fix(ai): clamp prompt_cache_key to 64 chars for all chat protocols (#44571)

packages/ai/src/protocols/open-responses.ts

@@ -633,6 +633,7 @@ const lowerMessages = Effect.fn("OpenResponses.lowerMessages")(function* (reques
 
 const lowerOptions = (request: LLMRequest) => {
   const options = OpenResponsesOptions.resolve(request)
+  const cacheKey = ProviderShared.clampPromptCacheKey(request.promptCacheKey)
   return {
     ...(options.instructions ? { instructions: options.instructions } : {}),
     ...(options.store !== undefined ? { store: options.store } : {}),
@@ -642,7 +643,7 @@ const lowerOptions = (request: LLMRequest) => {
       ? { stream_options: { include_obfuscation: options.streamOptions.includeObfuscation } }
       : {}),
     ...(options.topLogprobs !== undefined ? { top_logprobs: options.topLogprobs } : {}),
-    ...(request.promptCacheKey ? { prompt_cache_key: request.promptCacheKey } : {}),
+    ...(cacheKey ? { prompt_cache_key: cacheKey } : {}),
     ...(options.include ? { include: options.include } : {}),
     ...(options.reasoningEffort || options.reasoningSummary
       ? { reasoning: { effort: options.reasoningEffort, summary: options.reasoningSummary } }

packages/ai/src/protocols/openai-chat.ts

@@ -530,9 +530,10 @@ const hasToolHistory = (messages: ReadonlyArray<LLMRequest["messages"][number]>)
 
 const lowerOptions = (request: LLMRequest) => {
   const options = OpenAIOptions.resolve(request)
+  const cacheKey = ProviderShared.clampPromptCacheKey(request.promptCacheKey)
   return {
     ...(options.store !== undefined ? { store: options.store } : {}),
-    ...(request.promptCacheKey ? { prompt_cache_key: request.promptCacheKey } : {}),
+    ...(cacheKey ? { prompt_cache_key: cacheKey } : {}),
     ...(options.reasoningEffort ? { reasoning_effort: options.reasoningEffort } : {}),
   }
 }

packages/ai/src/protocols/shared.ts

@@ -24,6 +24,17 @@ export const JsonObject = Schema.Record(Schema.String, Schema.Unknown)
 export const optionalArray = <const S extends Schema.Top>(schema: S) => Schema.optional(Schema.Array(schema))
 export const optionalNull = <const S extends Schema.Top>(schema: S) => Schema.optional(Schema.NullOr(schema))
 
+export const OPENAI_PROMPT_CACHE_KEY_MAX_LENGTH = 64
+
+// OpenAI limits `prompt_cache_key` to 64 chars; DeepSeek and Zai inherit the same
+// limit via their OpenAI-compatible APIs. Clamp with unicode-aware slicing.
+export const clampPromptCacheKey = (key: string | undefined): string | undefined => {
+  if (key === undefined) return undefined
+  const chars = Array.from(key)
+  if (chars.length <= OPENAI_PROMPT_CACHE_KEY_MAX_LENGTH) return key
+  return chars.slice(0, OPENAI_PROMPT_CACHE_KEY_MAX_LENGTH).join("")
+}
+
 /**
  * Streaming tool-call accumulator. Adapters that build a tool call across
  * multiple `tool-input-delta` chunks store the partial JSON input string here

packages/ai/src/providers/openrouter.ts

@@ -9,7 +9,7 @@ import type { ProviderPackage } from "../provider-package.js"
 import * as OpenAICompatibleProfiles from "./openai-compatible-profile.js"
 import * as OpenAIChat from "../protocols/openai-chat.js"
 import { newBreakpoints, ttlBucket } from "../protocols/utils/cache.js"
-import { isRecord } from "../protocols/shared.js"
+import { isRecord, ProviderShared } from "../protocols/shared.js"
 
 export const profile = OpenAICompatibleProfiles.profiles.openrouter
 export const id = ProviderID.make(profile.provider)
@@ -115,11 +115,12 @@ export const protocol = Protocol.make({
               reasoning_details: reasoningDetails,
             }
           })
+          const cacheKey = ProviderShared.clampPromptCacheKey(request.promptCacheKey)
           return {
             ...body,
             messages,
             ...bodyOptions(request.providerOptions),
-            ...(request.promptCacheKey ? { prompt_cache_key: request.promptCacheKey } : {}),
+            ...(cacheKey ? { prompt_cache_key: cacheKey } : {}),
           } as OpenRouterBody
         }),
       ),