chore: generate

packages/llm/README.md

@@ -82,12 +82,12 @@ LLM.request({
 
 ### Provider behavior table
 
-| Protocol | `cache: "auto"` |
+| Protocol                | `cache: "auto"`                                                           |
-|---|---|
+| ----------------------- | ------------------------------------------------------------------------- |
-| Anthropic Messages | emits up to 3 `cache_control` markers (4-breakpoint cap enforced) |
+| Anthropic Messages      | emits up to 3 `cache_control` markers (4-breakpoint cap enforced)         |
-| Bedrock Converse | emits up to 3 `cachePoint` blocks (4-breakpoint cap enforced) |
+| Bedrock Converse        | emits up to 3 `cachePoint` blocks (4-breakpoint cap enforced)             |
-| OpenAI Chat / Responses | no-op (implicit caching above 1024 tokens) |
+| OpenAI Chat / Responses | no-op (implicit caching above 1024 tokens)                                |
-| Gemini | no-op (implicit caching on 2.5+; explicit `CachedContent` is out-of-band) |
+| Gemini                  | no-op (implicit caching on 2.5+; explicit `CachedContent` is out-of-band) |
 
 Normalized cache usage is read back into `response.usage.cacheReadInputTokens` and `cacheWriteInputTokens` across every provider.
 

packages/llm/src/cache-policy.ts

@@ -44,10 +44,7 @@ const RESPECTS_INLINE_HINTS = new Set(["anthropic-messages", "bedrock-converse"]
 const makeHint = (ttlSeconds: number | undefined): CacheHint =>
   ttlSeconds !== undefined ? new CacheHint({ type: "ephemeral", ttlSeconds }) : new CacheHint({ type: "ephemeral" })
 
-const markLastTool = (
+const markLastTool = (tools: ReadonlyArray<ToolDefinition>, hint: CacheHint): ReadonlyArray<ToolDefinition> => {
-  tools: ReadonlyArray<ToolDefinition>,
-  hint: CacheHint,
-): ReadonlyArray<ToolDefinition> => {
   if (tools.length === 0) return tools
   const last = tools.length - 1
   if (tools[last]!.cache) return tools
@@ -67,11 +64,7 @@ const lastIndexOfRole = (messages: ReadonlyArray<Message>, role: Message["role"]
 // Mark the last text part of `messages[index]`. If no text part exists, mark
 // the last content part regardless of type — that's the breakpoint position
 // in tool-result-only messages too.
-const markMessageAt = (
+const markMessageAt = (messages: ReadonlyArray<Message>, index: number, hint: CacheHint): ReadonlyArray<Message> => {
-  messages: ReadonlyArray<Message>,
-  index: number,
-  hint: CacheHint,
-): ReadonlyArray<Message> => {
   if (index < 0 || index >= messages.length) return messages
   const target = messages[index]!
   if (target.content.length === 0) return messages
@@ -79,9 +72,7 @@ const markMessageAt = (
   const markAt = lastTextIndex >= 0 ? lastTextIndex : target.content.length - 1
   const existing = target.content[markAt]!
   if ("cache" in existing && existing.cache) return messages
-  const nextContent = target.content.map((part, i) =>
+  const nextContent = target.content.map((part, i) => (i === markAt ? ({ ...part, cache: hint } as ContentPart) : part))
-    i === markAt ? ({ ...part, cache: hint } as ContentPart) : part,
-  )
   const next = new Message({ ...target, content: nextContent })
   // Single pass over `messages`, substituting the one updated entry. Long
   // conversations call this on every request, so avoid `.map()` here — its

packages/llm/src/schema/options.ts

@@ -226,9 +226,5 @@ export const CachePolicyObject = Schema.Struct({
 })
 export type CachePolicyObject = Schema.Schema.Type<typeof CachePolicyObject>
 
-export const CachePolicy = Schema.Union([
+export const CachePolicy = Schema.Union([Schema.Literal("auto"), Schema.Literal("none"), CachePolicyObject])
-  Schema.Literal("auto"),
-  Schema.Literal("none"),
-  CachePolicyObject,
-])
 export type CachePolicy = Schema.Schema.Type<typeof CachePolicy>

packages/llm/test/cache-policy.test.ts

@@ -56,11 +56,7 @@ describe("applyCachePolicy", () => {
           model: anthropicModel,
           system: "Sys A",
           tools: [{ name: "t1", description: "t1", inputSchema: { type: "object", properties: {} } }],
-          messages: [
+          messages: [LLM.user("first user"), LLM.assistant("assistant reply"), LLM.user("latest user message")],
-            LLM.user("first user"),
-            LLM.assistant("assistant reply"),
-            LLM.user("latest user message"),
-          ],
           cache: "auto",
         }),
       )

packages/llm/test/fixtures/recordings/anthropic-messages-cache/writes-then-reads-cache-control-on-identical-second-call.json

@@ -3,12 +3,7 @@
   "metadata": {
     "name": "anthropic-messages-cache/writes-then-reads-cache-control-on-identical-second-call",
     "recordedAt": "2026-05-11T01:52:54.319Z",
-    "tags": [
+    "tags": ["prefix:anthropic-messages-cache", "provider:anthropic", "protocol:anthropic-messages", "cache"]
-      "prefix:anthropic-messages-cache",
-      "provider:anthropic",
-      "protocol:anthropic-messages",
-      "cache"
-    ]
   },
   "interactions": [
     {

packages/llm/test/fixtures/recordings/gemini-cache/reports-cachedcontenttokencount-on-identical-second-call.json

@@ -3,12 +3,7 @@
   "metadata": {
     "name": "gemini-cache/reports-cachedcontenttokencount-on-identical-second-call",
     "recordedAt": "2026-05-11T01:55:40.600Z",
-    "tags": [
+    "tags": ["prefix:gemini-cache", "provider:google", "protocol:gemini", "cache"]
-      "prefix:gemini-cache",
-      "provider:google",
-      "protocol:gemini",
-      "cache"
-    ]
   },
   "interactions": [
     {

packages/llm/test/fixtures/recordings/openai-responses-cache/reports-cached-tokens-on-identical-second-call.json

@@ -3,12 +3,7 @@
   "metadata": {
     "name": "openai-responses-cache/reports-cached-tokens-on-identical-second-call",
     "recordedAt": "2026-05-11T01:41:58.951Z",
-    "tags": [
+    "tags": ["prefix:openai-responses-cache", "provider:openai", "protocol:openai-responses", "cache"]
-      "prefix:openai-responses-cache",
-      "provider:openai",
-      "protocol:openai-responses",
-      "cache"
-    ]
   },
   "interactions": [
     {