ function resolveContextInputTokens(message) {
 function isContextOverflow(message, contextWindow) {
 	if (message.stopReason === "error" && message.errorMessage) {
 		const errorMessage = message.errorMessage;
-		if (!NON_OVERFLOW_PATTERNS.some((p) => p.test(errorMessage)) && OVERFLOW_PATTERNS.some((p) => p.test(errorMessage))) return true;
+		if (!NON_OVERFLOW_PATTERNS.some((p) => p.test(errorMessage)) && matchesContextOverflowMessage(errorMessage, "assistant-error")) return true;
 	}
 	if (contextWindow && message.stopReason === "stop") {
 		const inputTokens = resolveContextInputTokens(message);
@@ -185,4 +198,4 @@ function isContextOverflow(message, contextWindow) {
 	return false;
 }
 //#endregion
-export { applyProviderReportedUsageCost, calculateCost, clampThinkingLevel, cleanupSessionResources, clearApiProviders, complete, completeSimple, createDeferredEventBuffer, createFirstStreamEventAbortController, createFirstStreamEventTimeoutError, createReasoningTagTextPartitioner, createSseByteGuard, decodeOpenAICodexJwtPayload, defaultApiRegistry, defaultLlmRuntime, findEnvKeys, getApiProvider, getApiProviders, getEnvApiKey, getFirstStreamEventTimeoutHandler, getFirstStreamEventTimeoutMs, getSupportedThinkingLevels, headersToRecord, isConfiguredContextSizeOverflowError, isContextOverflow, modelsAreEqual, notifyLlmRequestActivity, onLlmRequestActivity, parseJsonWithRepair, parseStreamingJson, registerSessionResourceCleanup, repairJson, resolveOpenAICodexAccountId, sanitizeSurrogates, shortHash, stream, streamSimple, withFirstStreamEventTimeout };
+export { applyProviderReportedUsageCost, calculateCost, clampThinkingLevel, cleanupSessionResources, clearApiProviders, complete, completeSimple, createDeferredEventBuffer, createFirstStreamEventAbortController, createFirstStreamEventTimeoutError, createReasoningTagTextPartitioner, createSseByteGuard, decodeOpenAICodexJwtPayload, defaultApiRegistry, defaultLlmRuntime, findEnvKeys, getApiProvider, getApiProviders, getEnvApiKey, getFirstStreamEventTimeoutHandler, getFirstStreamEventTimeoutMs, getSupportedThinkingLevels, headersToRecord, isConfiguredContextSizeOverflowError, isContextOverflow, matchesContextOverflowMessage, modelsAreEqual, notifyLlmRequestActivity, onLlmRequestActivity, parseJsonWithRepair, parseStreamingJson, registerSessionResourceCleanup, repairJson, resolveOpenAICodexAccountId, sanitizeSurrogates, shortHash, stream, streamSimple, withFirstStreamEventTimeout };

package/dist/internal/shared.d.mts

@@ -1,4 +1,4 @@
-import { E as Model, F as SimpleStreamOptions, H as ThinkingBudgets, T as Message, W as ThinkingLevel, i as AssistantMessage, n as Api, z as StreamOptions } from "../types-CH7ReIcU.mjs";
+import { D as Model, E as Message, J as ThinkingLevel, K as ThinkingBudgets, U as StreamOptions, a as AssistantMessage, n as Api, z as SimpleStreamOptions } from "../types-DTAP49HF.mjs";
 import { t as sanitizeSurrogates } from "../sanitize-unicode-BZiVbGwK.mjs";
 import { n as normalizeStructuredPromptSection, r as sortPromptCacheToolsByName, t as normalizePromptCapabilityIds } from "../prompt-cache-stability-Cwcjv_fx.mjs";
 
@@ -29,14 +29,23 @@ declare function isImageWithMediaPayload<T>(block: T): block is T & {
 };
 declare function describeToolResultMediaPlaceholder(blocks: readonly unknown[]): string | undefined;
 declare function extractToolResultBlockText(block: unknown): string | undefined;
-declare function extractToolResultText(blocks: readonly unknown[]): string;
+declare function extractToolResultText(blocks: readonly unknown[], options?: {
+  includeStructured?: boolean;
+}): string;
+type ToolResultMediaSupport = {
+  images: boolean;
+  audio: boolean;
+};
+/** Describe media that cannot be represented on the target provider wire. */
+declare function describeUnsupportedToolResultMedia(blocks: readonly unknown[], support: ToolResultMediaSupport): string | undefined;
+declare function formatToolResultText(params: {
+  text: string;
+  mediaPlaceholder?: string;
+  omittedMediaPlaceholder?: string;
+  isError: boolean;
+}): string;
 //#endregion
 //#region packages/ai/src/transcript-transform.d.ts
-/**
- * Normalize tool call ID for cross-provider compatibility.
- * OpenAI Responses API generates IDs that are 450+ chars with special characters like `|`.
- * Anthropic APIs require IDs matching ^[a-zA-Z0-9_-]+$ (max 64 chars).
- */
 declare function transformMessages<TApi extends Api>(messages: Message[], model: Model<TApi>, normalizeToolCallId?: (id: string, model: Model<TApi>, source: AssistantMessage) => string): Message[];
 //#endregion
 //#region packages/ai/src/utils/system-prompt-cache-boundary.d.ts
@@ -62,4 +71,4 @@ type TlsCertificateErrorDetails = {
 /** Classify deterministic Node/OpenSSL certificate validation failures. */
 declare function inspectTlsCertificateError(error: unknown): TlsCertificateErrorDetails | null;
 //#endregion
-export { SYSTEM_PROMPT_CACHE_BOUNDARY, TlsCertificateErrorDetails, TlsCertificateErrorKind, adjustMaxTokensForThinking, buildBaseOptions, clampMaxTokensToModel, clampReasoning, describeToolResultMediaPlaceholder, ensureSystemPromptCacheBoundary, extractToolResultBlockText, extractToolResultText, hasMediaPayload, inspectTlsCertificateError, isImageWithMediaPayload, normalizePromptCapabilityIds, normalizeStructuredPromptSection, prependSystemPromptAdditionAfterCacheBoundary, sanitizeSurrogates, sortPromptCacheToolsByName, splitSystemPromptCacheBoundary, stripSystemPromptCacheBoundary, transformMessages };
+export { SYSTEM_PROMPT_CACHE_BOUNDARY, TlsCertificateErrorDetails, TlsCertificateErrorKind, adjustMaxTokensForThinking, buildBaseOptions, clampMaxTokensToModel, clampReasoning, describeToolResultMediaPlaceholder, describeUnsupportedToolResultMedia, ensureSystemPromptCacheBoundary, extractToolResultBlockText, extractToolResultText, formatToolResultText, hasMediaPayload, inspectTlsCertificateError, isImageWithMediaPayload, normalizePromptCapabilityIds, normalizeStructuredPromptSection, prependSystemPromptAdditionAfterCacheBoundary, sanitizeSurrogates, sortPromptCacheToolsByName, splitSystemPromptCacheBoundary, stripSystemPromptCacheBoundary, transformMessages };

package/dist/internal/shared.mjs

@@ -1,6 +1,5 @@
-import { i as transformMessages } from "../host-Bl7Kgddo.mjs";
-import { t as sanitizeSurrogates } from "../sanitize-unicode-DT5o51ur.mjs";
-import { _ as normalizeStructuredPromptSection, a as isImageWithMediaPayload, c as buildBaseOptions, d as SYSTEM_PROMPT_CACHE_BOUNDARY, f as ensureSystemPromptCacheBoundary, g as normalizePromptCapabilityIds, h as stripSystemPromptCacheBoundary, i as hasMediaPayload, l as clampMaxTokensToModel, m as splitSystemPromptCacheBoundary, n as extractToolResultBlockText, p as prependSystemPromptAdditionAfterCacheBoundary, r as extractToolResultText, s as adjustMaxTokensForThinking, t as describeToolResultMediaPlaceholder, u as clampReasoning, v as sortPromptCacheToolsByName } from "../tool-result-text-Dvkp2Dus.mjs";
-import "../transform-messages-DfjpNXNQ.mjs";
+import { t as sanitizeSurrogates } from "../sanitize-unicode-BP7HLLZn.mjs";
+import { a as describeToolResultMediaPlaceholder, c as extractToolResultText, d as isImageWithMediaPayload, i as transformMessages, l as formatToolResultText, o as describeUnsupportedToolResultMedia, s as extractToolResultBlockText, u as hasMediaPayload } from "../host-DF_yxOsA.mjs";
+import { A as splitSystemPromptCacheBoundary, D as SYSTEM_PROMPT_CACHE_BOUNDARY, M as normalizePromptCapabilityIds, N as normalizeStructuredPromptSection, O as ensureSystemPromptCacheBoundary, P as sortPromptCacheToolsByName, i as clampReasoning, j as stripSystemPromptCacheBoundary, k as prependSystemPromptAdditionAfterCacheBoundary, n as buildBaseOptions, r as clampMaxTokensToModel, t as adjustMaxTokensForThinking } from "../simple-options-oqsXB6U3.mjs";
 import { t as inspectTlsCertificateError } from "../tls-certificate-errors-DXSpluKI.mjs";
-export { SYSTEM_PROMPT_CACHE_BOUNDARY, adjustMaxTokensForThinking, buildBaseOptions, clampMaxTokensToModel, clampReasoning, describeToolResultMediaPlaceholder, ensureSystemPromptCacheBoundary, extractToolResultBlockText, extractToolResultText, hasMediaPayload, inspectTlsCertificateError, isImageWithMediaPayload, normalizePromptCapabilityIds, normalizeStructuredPromptSection, prependSystemPromptAdditionAfterCacheBoundary, sanitizeSurrogates, sortPromptCacheToolsByName, splitSystemPromptCacheBoundary, stripSystemPromptCacheBoundary, transformMessages };
+export { SYSTEM_PROMPT_CACHE_BOUNDARY, adjustMaxTokensForThinking, buildBaseOptions, clampMaxTokensToModel, clampReasoning, describeToolResultMediaPlaceholder, describeUnsupportedToolResultMedia, ensureSystemPromptCacheBoundary, extractToolResultBlockText, extractToolResultText, formatToolResultText, hasMediaPayload, inspectTlsCertificateError, isImageWithMediaPayload, normalizePromptCapabilityIds, normalizeStructuredPromptSection, prependSystemPromptAdditionAfterCacheBoundary, sanitizeSurrogates, sortPromptCacheToolsByName, splitSystemPromptCacheBoundary, stripSystemPromptCacheBoundary, transformMessages };

package/dist/{json-parse-BvXNt1-7.mjs → json-parse-CDnesDM_.mjs}

@@ -1,3 +1,4 @@
+import { t as asNonArrayRecord } from "./record-coerce-DdXsgUd_.mjs";
 import { parse } from "partial-json";
 //#region packages/ai/src/utils/json-parse.ts
 const VALID_JSON_ESCAPES = /* @__PURE__ */ new Set([
@@ -102,9 +103,6 @@ function looksLikeWindowsPathPrefix(prefix) {
 	const tail = prefix.slice(-160);
 	return /(?:^|[^A-Za-z0-9])[A-Za-z]:(?:[\\/][^"\\/:*?<>|\r\n]*)*$/.test(tail);
 }
-function asStreamingJsonRecord(value) {
-	return value !== null && typeof value === "object" && !Array.isArray(value) ? value : {};
-}
 /**
 * Attempts to parse potentially incomplete JSON during streaming.
 * Always returns a valid object, even if the JSON is incomplete.
@@ -115,13 +113,13 @@ function asStreamingJsonRecord(value) {
 function parseStreamingJson(partialJson) {
 	if (!partialJson || partialJson.trim() === "") return {};
 	try {
-		return asStreamingJsonRecord(parseJsonWithRepair(partialJson));
+		return asNonArrayRecord(parseJsonWithRepair(partialJson));
 	} catch {
 		try {
-			return asStreamingJsonRecord(parse(partialJson));
+			return asNonArrayRecord(parse(partialJson));
 		} catch {
 			try {
-				return asStreamingJsonRecord(parse(repairJson(partialJson)));
+				return asNonArrayRecord(parse(repairJson(partialJson)));
 			} catch {
 				return {};
 			}