@hebo-ai/gateway 0.6.1 → 0.6.2-rc0

package/README.md

@@ -569,7 +569,7 @@ Accepted request fields:
 
 - `prompt_cache_key` + `prompt_cache_retention` (OpenAI style)
 - `cache_control` (OpenRouter / Vercel / Claude style)
-- `cached_content` (Gemini style)
+- `extra_body { google: { cached_content } }` (Gemini style)
 
 ```json
 {

package/dist/endpoints/chat-completions/converters.js

@@ -6,9 +6,9 @@ import { toResponse } from "../../utils/response";
 import { parseDataUrl } from "../../utils/url";
 // --- Request Flow ---
 export function convertToTextCallOptions(params) {
-    const { messages, tools, tool_choice, temperature, max_tokens, max_completion_tokens, response_format, reasoning_effort, reasoning, prompt_cache_key, prompt_cache_retention, cached_content, cache_control, frequency_penalty, presence_penalty, seed, stop, top_p, ...rest } = params;
+    const { messages, tools, tool_choice, temperature, max_tokens, max_completion_tokens, response_format, reasoning_effort, reasoning, prompt_cache_key, prompt_cache_retention, extra_body, cache_control, frequency_penalty, presence_penalty, seed, stop, top_p, ...rest } = params;
     Object.assign(rest, parseReasoningOptions(reasoning_effort, reasoning));
-    Object.assign(rest, parsePromptCachingOptions(prompt_cache_key, prompt_cache_retention, cached_content, cache_control));
+    Object.assign(rest, parsePromptCachingOptions(prompt_cache_key, prompt_cache_retention, extra_body?.google?.cached_content, cache_control));
     const { toolChoice, activeTools } = convertToToolChoiceOptions(tool_choice);
     return {
         messages: convertToModelMessages(messages),

package/dist/endpoints/chat-completions/schema.d.ts

@@ -658,7 +658,11 @@ declare const ChatCompletionsInputsSchema: z.ZodObject<{
         in_memory: "in_memory";
         "24h": "24h";
     }>>;
-    cached_content: z.ZodOptional<z.ZodString>;
+    extra_body: z.ZodOptional<z.ZodObject<{
+        google: z.ZodOptional<z.ZodObject<{
+            cached_content: z.ZodOptional<z.ZodString>;
+        }, z.core.$strip>>;
+    }, z.core.$strip>>;
     cache_control: z.ZodOptional<z.ZodObject<{
         type: z.ZodLiteral<"ephemeral">;
         ttl: z.ZodOptional<z.ZodString>;
@@ -866,7 +870,11 @@ export declare const ChatCompletionsBodySchema: z.ZodObject<{
         in_memory: "in_memory";
         "24h": "24h";
     }>>;
-    cached_content: z.ZodOptional<z.ZodString>;
+    extra_body: z.ZodOptional<z.ZodObject<{
+        google: z.ZodOptional<z.ZodObject<{
+            cached_content: z.ZodOptional<z.ZodString>;
+        }, z.core.$strip>>;
+    }, z.core.$strip>>;
     cache_control: z.ZodOptional<z.ZodObject<{
         type: z.ZodLiteral<"ephemeral">;
         ttl: z.ZodOptional<z.ZodString>;

package/dist/endpoints/chat-completions/schema.js

@@ -213,7 +213,17 @@ const ChatCompletionsInputsSchema = z.object({
     prompt_cache_key: z.string().optional(),
     prompt_cache_retention: z.enum(["in_memory", "24h"]).optional(),
     // Extension origin: Gemini explicit cache handle
-    cached_content: z.string().optional().meta({ extension: true }),
+    // FUTURE: generalize extra_body handling
+    // https://docs.cloud.google.com/vertex-ai/generative-ai/docs/migrate/openai/overview
+    extra_body: z
+        .object({
+        google: z
+            .object({
+            cached_content: z.string().optional().meta({ extension: true }),
+        })
+            .optional(),
+    })
+        .optional(),
     // Extension origin: OpenRouter/Vercel/Anthropic
     cache_control: ChatCompletionsCacheControlSchema.optional().meta({ extension: true }),
     // Extension origin: OpenRouter

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hebo-ai/gateway",
-  "version": "0.6.1",
+  "version": "0.6.2-rc0",
   "description": "AI gateway as a framework. For full control over models, routing & lifecycle. OpenAI-compatible /chat/completions, /embeddings & /models.",
   "keywords": [
     "ai",
@@ -170,7 +170,7 @@
     "@ai-sdk/google-vertex": "^4.0.80",
     "@ai-sdk/groq": "^3.0.29",
     "@ai-sdk/openai": "^3.0.41",
-    "@aws-sdk/credential-providers": "^3.1002.0",
+    "@aws-sdk/credential-providers": "^3.1003.0",
     "@langfuse/otel": "^4.6.1",
     "@mjackson/node-fetch-server": "^0.7.0",
     "@opentelemetry/api": "^1.9.0",

package/src/endpoints/chat-completions/converters.ts

@@ -87,7 +87,7 @@ export function convertToTextCallOptions(params: ChatCompletionsInputs): TextCal
     reasoning,
     prompt_cache_key,
     prompt_cache_retention,
-    cached_content,
+    extra_body,
     cache_control,
     frequency_penalty,
     presence_penalty,
@@ -103,7 +103,7 @@ export function convertToTextCallOptions(params: ChatCompletionsInputs): TextCal
     parsePromptCachingOptions(
       prompt_cache_key,
       prompt_cache_retention,
-      cached_content,
+      extra_body?.google?.cached_content,
       cache_control,
     ),
   );

package/src/endpoints/chat-completions/schema.ts

@@ -254,7 +254,17 @@ const ChatCompletionsInputsSchema = z.object({
   prompt_cache_key: z.string().optional(),
   prompt_cache_retention: z.enum(["in_memory", "24h"]).optional(),
   // Extension origin: Gemini explicit cache handle
-  cached_content: z.string().optional().meta({ extension: true }),
+  // FUTURE: generalize extra_body handling
+  // https://docs.cloud.google.com/vertex-ai/generative-ai/docs/migrate/openai/overview
+  extra_body: z
+    .object({
+      google: z
+        .object({
+          cached_content: z.string().optional().meta({ extension: true }),
+        })
+        .optional(),
+    })
+    .optional(),
   // Extension origin: OpenRouter/Vercel/Anthropic
   cache_control: ChatCompletionsCacheControlSchema.optional().meta({ extension: true }),
   // Extension origin: OpenRouter