@kenkaiiii/gg-ai 4.4.0 → 4.6.0

package/dist/index.d.cts

@@ -23,6 +23,11 @@ interface VideoContent {
     type: "video";
     mediaType: string;
     data: string;
+    /** Moonshot/Kimi file id (e.g. "d4f0…") after uploading via the files API.
+     *  Moonshot rejects inline base64 video; the provider uploads the clip once
+     *  and caches the id here so later turns reference `ms://<fileId>` instead of
+     *  re-sending the bytes. */
+    fileId?: string;
 }
 interface ToolCall {
     type: "tool_call";
@@ -30,7 +35,7 @@ interface ToolCall {
     name: string;
     args: Record<string, unknown>;
 }
-type ToolResultContent = string | (TextContent | ImageContent)[];
+type ToolResultContent = string | (TextContent | ImageContent | VideoContent)[];
 interface ToolResult {
     type: "tool_result";
     toolCallId: string;
@@ -208,6 +213,11 @@ interface StreamOptions {
      *  version should pass it here. Ignored for non-Anthropic providers and for
      *  Anthropic requests using a regular API key. */
     userAgent?: string;
+    /** Extra HTTP headers attached to every model request. Used by providers
+     *  whose endpoint gates on client identity (e.g. Kimi For Coding requires a
+     *  `User-Agent: kimi-code-cli/...` and `X-Msh-*` device headers). Merged
+     *  into the underlying SDK's default headers. */
+    defaultHeaders?: Record<string, string>;
 }
 
 /**
@@ -330,7 +340,7 @@ declare const providerRegistry: ProviderRegistryImpl;
  *     Cannot read property 'foo' of undefined
  *     → This is a ggcoder bug — please report it.
  */
-type ErrorSource = "provider" | "ggcoder" | "network" | "auth";
+type ErrorSource = "provider" | "ggcoder" | "network" | "auth" | "capability";
 interface FormattedError {
     /** Plain-English headline, e.g. "OpenAI returned an error." */
     headline: string;

package/dist/index.d.ts

@@ -23,6 +23,11 @@ interface VideoContent {
     type: "video";
     mediaType: string;
     data: string;
+    /** Moonshot/Kimi file id (e.g. "d4f0…") after uploading via the files API.
+     *  Moonshot rejects inline base64 video; the provider uploads the clip once
+     *  and caches the id here so later turns reference `ms://<fileId>` instead of
+     *  re-sending the bytes. */
+    fileId?: string;
 }
 interface ToolCall {
     type: "tool_call";
@@ -30,7 +35,7 @@ interface ToolCall {
     name: string;
     args: Record<string, unknown>;
 }
-type ToolResultContent = string | (TextContent | ImageContent)[];
+type ToolResultContent = string | (TextContent | ImageContent | VideoContent)[];
 interface ToolResult {
     type: "tool_result";
     toolCallId: string;
@@ -208,6 +213,11 @@ interface StreamOptions {
      *  version should pass it here. Ignored for non-Anthropic providers and for
      *  Anthropic requests using a regular API key. */
     userAgent?: string;
+    /** Extra HTTP headers attached to every model request. Used by providers
+     *  whose endpoint gates on client identity (e.g. Kimi For Coding requires a
+     *  `User-Agent: kimi-code-cli/...` and `X-Msh-*` device headers). Merged
+     *  into the underlying SDK's default headers. */
+    defaultHeaders?: Record<string, string>;
 }
 
 /**
@@ -330,7 +340,7 @@ declare const providerRegistry: ProviderRegistryImpl;
  *     Cannot read property 'foo' of undefined
  *     → This is a ggcoder bug — please report it.
  */
-type ErrorSource = "provider" | "ggcoder" | "network" | "auth";
+type ErrorSource = "provider" | "ggcoder" | "network" | "auth" | "capability";
 interface FormattedError {
     /** Plain-English headline, e.g. "OpenAI returned an error." */
     headline: string;

package/dist/index.js

@@ -25,6 +25,12 @@ var GGAIError = class extends Error {
     this.hint = options?.hint;
   }
 };
+var VideoUnsupportedError = class extends GGAIError {
+  constructor() {
+    super("This model can't analyze video.", { source: "capability" });
+    this.name = "VideoUnsupportedError";
+  }
+};
 var ProviderError = class extends GGAIError {
   provider;
   statusCode;
@@ -140,6 +146,14 @@ function finaliseBySource(source, message, requestId, hint) {
         guidance: hint ?? providerGuidance(void 0, message, void 0),
         ...requestId ? { requestId } : {}
       };
+    case "capability":
+      return {
+        headline: message,
+        source,
+        message: "",
+        guidance: hint ?? "Only Kimi, Gemini, MiniMax, and MiMo-V2.5 can analyze video. Switch with /model.",
+        ...requestId ? { requestId } : {}
+      };
     case "ggcoder":
       return {
         headline: "ggcoder hit an unexpected error.",
@@ -500,6 +514,10 @@ function toolResultImages(content) {
   if (typeof content === "string") return [];
   return content.filter((b) => b.type === "image");
 }
+function toolResultVideos(content) {
+  if (typeof content === "string") return [];
+  return content.filter((b) => b.type === "video");
+}
 function toAnthropicCacheControl(retention, baseUrl) {
   const resolved = retention ?? "short";
   if (resolved === "none") return void 0;
@@ -510,6 +528,12 @@ function toAnthropicToolResultContent(content) {
   if (typeof content === "string") return content;
   return content.map((block) => {
     if (block.type === "text") return { type: "text", text: block.text };
+    if (block.type === "video") {
+      return {
+        type: "video",
+        source: { type: "base64", media_type: block.mediaType, data: block.data }
+      };
+    }
     return {
       type: "image",
       source: {
@@ -579,6 +603,8 @@ function toAnthropicMessages(messages, cacheControl) {
     if (msg.role === "tool") {
       out.push({
         role: "user",
+        // Cast covers the video block (used by the Anthropic-compatible MiniMax
+        // API), which isn't in the first-party Anthropic tool_result types.
         content: msg.content.map((result) => ({
           type: "tool_result",
           tool_use_id: remapAnthropicToolCallId(result.toolCallId, idMap),
@@ -719,11 +745,10 @@ function toOpenAIMessages(messages, options) {
             (part) => {
               if (part.type === "text") return { type: "text", text: part.text };
               if (part.type === "video") {
+                const videoUrl = part.fileId ? { url: `ms://${part.fileId}`, id: part.fileId } : { url: `data:${part.mediaType};base64,${part.data}` };
                 return {
                   type: "video_url",
-                  video_url: {
-                    url: `data:${part.mediaType};base64,${part.data}`
-                  }
+                  video_url: videoUrl
                 };
               }
               return {
@@ -768,29 +793,56 @@ function toOpenAIMessages(messages, options) {
       continue;
     }
     if (msg.role === "tool") {
-      const imageBlocks = [];
+      const isMoonshot = options?.provider === "moonshot";
+      const followUpMediaBlocks = [];
+      let followUpHasVideo = false;
       for (const result of msg.content) {
         const text = toolResultText(result.content);
         const images = toolResultImages(result.content);
+        const videos = toolResultVideos(result.content);
         const hasText = text.length > 0;
+        if (isMoonshot && videos.length > 0) {
+          const parts = [];
+          if (hasText) parts.push({ type: "text", text });
+          const videoParts = videos.map((v) => {
+            const videoUrl = v.fileId ? { url: `ms://${v.fileId}`, id: v.fileId } : { url: `data:${v.mediaType};base64,${v.data}` };
+            return { type: "video_url", video_url: videoUrl };
+          });
+          out.push({
+            role: "tool",
+            tool_call_id: remapToolCallId(result.toolCallId, idMap),
+            content: [...parts, ...videoParts]
+          });
+          continue;
+        }
         out.push({
           role: "tool",
           tool_call_id: remapToolCallId(result.toolCallId, idMap),
-          content: hasText ? text : "(see attached image)"
+          content: hasText ? text : "(see attached media)"
         });
         if (images.length > 0 && options?.supportsImages !== false) {
           for (const img of images) {
-            imageBlocks.push({
+            followUpMediaBlocks.push({
               type: "image_url",
               image_url: { url: `data:${img.mediaType};base64,${img.data}` }
             });
           }
         }
+        if (!isMoonshot && videos.length > 0) {
+          for (const v of videos) {
+            followUpMediaBlocks.push({
+              type: "video_url",
+              video_url: { url: `data:${v.mediaType};base64,${v.data}` }
+            });
+            followUpHasVideo = true;
+          }
+        }
       }
-      if (imageBlocks.length > 0) {
+      if (followUpMediaBlocks.length > 0) {
+        const label = followUpHasVideo ? "Attached media from tool result:" : "Attached image(s) from tool result:";
         out.push({
           role: "user",
-          content: [{ type: "text", text: "Attached image(s) from tool result:" }, ...imageBlocks]
+          content: [{ type: "text", text: label }, ...followUpMediaBlocks]
         });
       }
     }
@@ -1357,6 +1409,75 @@ function fnv1aHash(value) {
   return (hash >>> 0).toString(16).padStart(8, "0");
 }
 
+// src/utils/diag.ts
+var _diagFn = null;
+function setProviderDiagnostic(fn) {
+  _diagFn = fn;
+}
+function providerDiag(phase, data) {
+  _diagFn?.(phase, data);
+}
+
+// src/providers/moonshot-video.ts
+async function uploadMoonshotVideos(client, messages, signal) {
+  for (const msg of messages) {
+    if (typeof msg.content === "string") continue;
+    for (const part of msg.content) {
+      if (part.type === "video") {
+        await ensureUploaded(client, part, signal);
+        continue;
+      }
+      if (part.type === "tool_result" && Array.isArray(part.content)) {
+        for (const inner of part.content) {
+          if (inner.type === "video") {
+            await ensureUploaded(client, inner, signal);
+          }
+        }
+      }
+    }
+  }
+}
+async function ensureUploaded(client, video, signal) {
+  if (video.fileId) {
+    providerDiag("moonshot_video_cached", { fileId: video.fileId });
+    return;
+  }
+  if (!video.data) {
+    providerDiag("moonshot_video_skipped_no_data", {});
+    return;
+  }
+  providerDiag("moonshot_video_upload_start", {
+    mediaType: video.mediaType,
+    bytes: Math.floor(video.data.length * 3 / 4)
+  });
+  video.fileId = await uploadOne(client, video, signal);
+  providerDiag("moonshot_video_upload_done", { fileId: video.fileId });
+}
+async function uploadOne(client, video, signal) {
+  const bytes = Buffer.from(video.data, "base64");
+  const mediaType = video.mediaType || "video/mp4";
+  const filename = `upload.${extForMime(mediaType)}`;
+  const file = new File([new Uint8Array(bytes)], filename, { type: mediaType });
+  const uploaded = await client.files.create(
+    { file, purpose: "video" },
+    signal ? { signal } : void 0
+  );
+  return uploaded.id;
+}
+var MIME_TO_EXT = {
+  "video/mp4": "mp4",
+  "video/mpeg": "mpeg",
+  "video/quicktime": "mov",
+  "video/webm": "webm",
+  "video/x-matroska": "mkv",
+  "video/x-msvideo": "avi",
+  "video/x-flv": "flv",
+  "video/3gpp": "3gp"
+};
+function extForMime(mediaType) {
+  return MIME_TO_EXT[mediaType.toLowerCase()] ?? "mp4";
+}
+
 // src/utils/env.ts
 function getEnvironment() {
   return globalThis.process?.env;
@@ -1386,7 +1507,8 @@ function createClient2(options) {
   return new OpenAI({
     apiKey: options.apiKey,
     ...options.baseUrl ? { baseURL: options.baseUrl } : {},
-    ...options.fetch ? { fetch: options.fetch } : {}
+    ...options.fetch ? { fetch: options.fetch } : {},
+    ...options.defaultHeaders ? { defaultHeaders: options.defaultHeaders } : {}
   });
 }
 function streamOpenAI(options) {
@@ -1399,6 +1521,13 @@ async function* runStream2(options) {
   const usesThinkingParam = options.provider === "glm" || options.provider === "moonshot" || options.provider === "xiaomi";
   const downgradedImages = downgradeUnsupportedImages(options.messages, options.supportsImages);
   const downgradedMessages = downgradeUnsupportedVideos(downgradedImages, options.supportsVideo);
+  if (options.provider === "moonshot") {
+    try {
+      await uploadMoonshotVideos(client, downgradedMessages, options.signal);
+    } catch (err) {
+      throw toError2(err, providerName);
+    }
+  }
   const messages = toOpenAIMessages(downgradedMessages, {
     provider: options.provider,
     thinking: !!options.thinking,
@@ -1702,15 +1831,6 @@ function toError2(err, provider = "openai") {
 // src/providers/openai-codex.ts
 import os from "os";
 
-// src/utils/diag.ts
-var _diagFn = null;
-function setProviderDiagnostic(fn) {
-  _diagFn = fn;
-}
-function providerDiag(phase, data) {
-  _diagFn?.(phase, data);
-}
-
 // src/utils/sse.ts
 function parseSseBuffer(buffer) {
   const events = [];
@@ -2359,6 +2479,13 @@ ${msg.content}` : msg.content;
             }
           }
         });
+        if (typeof result.content !== "string") {
+          for (const block of result.content) {
+            if (block.type === "video") {
+              parts.push({ inlineData: { mimeType: block.mediaType, data: block.data } });
+            }
+          }
+        }
       }
       if (parts.length > 0) contents.push({ role: "user", parts });
     }
@@ -2756,6 +2883,7 @@ var providerRegistry = new ProviderRegistryImpl();
 
 // src/stream.ts
 var GLM_CODING_BASE_URL = "https://api.z.ai/api/coding/paas/v4";
+var KIMI_CODE_USER_AGENT = `kimi-code-cli/${process.env.KIMI_CODE_VERSION ?? "1.0.11"}`;
 providerRegistry.register("anthropic", {
   stream: (options) => streamAnthropic(options)
 });
@@ -2784,10 +2912,11 @@ providerRegistry.register("glm", {
   })
 });
 providerRegistry.register("moonshot", {
-  stream: (options) => streamOpenAI({
-    ...options,
-    baseUrl: options.baseUrl ?? "https://api.moonshot.ai/v1"
-  })
+  stream: (options) => {
+    const baseUrl = options.baseUrl ?? "https://api.moonshot.ai/v1";
+    const defaultHeaders = baseUrl.includes("api.kimi.com") ? { "User-Agent": KIMI_CODE_USER_AGENT, ...options.defaultHeaders } : options.defaultHeaders;
+    return streamOpenAI({ ...options, baseUrl, defaultHeaders });
+  }
 });
 providerRegistry.register("deepseek", {
   stream: (options) => streamOpenAI({
@@ -2820,8 +2949,23 @@ function stream(options) {
       `Unknown provider: "${options.provider}". Registered: ${providerRegistry.list().join(", ")}`
     );
   }
+  if (options.supportsVideo !== true && messagesContainVideo(options.messages)) {
+    throw new VideoUnsupportedError();
+  }
   return entry.stream(options);
 }
+function messagesContainVideo(messages) {
+  for (const msg of messages) {
+    if (typeof msg.content === "string" || !Array.isArray(msg.content)) continue;
+    for (const part of msg.content) {
+      if (part.type === "video") return true;
+      if (part.type === "tool_result" && Array.isArray(part.content)) {
+        if (part.content.some((block) => block.type === "video")) return true;
+      }
+    }
+  }
+  return false;
+}
 
 // src/error-classification.ts
 var CONTEXT_OVERFLOW_PATTERNS = [