@kenkaiiii/gg-ai 4.3.38 → 4.3.40

package/dist/index.d.cts

@@ -23,10 +23,11 @@ interface ToolCall {
     name: string;
     args: Record<string, unknown>;
 }
+type ToolResultContent = string | (TextContent | ImageContent)[];
 interface ToolResult {
     type: "tool_result";
     toolCallId: string;
-    content: string;
+    content: ToolResultContent;
     isError?: boolean;
 }
 interface ServerToolCall {
@@ -174,6 +175,10 @@ interface StreamOptions {
      *  where the default `globalThis.fetch` doesn't support streaming properly.
      *  Passed directly to the underlying provider SDK. */
     fetch?: typeof globalThis.fetch;
+    /** Whether the target model supports image input. When false, image content
+     *  in user messages and tool_result messages is downgraded to a text placeholder
+     *  before being sent to the provider. Default: true. */
+    supportsImages?: boolean;
     /** Use streaming transport (default: true). When false, providers issue a
      *  single non-streaming request and synthesize events from the full response.
      *  The agent loop flips this to `false` as a fallback after repeated stream
@@ -368,4 +373,4 @@ interface PalsuProviderConfig {
  */
 declare function registerPalsuProvider(config?: PalsuProviderConfig): PalsuProviderHandle;
 
-export { type AssistantMessage, type CacheRetention, type ContentPart, type DoneEvent, type ErrorEvent, EventStream, GGAIError, type ImageContent, type Message, type PalsuModelConfig, type PalsuModelHandle, type PalsuProviderConfig, type PalsuProviderHandle, type PalsuProviderState, type PalsuResponse, type PalsuResponseFactory, type Provider, type ProviderEntry, ProviderError, type ProviderStreamFn, type RawContent, type ServerToolCall, type ServerToolCallEvent, type ServerToolDefinition, type ServerToolResult, type ServerToolResultEvent, type StopReason, type StreamEvent, type StreamOptions, type StreamResponse, StreamResult, type SystemMessage, type TextContent, type TextDeltaEvent, type ThinkingContent, type ThinkingDeltaEvent, type ThinkingLevel, type Tool, type ToolCall, type ToolCallDeltaEvent, type ToolCallDoneEvent, type ToolChoice, type ToolResult, type ToolResultMessage, type Usage, type UserMessage, palsuAssistantMessage, palsuText, palsuThinking, palsuToolCall, providerRegistry, registerPalsuProvider, stream };
+export { type AssistantMessage, type CacheRetention, type ContentPart, type DoneEvent, type ErrorEvent, EventStream, GGAIError, type ImageContent, type Message, type PalsuModelConfig, type PalsuModelHandle, type PalsuProviderConfig, type PalsuProviderHandle, type PalsuProviderState, type PalsuResponse, type PalsuResponseFactory, type Provider, type ProviderEntry, ProviderError, type ProviderStreamFn, type RawContent, type ServerToolCall, type ServerToolCallEvent, type ServerToolDefinition, type ServerToolResult, type ServerToolResultEvent, type StopReason, type StreamEvent, type StreamOptions, type StreamResponse, StreamResult, type SystemMessage, type TextContent, type TextDeltaEvent, type ThinkingContent, type ThinkingDeltaEvent, type ThinkingLevel, type Tool, type ToolCall, type ToolCallDeltaEvent, type ToolCallDoneEvent, type ToolChoice, type ToolResult, type ToolResultContent, type ToolResultMessage, type Usage, type UserMessage, palsuAssistantMessage, palsuText, palsuThinking, palsuToolCall, providerRegistry, registerPalsuProvider, stream };

package/dist/index.d.ts

@@ -23,10 +23,11 @@ interface ToolCall {
     name: string;
     args: Record<string, unknown>;
 }
+type ToolResultContent = string | (TextContent | ImageContent)[];
 interface ToolResult {
     type: "tool_result";
     toolCallId: string;
-    content: string;
+    content: ToolResultContent;
     isError?: boolean;
 }
 interface ServerToolCall {
@@ -174,6 +175,10 @@ interface StreamOptions {
      *  where the default `globalThis.fetch` doesn't support streaming properly.
      *  Passed directly to the underlying provider SDK. */
     fetch?: typeof globalThis.fetch;
+    /** Whether the target model supports image input. When false, image content
+     *  in user messages and tool_result messages is downgraded to a text placeholder
+     *  before being sent to the provider. Default: true. */
+    supportsImages?: boolean;
     /** Use streaming transport (default: true). When false, providers issue a
      *  single non-streaming request and synthesize events from the full response.
      *  The agent loop flips this to `false` as a fallback after repeated stream
@@ -368,4 +373,4 @@ interface PalsuProviderConfig {
  */
 declare function registerPalsuProvider(config?: PalsuProviderConfig): PalsuProviderHandle;
 
-export { type AssistantMessage, type CacheRetention, type ContentPart, type DoneEvent, type ErrorEvent, EventStream, GGAIError, type ImageContent, type Message, type PalsuModelConfig, type PalsuModelHandle, type PalsuProviderConfig, type PalsuProviderHandle, type PalsuProviderState, type PalsuResponse, type PalsuResponseFactory, type Provider, type ProviderEntry, ProviderError, type ProviderStreamFn, type RawContent, type ServerToolCall, type ServerToolCallEvent, type ServerToolDefinition, type ServerToolResult, type ServerToolResultEvent, type StopReason, type StreamEvent, type StreamOptions, type StreamResponse, StreamResult, type SystemMessage, type TextContent, type TextDeltaEvent, type ThinkingContent, type ThinkingDeltaEvent, type ThinkingLevel, type Tool, type ToolCall, type ToolCallDeltaEvent, type ToolCallDoneEvent, type ToolChoice, type ToolResult, type ToolResultMessage, type Usage, type UserMessage, palsuAssistantMessage, palsuText, palsuThinking, palsuToolCall, providerRegistry, registerPalsuProvider, stream };
+export { type AssistantMessage, type CacheRetention, type ContentPart, type DoneEvent, type ErrorEvent, EventStream, GGAIError, type ImageContent, type Message, type PalsuModelConfig, type PalsuModelHandle, type PalsuProviderConfig, type PalsuProviderHandle, type PalsuProviderState, type PalsuResponse, type PalsuResponseFactory, type Provider, type ProviderEntry, ProviderError, type ProviderStreamFn, type RawContent, type ServerToolCall, type ServerToolCallEvent, type ServerToolDefinition, type ServerToolResult, type ServerToolResultEvent, type StopReason, type StreamEvent, type StreamOptions, type StreamResponse, StreamResult, type SystemMessage, type TextContent, type TextDeltaEvent, type ThinkingContent, type ThinkingDeltaEvent, type ThinkingLevel, type Tool, type ToolCall, type ToolCallDeltaEvent, type ToolCallDoneEvent, type ToolChoice, type ToolResult, type ToolResultContent, type ToolResultMessage, type Usage, type UserMessage, palsuAssistantMessage, palsuText, palsuThinking, palsuToolCall, providerRegistry, registerPalsuProvider, stream };

package/dist/index.js

@@ -128,12 +128,70 @@ function zodToJsonSchema(schema) {
 }
 
 // src/providers/transform.ts
+var NON_VISION_USER_IMAGE_PLACEHOLDER = "(image omitted: model does not support images)";
+var NON_VISION_TOOL_IMAGE_PLACEHOLDER = "(tool image omitted: model does not support images)";
+function stripImages(content, placeholder) {
+  const out = [];
+  let lastWasPlaceholder = false;
+  for (const block of content) {
+    if (block.type === "image") {
+      if (!lastWasPlaceholder) out.push({ type: "text", text: placeholder });
+      lastWasPlaceholder = true;
+      continue;
+    }
+    out.push(block);
+    lastWasPlaceholder = block.text === placeholder;
+  }
+  return out;
+}
+function downgradeUnsupportedImages(messages, supportsImages) {
+  if (supportsImages !== false) return messages;
+  return messages.map((msg) => {
+    if (msg.role === "user" && Array.isArray(msg.content)) {
+      return { ...msg, content: stripImages(msg.content, NON_VISION_USER_IMAGE_PLACEHOLDER) };
+    }
+    if (msg.role === "tool") {
+      return {
+        ...msg,
+        content: msg.content.map(
+          (tr) => Array.isArray(tr.content) ? {
+            ...tr,
+            content: stripImages(tr.content, NON_VISION_TOOL_IMAGE_PLACEHOLDER)
+          } : tr
+        )
+      };
+    }
+    return msg;
+  });
+}
+function toolResultText(content) {
+  if (typeof content === "string") return content;
+  return content.filter((b) => b.type === "text").map((b) => b.text).join("\n");
+}
+function toolResultImages(content) {
+  if (typeof content === "string") return [];
+  return content.filter((b) => b.type === "image");
+}
 function toAnthropicCacheControl(retention, baseUrl) {
   const resolved = retention ?? "short";
   if (resolved === "none") return void 0;
   const ttl = resolved === "long" && (!baseUrl || baseUrl.includes("api.anthropic.com")) ? "1h" : void 0;
   return { type: "ephemeral", ...ttl && { ttl } };
 }
+function toAnthropicToolResultContent(content) {
+  if (typeof content === "string") return content;
+  return content.map((block) => {
+    if (block.type === "text") return { type: "text", text: block.text };
+    return {
+      type: "image",
+      source: {
+        type: "base64",
+        media_type: block.mediaType,
+        data: block.data
+      }
+    };
+  });
+}
 function toAnthropicMessages(messages, cacheControl) {
   let systemText;
   const out = [];
@@ -197,7 +255,7 @@ function toAnthropicMessages(messages, cacheControl) {
         content: msg.content.map((result) => ({
           type: "tool_result",
           tool_use_id: result.toolCallId,
-          content: result.content,
+          content: toAnthropicToolResultContent(result.content),
           is_error: result.isError
         }))
       });
@@ -368,11 +426,29 @@ function toOpenAIMessages(messages, options) {
       continue;
     }
     if (msg.role === "tool") {
+      const imageBlocks = [];
       for (const result of msg.content) {
+        const text = toolResultText(result.content);
+        const images = toolResultImages(result.content);
+        const hasText = text.length > 0;
         out.push({
           role: "tool",
           tool_call_id: remapToolCallId(result.toolCallId, idMap),
-          content: result.content
+          content: hasText ? text : "(see attached image)"
+        });
+        if (images.length > 0 && options?.supportsImages !== false) {
+          for (const img of images) {
+            imageBlocks.push({
+              type: "image_url",
+              image_url: { url: `data:${img.mediaType};base64,${img.data}` }
+            });
+          }
+        }
+      }
+      if (imageBlocks.length > 0) {
+        out.push({
+          role: "user",
+          content: [{ type: "text", text: "Attached image(s) from tool result:" }, ...imageBlocks]
         });
       }
     }
@@ -454,7 +530,8 @@ async function* runStream(options) {
   const isOAuth = options.apiKey?.startsWith("sk-ant-oat");
   const useStreaming = options.streaming !== false;
   const cacheControl = toAnthropicCacheControl(options.cacheRetention, options.baseUrl);
-  const { system: rawSystem, messages } = toAnthropicMessages(options.messages, cacheControl);
+  const downgradedMessages = downgradeUnsupportedImages(options.messages, options.supportsImages);
+  const { system: rawSystem, messages } = toAnthropicMessages(downgradedMessages, cacheControl);
   const system = isOAuth ? [
     {
       type: "text",
@@ -846,9 +923,11 @@ async function* runStream2(options) {
   const useStreaming = options.streaming !== false;
   const client = createClient2(options);
   const usesThinkingParam = options.provider === "glm" || options.provider === "moonshot" || options.provider === "xiaomi";
-  const messages = toOpenAIMessages(options.messages, {
+  const downgradedMessages = downgradeUnsupportedImages(options.messages, options.supportsImages);
+  const messages = toOpenAIMessages(downgradedMessages, {
     provider: options.provider,
-    thinking: !!options.thinking
+    thinking: !!options.thinking,
+    supportsImages: options.supportsImages
   });
   const defaultTemp = options.provider === "glm" ? 0.6 : void 0;
   const effectiveTemp = options.temperature ?? defaultTemp;
@@ -1140,7 +1219,8 @@ function streamOpenAICodex(options) {
 async function* runStream3(options) {
   const baseUrl = (options.baseUrl || DEFAULT_BASE_URL).replace(/\/+$/, "");
   const url = `${baseUrl}/codex/responses`;
-  const { system, input } = toCodexInput(options.messages);
+  const downgraded = downgradeUnsupportedImages(options.messages, options.supportsImages);
+  const { system, input } = toCodexInput(downgraded, { supportsImages: options.supportsImages });
   const body = {
     model: options.model,
     store: false,
@@ -1360,7 +1440,11 @@ function remapCodexId(id, idMap) {
   idMap.set(id, mapped);
   return mapped;
 }
-function toCodexInput(messages) {
+function codexToolResultText(content) {
+  if (typeof content === "string") return content;
+  return content.filter((b) => b.type === "text").map((b) => b.text).join("\n");
+}
+function toCodexInput(messages, options) {
   let system;
   const input = [];
   const idMap = /* @__PURE__ */ new Map();
@@ -1413,12 +1497,33 @@ function toCodexInput(messages) {
       continue;
     }
     if (msg.role === "tool") {
+      const toolImages = [];
       for (const result of msg.content) {
         const [callId] = result.toolCallId.includes("|") ? result.toolCallId.split("|", 2) : [result.toolCallId];
+        const text = codexToolResultText(result.content);
         input.push({
           type: "function_call_output",
           call_id: remapCodexId(callId, idMap),
-          output: result.content
+          output: text.length > 0 ? text : "(see attached image)"
+        });
+        if (options?.supportsImages !== false && Array.isArray(result.content)) {
+          for (const block of result.content) {
+            if (block.type === "image") toolImages.push(block);
+          }
+        }
+      }
+      if (toolImages.length > 0) {
+        input.push({
+          type: "message",
+          role: "user",
+          content: [
+            { type: "input_text", text: "Attached image(s) from tool result:" },
+            ...toolImages.map((img) => ({
+              type: "input_image",
+              detail: "auto",
+              image_url: `data:${img.mediaType};base64,${img.data}`
+            }))
+          ]
         });
       }
     }