@linxiraos/pi-ai 1.1.6 → 1.1.8

package/CHANGELOG.md

@@ -2,6 +2,11 @@
 
 ## [Unreleased]
 
+## [1.1.8] - 2026-09-04
+
+- OMP sync v18.1.2–v18.1.5: Copilot auth standardization, transport-error-safe retries after tool calls, and prefix-binding control state.
+- Fixed the Z.ai OAuth key name to the Zeta brand (`zeta`) — the merge had reintroduced upstream's `oh-my-pi` fixture rows.
+
 ## [1.1.6] - 2026-08-30
 
 - 同步上游 OMP v18.0.9（`cc14e04f075d`）。
@@ -10,12 +15,6 @@
 
 - 同步上游 OMP v18.0.5 / v18.0.6：新增 Yolo-Auto / OpenRouter 浏览器登录与 DeepInfra 支持，空补全重试重构（withReplaySafeStreamRetry）。
 
-## [1.1.3] - 2026-08-25
-
-### Fixed
-
-- Republished as 1.1.3 to reset the latest tag after the broken 1.1.2 (no functional change over 1.1.1).
-
 ## [1.1.2] - 2026-08-25
 
 ### Fixed
@@ -43,3 +42,21 @@
 ### Changed
 
 - Reset the version to 1.0.0 and republished under the `@linxiraos/*` scope, breaking from the `@linxiraos` version lineage.
+- Fixed Gemini thought summaries occasionally leaking a raw `` ```thinking `` / `` ``````thinking `` fence delimiter into the reasoning block, so it no longer shows up as fence spam in the thinking display or persisted transcripts ([#8719](https://github.com/can1357/oh-my-pi/issues/8719)).
+- Fixed the OpenCode Go login prompting for an "OpenCode Zen API key": the shared login flow now names the provider you selected, so connecting OpenCode Go asks for an OpenCode Go key (the `opencode.ai/auth` console is still shared, as documented upstream) ([#8738](https://github.com/can1357/oh-my-pi/issues/8738)).
+- Fixed Anthropic-compatible endpoints with strict prompt validation (e.g. Z.AI GLM `api.z.ai/api/anthropic`, which rejects the whole request with `400 code 1213 "The prompt parameter was not received normally"`) failing sessions once a tool returned empty output on a vision-capable model: empty successful `tool_result` blocks now encode as `content: ""` instead of `content: []`, which both the official API and strict compatible endpoints accept.
+- Fixed `retry.usageReservePct` (Reserve Margin) ignoring Claude Fable/Mythos weekly tier usage until it hit 100%, so a Fable model kept serving turns past the configured reserve; reserve health now honors the mapped tier row while credential-wide hard blocks still require confirmed exhaustion ([#8773](https://github.com/can1357/oh-my-pi/issues/8773)).
+- Fixed `cursor-agent` streams stalling with "Provider stream stalled while waiting for the next event" when Cursor asked the client to approve a hosted WebFetch / web search (reproduced on `cursor-grok-4.6-xhigh` after "I'll fetch the page…"). Those `interaction_query` frames — including the newer WebFetch field 9 this proto did not name — were dropped, so the server waited forever and the idle watchdog aborted a live connection. Permission queries are now answered; hosted search/fetch is approved, unnamed permission fields get an `approved` reply on the same field number, and prompts this client cannot serve are rejected so the turn can continue.
+
+### Fixed
+
+- Fixed thinking effort selections being ignored for local Qwen 3.8+ models on llama.cpp and vLLM: the Qwen chat-completions dialects only toggled `enable_thinking`, so the chat template always reasoned at its `xhigh` default no matter which level was selected. The encoder now routes the requested effort onto the template's `reasoning_effort` kwarg (`chat_template_kwargs` for both Qwen dialects, plus the top-level field newer llama.cpp builds map natively).
+- Fixed OpenAI Completions, Amazon Bedrock, and Cursor providers ignoring `onPayload` replacement payloads. The hook now transforms the actual request body sent upstream on these providers, matching the Anthropic/Gemini/OpenAI Responses replacement contract. `devin-agent` still does not fire the hook (its payload is a protobuf object).
+- Fixed Codex requests failing outright when the signed-in ChatGPT account is not entitled to the requested model; the exact model denial is now classified as an account-policy error so credential rotation can reach an entitled sibling account
+- Fixed Perplexity email-OTP login after its verification response renamed the encrypted session token from `token` to `challenge_token`.
+- Cloud Code Assist Gemini 3.6/3.7 Flash requests at `minimal` now send `thinkingLevel: LOW` on the aliased `-low` SKU instead of `MINIMAL`, which the API rejects with HTTP 400.
+- Answer Cursor `interaction_query` permission gates (hosted web search, Exa, unnamed field-9 WebFetch) so the Run RPC continues instead of sitting silent until the 300s idle watchdog.
+- Fixed provider tool calls arriving with flattened array argument paths (e.g. Gemini's `questions[0].id`) being stripped and rejected by argument validation; well-formed flattened paths are now rebuilt into the nested arrays the tool schema expects ([#8886](https://github.com/can1357/oh-my-pi/issues/8886)).
+- Fixed opencode-go (Console Go) rejecting Responses turns with `400 No tool output found for tool call …` (naming a random call of the batch on each retry) when a model streamed a trailing text/thinking block after its tool calls: `buildResponsesInput` emitted that block as an assistant `message` item wedged between the `function_call` batch and its `function_call_output` items. Such interleaved messages are now hoisted ahead of their call batch (canonical `message(s) → calls → outputs`), which the strict gateway validator accepts; content is unchanged ([#8789](https://github.com/can1357/oh-my-pi/issues/8789)).
+- Fixed the OpenAI-wire transport sleeping on a LiteLLM concurrency-admission 429 (`rate_limit_type: max_parallel_requests`, `Retry-After: 60`) and retrying it up to 6 times (~300s) before session recovery saw the error. Because a 60s hint equals the transport's `maxDelayMs` cap, `fetchWithRetry` kept sleeping and retrying; the request now surfaces on the first attempt so `TurnRecovery`'s concurrency backoff/model fallback runs promptly. Genuine RPM/quota 429s (no such marker) still honor `Retry-After` ([#8854](https://github.com/can1357/oh-my-pi/issues/8854)).
+- Fixed OAuth login (Codex `localhost:1455`, and any `localhost` callback flow) failing on hosts with IPv6 disabled at the kernel (`ipv6.disable=1`). The `::1` companion listener added in #8081 fails there with Bun's generic "Is port X in use?" message (oven-sh/bun#7187), which the in-use check misread as a real collision — tearing down the healthy IPv4 listener and surfacing a bogus "port 1455 is in use" error. The dual-bind path now detects the missing IPv6 loopback up front and serves IPv4 alone ([#8814](https://github.com/can1357/oh-my-pi/issues/8814)).

package/dist/types/auth-gateway/types.d.ts

@@ -60,6 +60,8 @@ export interface AuthGatewayParsedRequestOptions {
     hideThinkingSummary?: boolean;
     /** Anthropic `output_config.task_budget` advisory loop budget. */
     taskBudget?: TokenTaskBudget;
+    /** Anthropic preserved-thinking behavior for a changed conversation prefix. */
+    anthropicPrefixMismatchBehavior?: "drop_block" | "error";
     /** OpenAI service tier (auto|default|flex|scale|priority). */
     serviceTier?: ServiceTier;
     /** Cache retention hint derived from inbound `cache_control` markers. */

package/dist/types/auth-storage.d.ts

@@ -3,7 +3,7 @@ import type { OAuthAuthInfo, OAuthController, OAuthCredentials, OAuthProviderId
 import type { Provider } from "./types.js";
 import type { ClientUsageIdentity, ClientUsageReport, ClientUsageSummary, CredentialRankingStrategy, ObservedUsageEntry, UsageHistoryEntry, UsageHistoryQuery, UsageLogger, UsageProvider, UsageReport } from "./usage.js";
 import { type CodexResetConsumeCode, type CodexResetCredit } from "./usage/openai-codex-reset.js";
-export { isSqliteBusyError, isSqliteCorruptionError, SqliteAuthCredentialStore, } from "./auth/sqlite-credential-store.js";
+export { isSqliteBusyError, isSqliteCorruptionError, SqliteAuthCredentialStore } from "./auth/sqlite-credential-store.js";
 export type ApiKeyCredential = {
     type: "api_key";
     key: string;
@@ -468,7 +468,7 @@ export type AuthStorageOptions = {
      * so the TUI can show where a token came from (broker URL or local SQLite path).
      *
      * Examples:
-     * - `"local ~/.zeta/agent/agent.db"`
+     * - `"local ~/.omp/agent/agent.db"`
      * - `"broker http://omp.internal:8765"`
      */
     sourceLabel?: string;

package/dist/types/dialect/thinking.d.ts

@@ -1,6 +1,15 @@
 import type { InbandScanEvent, InbandScanner } from "./types.js";
+export interface ThinkingInbandScannerOptions {
+    /**
+     * Report a bare reasoning close tag (no open seen) as an `impliedThinkingEnd`
+     * event instead of passing it through as text. Off by default: only a
+     * consumer that owns the whole message can reclassify the text before it.
+     */
+    readonly impliedOpen?: boolean;
+}
 export declare class ThinkingInbandScanner implements InbandScanner {
     #private;
+    constructor(options?: ThinkingInbandScannerOptions);
     feed(text: string): InbandScanEvent[];
     flush(): InbandScanEvent[];
 }

package/dist/types/dialect/types.d.ts

@@ -12,6 +12,16 @@ export type InbandScanEvent = {
 } | {
     type: "thinkingEnd";
     thinking: string;
+}
+/**
+ * A reasoning close tag with no open in this stream. Chat templates that
+ * prefill the opener into the prompt (DeepSeek-R1, Qwen3-Thinking) make the
+ * model stream only the close, so everything emitted as `text` before this
+ * event was reasoning. Only {@link ThinkingInbandScanner} with `impliedOpen`
+ * emits it.
+ */
+ | {
+    type: "impliedThinkingEnd";
 } | {
     type: "toolStart";
     id: string;

package/dist/types/error/flags.d.ts

@@ -81,12 +81,12 @@ export declare function isGrammarError(error: unknown): boolean;
  */
 export declare function isFastModeUnsupported(error: unknown): boolean;
 /**
- * GitHub Copilot 400 `model_not_supported` response for a model advertised by
- * `/models` — transient fleet skew, not a malformed request. Reads the
- * structural `code` through the SDK/body envelopes, then falls back to the
- * stringified body both SDK families put in `message`.
+ * Cline's gateway gates some roster entries (certain free-tier models) to its
+ * own product surfaces with a 403. The API key is valid — the restriction is
+ * per-model client policy — so it must neither rotate sibling credentials (they
+ * fail identically) nor surface as an auth failure.
  */
-export declare function isCopilotTransientModelError(error: unknown): boolean;
+export declare function isClinePassSurfaceGateMessage(errorMessage: string | undefined): boolean;
 export declare function classifyMessage(message: {
     api?: Api;
     provider?: string;

package/dist/types/error/rate-limit.d.ts

@@ -24,13 +24,15 @@ export declare function calculateRateLimitBackoffMs(reason: RateLimitReason): nu
 /**
  * HTTP status codes that, absent richer body classification, represent an
  * account-local usage cap rather than a bad credential or a transient blip.
- * HTTP 402 Payment Required is categorically an account-billing cap (xAI
+ * HTTP 402 Payment Required represents an account-billing cap (xAI
  * Grok Build "usage balance exhausted", DeepSeek "Insufficient Balance",
- * OpenRouter credit exhaustion) — never a transient blip or bad credential.
- * Always combine with {@link isUsageLimitOutcome} when a message is available
- * — a 429 carrying transient rate-limit wording is NOT a usage cap.
+ * OpenRouter credit exhaustion) when opaque, payment/deactivation/balance-worded,
+ * or QUOTA_EXHAUSTED/CONCURRENT_LIMIT, while informative non-quota 402s (e.g.
+ * endpoint subscription requirements) remain non-usage-limits. Always combine
+ * with {@link isUsageLimitOutcome} when a message is available.
  */
 export declare function isUsageLimitStatus(status: number | undefined): boolean;
+export declare function is402BillingCapBody(message: string | undefined): boolean;
 /**
  * Returns true for failures that should burn one credential and rotate to a
  * sibling account. Decision tree:
@@ -43,13 +45,13 @@ export declare function isUsageLimitStatus(status: number | undefined): boolean;
  *     empty JSON, HTTP framing only) → rotate conservatively: the server
  *     gave us nothing else to go on.
  *  4. Body has content → defer to {@link parseRateLimitReason}. `QUOTA_EXHAUSTED`
- *     rotates; for the categorical 402 billing cap a `CONCURRENT_LIMIT` body
- *     also rotates (the cap is concurrent-worded but the status is still an
- *     exhausted billing cap). `RATE_LIMIT_EXCEEDED` (`Too many requests`,
- *     per-minute caps), `MODEL_CAPACITY_EXHAUSTED` (`Service overloaded`),
- *     `SERVER_ERROR`, and `UNKNOWN` (`Please retry in 5s`) stay in the
- *     provider's own backoff layer so transient 429s don't burn sibling
- *     credentials.
+ *     rotates; for a 402 status a `CONCURRENT_LIMIT` body also rotates (the cap
+ *     is concurrent-worded but the status is an exhausted billing cap).
+ *     `RATE_LIMIT_EXCEEDED` (`Too many requests`, per-minute caps),
+ *     `MODEL_CAPACITY_EXHAUSTED` (`Service overloaded`), `SERVER_ERROR`, and
+ *     `UNKNOWN` (e.g. "A subscription is required for this endpoint" or
+ *     "Please retry in 5s") stay in the provider's own backoff / failure
+ *     layer so transient or non-quota responses don't burn sibling credentials.
  */
 export declare function isUsageLimitOutcome(status: number | undefined, message: string | undefined): boolean;
 /**
@@ -78,6 +80,6 @@ export declare function isAccountScopedCapText(message: string): boolean;
 /**
  * A concurrency cap on a non-billing status is shed-and-backoff, not
  * credential-rotatable. This mirrors the exclusion in {@link isUsageLimitOutcome}
- * for the 403 auth-retry entry points. A 402 remains a categorical billing cap.
+ * for the 403 auth-retry entry points. A 402 remains an account-billing cap.
  */
 export declare function isConcurrencyCapExclusion(status: number | undefined, message: string | undefined): boolean;

package/dist/types/error/retryable.d.ts

@@ -7,13 +7,6 @@
  * may match more. Use this when you only have a `status: number`.
  */
 export declare function isTransientStatus(status: number | undefined): boolean;
-/** Hook for provider-specific transient detection that the error module must not import directly. */
-export interface ProviderRetryableHooks {
-    /** Provider id of the failing request, used to gate provider-specific checks. */
-    provider?: string;
-    /** Provider-specific transient predicate (e.g. Copilot model-availability 400s). */
-    isProviderTransient?: (error: Error) => boolean;
-}
 /**
  * Whether a provider stream error should be retried against the same credential.
  *
@@ -21,7 +14,7 @@ export interface ProviderRetryableHooks {
  * here — they are owned by the credential-rotation layer (auth-gateway /
  * `streamSimple` a/b/c policy), not this seconds-scale provider backoff.
  *
- * Provider-specific transient cases are injected via {@link ProviderRetryableHooks}
- * so this stays free of provider imports.
+ * Every 4xx other than 408/429 is terminal: a request the provider rejected
+ * as malformed, unauthorized, or unentitled fails identically on replay.
  */
-export declare function isProviderRetryableError(error: unknown, hooks?: ProviderRetryableHooks): boolean;
+export declare function isProviderRetryableError(error: unknown): boolean;

package/dist/types/index.d.ts

@@ -46,7 +46,6 @@ export * from "./utils/anthropic-auth.js";
 export * from "./utils/event-stream.js";
 export * from "./utils/glyph-codec.js";
 export * from "./utils/openrouter-headers.js";
-export * from "./utils/retry.js";
 export * from "./utils/schema/index.js";
 export * from "./utils/thinking-loop.js";
 export * from "./utils/tool-call-loop-guard.js";

package/dist/types/providers/anthropic-messages-server-schema.d.ts

@@ -196,24 +196,56 @@ export declare const userMessageSchema: import("@linxiraos/pi-omptype").FluentTy
     role: "user";
 }>;
 export declare const systemMessageSchema: import("@linxiraos/pi-omptype").FluentType<{
-    content: string | {
+    clear_at?: "never" | "next_user_message" | undefined;
+    content: string | ({
         cache_control?: {
             ttl?: "1h" | "5m" | undefined;
             type: "ephemeral";
         } | undefined;
         text: string;
         type: "text";
-    }[];
+    } | {
+        tool: {
+            name: string;
+            type: "tool_reference";
+        };
+        type: "tool_addition";
+    } | {
+        tool: {
+            name: string;
+            type: "tool_reference";
+        };
+        type: "tool_removal";
+    })[];
+    output_config?: {
+        effort?: "high" | "low" | "max" | "medium" | "xhigh" | undefined;
+    } | undefined;
     role: "system";
 }, {
-    content: string | {
+    clear_at?: "never" | "next_user_message" | undefined;
+    content: string | ({
         cache_control?: {
             ttl?: "1h" | "5m" | undefined;
             type: "ephemeral";
         } | undefined;
         text: string;
         type: "text";
-    }[];
+    } | {
+        tool: {
+            name: string;
+            type: "tool_reference";
+        };
+        type: "tool_addition";
+    } | {
+        tool: {
+            name: string;
+            type: "tool_reference";
+        };
+        type: "tool_removal";
+    })[];
+    output_config?: {
+        effort?: "high" | "low" | "max" | "medium" | "xhigh" | undefined;
+    } | undefined;
     role: "system";
 }>;
 export declare const assistantMessageSchema: import("@linxiraos/pi-omptype").FluentType<{
@@ -352,14 +384,30 @@ export declare const messageSchema: import("@linxiraos/pi-omptype").FluentType<{
     })[];
     role: "user";
 } | {
-    content: string | {
+    clear_at?: "never" | "next_user_message" | undefined;
+    content: string | ({
         cache_control?: {
             ttl?: "1h" | "5m" | undefined;
             type: "ephemeral";
         } | undefined;
         text: string;
         type: "text";
-    }[];
+    } | {
+        tool: {
+            name: string;
+            type: "tool_reference";
+        };
+        type: "tool_addition";
+    } | {
+        tool: {
+            name: string;
+            type: "tool_reference";
+        };
+        type: "tool_removal";
+    })[];
+    output_config?: {
+        effort?: "high" | "low" | "max" | "medium" | "xhigh" | undefined;
+    } | undefined;
     role: "system";
 } | {
     content: string | ({
@@ -460,14 +508,30 @@ export declare const messageSchema: import("@linxiraos/pi-omptype").FluentType<{
     })[];
     role: "user";
 } | {
-    content: string | {
+    clear_at?: "never" | "next_user_message" | undefined;
+    content: string | ({
         cache_control?: {
             ttl?: "1h" | "5m" | undefined;
             type: "ephemeral";
         } | undefined;
         text: string;
         type: "text";
-    }[];
+    } | {
+        tool: {
+            name: string;
+            type: "tool_reference";
+        };
+        type: "tool_addition";
+    } | {
+        tool: {
+            name: string;
+            type: "tool_reference";
+        };
+        type: "tool_removal";
+    })[];
+    output_config?: {
+        effort?: "high" | "low" | "max" | "medium" | "xhigh" | undefined;
+    } | undefined;
     role: "system";
 } | {
     content: string | ({
@@ -511,6 +575,7 @@ export declare const toolSchema: import("@linxiraos/pi-omptype").FluentType<{
         ttl?: "1h" | "5m" | undefined;
         type: "ephemeral";
     } | undefined;
+    defer_loading?: boolean | undefined;
     description?: string | undefined;
     input_schema: Record<string, unknown>;
     name: string;
@@ -519,6 +584,7 @@ export declare const toolSchema: import("@linxiraos/pi-omptype").FluentType<{
         ttl?: "1h" | "5m" | undefined;
         type: "ephemeral";
     } | undefined;
+    defer_loading?: boolean | undefined;
     description?: string | undefined;
     input_schema: Record<string, unknown>;
     name: string;
@@ -554,14 +620,23 @@ export declare const thinkingConfigSchema: import("@linxiraos/pi-omptype").Fluen
     display?: unknown;
     type: "disabled";
 } | {
+    block_binding?: {
+        prefix_mismatch_behavior: "drop_block" | "error";
+    } | undefined;
     budget_tokens?: number | undefined;
     display?: unknown;
     type: "adaptive";
 } | {
+    block_binding?: {
+        prefix_mismatch_behavior: "drop_block" | "error";
+    } | undefined;
     budget_tokens: number;
     display?: unknown;
     type: "enabled";
 }, {
+    block_binding?: {
+        prefix_mismatch_behavior: "drop_block" | "error";
+    } | undefined;
     budget_tokens: number;
     display?: unknown;
     type: "enabled";
@@ -569,6 +644,9 @@ export declare const thinkingConfigSchema: import("@linxiraos/pi-omptype").Fluen
     display?: unknown;
     type: "disabled";
 } | {
+    block_binding?: {
+        prefix_mismatch_behavior: "drop_block" | "error";
+    } | undefined;
     budget_tokens?: number | undefined;
     display?: unknown;
     type: "adaptive";
@@ -641,14 +719,30 @@ export declare const anthropicMessagesRequestSchema: import("@linxiraos/pi-ompty
         })[];
         role: "user";
     } | {
-        content: string | {
+        clear_at?: "never" | "next_user_message" | undefined;
+        content: string | ({
             cache_control?: {
                 ttl?: "1h" | "5m" | undefined;
                 type: "ephemeral";
             } | undefined;
             text: string;
             type: "text";
-        }[];
+        } | {
+            tool: {
+                name: string;
+                type: "tool_reference";
+            };
+            type: "tool_addition";
+        } | {
+            tool: {
+                name: string;
+                type: "tool_reference";
+            };
+            type: "tool_removal";
+        })[];
+        output_config?: {
+            effort?: "high" | "low" | "max" | "medium" | "xhigh" | undefined;
+        } | undefined;
         role: "system";
     } | {
         content: string | ({
@@ -714,10 +808,16 @@ export declare const anthropicMessagesRequestSchema: import("@linxiraos/pi-ompty
         display?: unknown;
         type: "disabled";
     } | {
+        block_binding?: {
+            prefix_mismatch_behavior: "drop_block" | "error";
+        } | undefined;
         budget_tokens?: number | undefined;
         display?: unknown;
         type: "adaptive";
     } | {
+        block_binding?: {
+            prefix_mismatch_behavior: "drop_block" | "error";
+        } | undefined;
         budget_tokens: number;
         display?: unknown;
         type: "enabled";
@@ -741,6 +841,7 @@ export declare const anthropicMessagesRequestSchema: import("@linxiraos/pi-ompty
             ttl?: "1h" | "5m" | undefined;
             type: "ephemeral";
         } | undefined;
+        defer_loading?: boolean | undefined;
         description?: string | undefined;
         input_schema: Record<string, unknown>;
         name: string;
@@ -815,14 +916,30 @@ export declare const anthropicMessagesRequestSchema: import("@linxiraos/pi-ompty
         })[];
         role: "user";
     } | {
-        content: string | {
+        clear_at?: "never" | "next_user_message" | undefined;
+        content: string | ({
             cache_control?: {
                 ttl?: "1h" | "5m" | undefined;
                 type: "ephemeral";
             } | undefined;
             text: string;
             type: "text";
-        }[];
+        } | {
+            tool: {
+                name: string;
+                type: "tool_reference";
+            };
+            type: "tool_addition";
+        } | {
+            tool: {
+                name: string;
+                type: "tool_reference";
+            };
+            type: "tool_removal";
+        })[];
+        output_config?: {
+            effort?: "high" | "low" | "max" | "medium" | "xhigh" | undefined;
+        } | undefined;
         role: "system";
     } | {
         content: string | ({
@@ -885,6 +1002,9 @@ export declare const anthropicMessagesRequestSchema: import("@linxiraos/pi-ompty
     }[] | undefined;
     temperature?: number | undefined;
     thinking?: {
+        block_binding?: {
+            prefix_mismatch_behavior: "drop_block" | "error";
+        } | undefined;
         budget_tokens: number;
         display?: unknown;
         type: "enabled";
@@ -892,6 +1012,9 @@ export declare const anthropicMessagesRequestSchema: import("@linxiraos/pi-ompty
         display?: unknown;
         type: "disabled";
     } | {
+        block_binding?: {
+            prefix_mismatch_behavior: "drop_block" | "error";
+        } | undefined;
         budget_tokens?: number | undefined;
         display?: unknown;
         type: "adaptive";
@@ -915,6 +1038,7 @@ export declare const anthropicMessagesRequestSchema: import("@linxiraos/pi-ompty
             ttl?: "1h" | "5m" | undefined;
             type: "ephemeral";
         } | undefined;
+        defer_loading?: boolean | undefined;
         description?: string | undefined;
         input_schema: Record<string, unknown>;
         name: string;

package/dist/types/providers/anthropic-messages-server.d.ts

@@ -8,7 +8,7 @@ import type { AuthGatewayStreamControl, AuthGatewayParsedRequest as ParsedReques
 export type { ParsedRequest };
 export declare function parseRequest(body: unknown, headers?: Headers): ParsedRequest;
 export declare function encodeResponse(message: AssistantMessage, requestedModelId: string): Record<string, unknown>;
-export declare function encodeStream(events: AssistantMessageEventStream, requestedModelId: string, _options?: ParsedRequest["options"], control?: AuthGatewayStreamControl): ReadableStream<Uint8Array>;
+export declare function encodeStream(events: AssistantMessageEventStream, requestedModelId: string, options?: ParsedRequest["options"], control?: AuthGatewayStreamControl): ReadableStream<Uint8Array>;
 /**
  * Anthropic error envelope: `{ type: "error", error: { type, message } }`.
  * See https://docs.anthropic.com/en/api/errors. Returned as a `Response` so

package/dist/types/providers/anthropic-wire.d.ts

@@ -11,7 +11,9 @@
  * `scope`, tool `strict`/`eager_input_streaming`, mid-conversation `system`
  * role) are first-class here instead of being patched in via casts.
  */
-import type { TokenTaskBudget } from "../types.js";
+import type { ProviderInputTransformation, TokenTaskBudget } from "../types.js";
+/** Beta enabling preserved-thinking block controls and transformation reports. */
+export declare const THINKING_BINDING_CONTROLS_BETA = "thinking-binding-controls-2026-08-01";
 /** Ephemeral prefix-cache breakpoint marker. */
 export type CacheControlEphemeral = {
     type: "ephemeral";
@@ -87,6 +89,18 @@ export type ToolSearchToolResultBlockParam = {
     content: unknown;
     [key: string]: unknown;
 };
+export type ToolChangeReferenceParam = {
+    type: "tool_reference";
+    name: string;
+};
+export type ToolAdditionBlockParam = {
+    type: "tool_addition";
+    tool: ToolChangeReferenceParam;
+};
+export type ToolRemovalBlockParam = {
+    type: "tool_removal";
+    tool: ToolChangeReferenceParam;
+};
 /** Anthropic server-tool history variants omp can replay atomically. */
 export type AnthropicServerToolHistoryBlockParam = WebSearchServerToolUseBlockParam | WebSearchToolResultBlockParam | ToolSearchServerToolUseBlockParam | ToolSearchToolResultBlockParam;
 /** True when a block is complete Anthropic server-tool history omp can replay. */
@@ -121,7 +135,7 @@ export type FallbackBlockParam = {
         model: string;
     };
 };
-export type ContentBlockParam = TextBlockParam | ImageBlockParam | ToolUseBlockParam | ToolResultBlockParam | ServerToolUseBlockParam | WebSearchToolResultBlockParam | ToolSearchToolResultBlockParam | ThinkingBlockParam | RedactedThinkingBlockParam | FallbackBlockParam;
+export type ContentBlockParam = TextBlockParam | ImageBlockParam | ToolUseBlockParam | ToolResultBlockParam | ServerToolUseBlockParam | WebSearchToolResultBlockParam | ToolSearchToolResultBlockParam | ToolAdditionBlockParam | ToolRemovalBlockParam | ThinkingBlockParam | RedactedThinkingBlockParam | FallbackBlockParam;
 /**
  * A single conversation turn.
  *
@@ -132,6 +146,10 @@ export type ContentBlockParam = TextBlockParam | ImageBlockParam | ToolUseBlockP
 export type MessageParam = {
     role: "user" | "assistant" | "system";
     content: string | ContentBlockParam[];
+    /** Turn-scoped system-message lifetime. */
+    clear_at?: "never" | "next_user_message";
+    /** Per-message effort override. */
+    output_config?: OutputConfig;
 };
 export type ToolInputSchema = {
     type: "object";
@@ -148,6 +166,8 @@ export type Tool = {
     strict?: boolean;
     /** Fine-grained tool streaming beta: stream tool input as it is generated. */
     eager_input_streaming?: boolean;
+    /** Withhold this tool until a later `tool_addition` block references it. */
+    defer_loading?: boolean;
 };
 export type ToolChoiceAuto = {
     type: "auto";
@@ -169,11 +189,16 @@ export type ToolChoice = ToolChoiceAuto | ToolChoiceAny | ToolChoiceTool | ToolC
 export type Metadata = {
     user_id?: string | null;
 };
+export type ThinkingBlockBinding = {
+    prefix_mismatch_behavior: "drop_block" | "error";
+};
 export type ThinkingConfigEnabled = {
     type: "enabled";
     budget_tokens: number;
     /** Opus 4.7+ reasoning display mode. */
     display?: "summarized" | "omitted";
+    /** Preserved-thinking prefix mismatch policy. */
+    block_binding?: ThinkingBlockBinding;
 };
 export type ThinkingConfigDisabled = {
     type: "disabled";
@@ -182,6 +207,8 @@ export type ThinkingConfigAdaptive = {
     type: "adaptive";
     /** Opus 4.7+ reasoning display mode. */
     display?: "summarized" | "omitted";
+    /** Preserved-thinking prefix mismatch policy. */
+    block_binding?: ThinkingBlockBinding;
 };
 export type ThinkingConfigParam = ThinkingConfigEnabled | ThinkingConfigDisabled | ThinkingConfigAdaptive;
 export type OutputConfig = {
@@ -228,6 +255,8 @@ export type MessageCreateParams = {
     speed?: "fast";
     /** Claude Code context-management beta. */
     context_management?: ContextManagement;
+    /** Google Cloud rawPredict carries Anthropic beta names in the body. */
+    anthropic_beta?: string[];
     /**
      * Server-side fallback beta chain — up to three fallback models the API
      * retries when a classifier blocks the primary. Required companion beta
@@ -271,6 +300,14 @@ export type Usage = {
     iterations?: UsageIteration[] | null;
 };
 /** The `message` envelope carried by `message_start`. */
+export type InputTransformation = {
+    type: string;
+    path?: string;
+    reason?: string;
+    [key: string]: unknown;
+};
+/** Parse Anthropic's forward-compatible input transformation list. */
+export declare function parseAnthropicInputTransformations(value: unknown): ProviderInputTransformation[];
 export type ResponseMessage = {
     id: string;
     type?: "message";
@@ -279,6 +316,7 @@ export type ResponseMessage = {
     content?: unknown[];
     stop_reason?: StopReason | null;
     stop_sequence?: string | null;
+    input_transformations?: InputTransformation[];
     usage: Usage;
 };
 /** `content_block` payload carried by `content_block_start`. */
@@ -351,6 +389,7 @@ export type RawMessageDeltaEvent = {
     type: "message_delta";
     delta: MessageDelta;
     usage: Usage;
+    input_transformations?: InputTransformation[];
 };
 export type RawMessageStopEvent = {
     type: "message_stop";