From 4aaba9f2bb13a5a508a0ab9193ade25e961904cf Mon Sep 17 00:00:00 2001
From: suyx <dada20041002@163.com>
Date: Tue, 4 Aug 2026 18:28:44 +0800
Subject: [PATCH] feat(wp7-02): support Gemini image interactions

---
 scripts/lib/wp7-02-controlled-executor.mjs | 42 ++++++++++++++++++++--
 1 file changed, 40 insertions(+), 2 deletions(-)

diff --git a/scripts/lib/wp7-02-controlled-executor.mjs b/scripts/lib/wp7-02-controlled-executor.mjs
index 82d6177..41e1e56 100644
--- a/scripts/lib/wp7-02-controlled-executor.mjs
+++ b/scripts/lib/wp7-02-controlled-executor.mjs
@@ -26,7 +26,12 @@ function assertModelConfig(modelConfig) {
   const profile = modelConfig.route_profile;
   if (!profile || typeof profile !== "object" || typeof profile.endpoint !== "string"
     || !profile.endpoint.startsWith("https://oneapi.intelligrow.cn/")
-    || !["gemini-native-v1beta", "openai-images-v1"].includes(profile.protocol_version)) {
+    || !["gemini-interactions-v1beta", "gemini-native-v1beta", "openai-images-v1"].includes(profile.protocol_version)) {
+    throw new Error("WP7_02_ROUTE_PROFILE_INVALID");
+  }
+  if (profile.protocol_version === "gemini-interactions-v1beta"
+    && (profile.endpoint !== "https://oneapi.intelligrow.cn/v1beta/interactions"
+      || profile.provider_model_id !== "gemini-3.1-flash-image")) {
     throw new Error("WP7_02_ROUTE_PROFILE_INVALID");
   }
   if (profile.protocol_version === "openai-images-v1"
@@ -69,6 +74,20 @@ export function buildProviderRequest({ modelConfig, prompt, ratio, reference })
     throw new Error("WP7_02_REFERENCE_FIXTURE_INVALID");
   }
   const headers = { "content-type": "application/json" };
+  if (config.route_profile.protocol_version === "gemini-interactions-v1beta") {
+    const input = [{ text: prompt, type: "text" }];
+    if (reference) input.push({ data: reference.bytes.toString("base64"), mime_type: reference.mime_type, type: "image" });
+    return {
+      body: {
+        input,
+        model: config.route_profile.provider_model_id,
+        response_format: { aspect_ratio: ratio, image_size: "1K", type: "image" },
+      },
+      headers,
+      method: "POST",
+      url: config.route_profile.endpoint,
+    };
+  }
   if (config.route_profile.protocol_version === "gemini-native-v1beta") {
     const parts = [{ text: prompt }];
     if (reference) parts.push({ inlineData: { data: reference.bytes.toString("base64"), mimeType: reference.mime_type } });
@@ -175,12 +194,31 @@ function openAiUsage(response) {
   };
 }
 
+function interactionUsage(response) {
+  const usage = response?.usage;
+  return {
+    input_units: integerOrZero(usage?.total_input_tokens),
+    output_units: integerOrZero(usage?.total_output_tokens),
+    total_units: integerOrZero(usage?.total_tokens),
+  };
+}
+
 export function normalizeProviderResponse(modelConfig, response) {
   const config = assertModelConfig(modelConfig);
   let bytes;
   let mime;
   let usageSummary;
-  if (config.route_profile.protocol_version === "gemini-native-v1beta") {
+  if (config.route_profile.protocol_version === "gemini-interactions-v1beta") {
+    const stepImages = response?.steps?.flatMap((step) => step?.type === "model_output" ? step?.content ?? [] : [])
+      .filter((content) => content?.type === "image" && content?.data) ?? [];
+    const images = stepImages.length > 0
+      ? stepImages
+      : [response?.output_image].filter((content) => content?.data);
+    if (images.length !== 1) throw new Error("WP7_02_RESPONSE_SINGLE_IMAGE_REQUIRED");
+    mime = images[0].mime_type ?? images[0].mimeType;
+    bytes = Buffer.from(images[0].data, "base64");
+    usageSummary = interactionUsage(response);
+  } else if (config.route_profile.protocol_version === "gemini-native-v1beta") {
     const parts = response?.candidates?.flatMap((candidate) => candidate?.content?.parts ?? []) ?? [];
     const images = parts.map((part) => part?.inlineData ?? part?.inline_data).filter((entry) => entry?.data);
     if (images.length !== 1) throw new Error("WP7_02_RESPONSE_SINGLE_IMAGE_REQUIRED");