perf(ai): build request bodies from the previous request's bytes

packages/ai/src/api/anthropic-messages.ts

@@ -40,11 +40,11 @@ import type {
 import { appendAssistantMessageDiagnostic } from "../utils/diagnostics.ts";
 import { AssistantMessageEventStream } from "../utils/event-stream.ts";
 import { headersToRecord } from "../utils/headers.ts";
-import { cacheJsonSerialization } from "../utils/json-body.ts";
 import { parseJsonWithRepair, parseStreamingJsonWhileStreaming, parseToolCallArguments } from "../utils/json-parse.ts";
 import { getPiUserAgent } from "../utils/pi-user-agent.ts";
 import { getProviderEnvValue } from "../utils/provider-env.ts";
 import { retryProviderRequest } from "../utils/provider-retry.ts";
+import { cacheRequestBody } from "../utils/request-body.ts";
 import { sanitizeSurrogates } from "../utils/sanitize-unicode.ts";
 import { iterateSseMessages } from "../utils/sse.ts";
 import { getSystemMessageText, renderSystemMessageUpdate } from "../utils/text.ts";
@@ -514,7 +514,8 @@ export const stream: StreamFunction<"anthropic-messages", AnthropicOptions> = (
 				if (nextParams !== undefined) {
 					params = { ...(nextParams as MessageCreateParamsStreaming), stream: true };
 				}
-				cacheJsonSerialization(params.messages);
+				// A client passed in by the caller is not wrapped.
+				cacheRequestBody(options?.client ? undefined : client, params.messages, "messages", options?.sessionId);
 				try {
 					response = await retryProviderRequest(
 						() => client.beta.messages.create(params, requestOptions).asResponse(),

packages/ai/src/api/azure-openai-responses.ts

@@ -5,9 +5,9 @@ import type { Api, AssistantMessage, Model, SimpleStreamOptions, StreamFunction,
 import { formatProviderError, normalizeProviderError } from "../utils/error-body.ts";
 import { AssistantMessageEventStream } from "../utils/event-stream.ts";
 import { headersToRecord } from "../utils/headers.ts";
-import { cacheJsonSerialization } from "../utils/json-body.ts";
 import { getPiUserAgent } from "../utils/pi-user-agent.ts";
 import { retryProviderRequest } from "../utils/provider-retry.ts";
+import { cacheRequestBody } from "../utils/request-body.ts";
 import { getDeclaredTools, resolveTranscript, resolveTranscriptTools } from "../utils/transcript.ts";
 import { type AzureEndpointOptions, resolveAzureConfig, resolveDeploymentName } from "./azure-openai-config.ts";
 import { createGrammarToolInputProperties } from "./constrained-sampling.ts";
@@ -80,7 +80,7 @@ export const stream: StreamFunction<"azure-openai-responses", AzureOpenAIRespons
 			if (nextParams !== undefined) {
 				params = nextParams as ResponseCreateParamsStreaming;
 			}
-			cacheJsonSerialization(params.input);
+			cacheRequestBody(client, params.input, "input", options?.sessionId);
 			const requestOptions = {
 				...(options?.signal ? { signal: options.signal } : {}),
 				...(options?.timeoutMs !== undefined ? { timeout: options.timeoutMs } : {}),

packages/ai/src/api/openai-completions.ts

@@ -39,11 +39,11 @@ import { formatProviderError, normalizeProviderError } from "../utils/error-body
 import { AssistantMessageEventStream } from "../utils/event-stream.ts";
 import { shortHash } from "../utils/hash.ts";
 import { headersToRecord } from "../utils/headers.ts";
-import { cacheJsonSerialization } from "../utils/json-body.ts";
 import { parseStreamingJsonWhileStreaming, parseToolCallArguments } from "../utils/json-parse.ts";
 import { getPiUserAgent } from "../utils/pi-user-agent.ts";
 import { getProviderEnvValue } from "../utils/provider-env.ts";
 import { retryProviderRequest } from "../utils/provider-retry.ts";
+import { cacheRequestBody } from "../utils/request-body.ts";
 import { sanitizeSurrogates } from "../utils/sanitize-unicode.ts";
 import { getSystemMessageText, renderSystemMessageUpdate } from "../utils/text.ts";
 import {
@@ -370,7 +370,7 @@ export const stream: StreamFunction<"openai-completions", OpenAICompletionsOptio
 			if (nextParams !== undefined) {
 				params = nextParams as OpenAI.Chat.Completions.ChatCompletionCreateParamsStreaming;
 			}
-			cacheJsonSerialization(params.messages);
+			cacheRequestBody(client, params.messages, "messages", options?.sessionId);
 			const requestOptions = {
 				...(options?.signal ? { signal: options.signal } : {}),
 				...(options?.timeoutMs !== undefined ? { timeout: options.timeoutMs } : {}),

packages/ai/src/api/openai-responses.ts

@@ -18,10 +18,10 @@ import type {
 import { formatProviderError, normalizeProviderError } from "../utils/error-body.ts";
 import { AssistantMessageEventStream } from "../utils/event-stream.ts";
 import { headersToRecord } from "../utils/headers.ts";
-import { cacheJsonSerialization } from "../utils/json-body.ts";
 import { getPiUserAgent } from "../utils/pi-user-agent.ts";
 import { getProviderEnvValue } from "../utils/provider-env.ts";
 import { retryProviderRequest } from "../utils/provider-retry.ts";
+import { cacheRequestBody } from "../utils/request-body.ts";
 import { getDeclaredTools, resolveTranscript, resolveTranscriptTools } from "../utils/transcript.ts";
 import { createGrammarToolInputProperties } from "./constrained-sampling.ts";
 import { buildCopilotDynamicHeaders, hasCopilotVisionInput } from "./github-copilot-headers.ts";
@@ -206,7 +206,7 @@ export const stream: StreamFunction<"openai-responses", OpenAIResponsesOptions>
 			if (nextParams !== undefined) {
 				params = nextParams as ResponseCreateParamsStreaming;
 			}
-			cacheJsonSerialization(params.input);
+			cacheRequestBody(client, params.input, "input", options?.sessionId);
 			const requestOptions = {
 				...(options?.signal ? { signal: options.signal } : {}),
 				...(options?.timeoutMs !== undefined ? { timeout: options.timeoutMs } : {}),

packages/ai/src/utils/json-body.ts

@@ -99,15 +99,29 @@ function substituteItems(
 	return out;
 }
 
+/**
+ * The toJSON cacheJsonSerialization gives an array: a copy of it with long strings replaced by their cached raw JSON,
+ * which JSON.stringify serializes to the same text as the array.
+ */
+export function cachedArrayToJSON(this: unknown[]): unknown {
+	try {
+		if (!rawJSON) return this;
+		startGeneration();
+		return substituteItems(this, rawJSON, new Set());
+	} catch {
+		// JSON.stringify serializes what toJSON returns without calling it again: the array as it is.
+		return this;
+	}
+}
+
 /**
  * Makes JSON.stringify of `items` (a request's messages) reuse the serialized form of long strings serialized before,
  * by defining a non-enumerable toJSON on the array. Does nothing to anything but an extensible plain array without a
  * toJSON (an onPayload hook may return a frozen one). Never throws.
  */
 export function cacheJsonSerialization(items: unknown): void {
-	const toRaw = rawJSON;
 	if (
-		!toRaw ||
+		!rawJSON ||
 		!isRawJSON ||
 		!Array.isArray(items) ||
 		Object.getPrototypeOf(items) !== Array.prototype ||
@@ -118,15 +132,7 @@ export function cacheJsonSerialization(items: unknown): void {
 	}
 	try {
 		Object.defineProperty(items, "toJSON", {
-			value(this: unknown[]): unknown {
-				try {
-					startGeneration();
-					return substituteItems(this, toRaw, new Set());
-				} catch {
-					// JSON.stringify serializes what toJSON returns without calling it again: the array as it is.
-					return this;
-				}
-			},
+			value: cachedArrayToJSON,
 			configurable: true,
 			writable: true,
 			enumerable: false,