@ai-setting/roy-agent-core 1.5.91 → 1.5.93

package/dist/shared/@ai-setting/roy-agent-core-t7rw68q8.js

@@ -0,0 +1,185 @@
+// src/env/tool/built-in/understand-image.ts
+import { z } from "zod";
+
+// src/utils/media.ts
+import * as fs from "node:fs/promises";
+import * as path from "node:path";
+var ATTACH_FILE_MAX_BYTES = 10 * 1024 * 1024;
+function startsWith(buf, sig) {
+  if (buf.length < sig.length)
+    return false;
+  for (let i = 0;i < sig.length; i++) {
+    if (buf[i] !== sig[i])
+      return false;
+  }
+  return true;
+}
+function sniffMimeType(buffer, fallbackName) {
+  if (startsWith(buffer, [137, 80, 78, 71, 13, 10, 26, 10])) {
+    return "image/png";
+  }
+  if (startsWith(buffer, [255, 216, 255])) {
+    return "image/jpeg";
+  }
+  if (startsWith(buffer, [71, 73, 70, 56])) {
+    return "image/gif";
+  }
+  if (startsWith(buffer, [66, 77])) {
+    return "image/bmp";
+  }
+  if (startsWith(buffer, [37, 80, 68, 70, 45])) {
+    return "application/pdf";
+  }
+  if (startsWith(buffer, [82, 73, 70, 70]) && startsWith(buffer.subarray(8), [87, 69, 66, 80])) {
+    return "image/webp";
+  }
+  if (fallbackName) {
+    const ext = path.extname(fallbackName).toLowerCase();
+    const extMap = {
+      ".png": "image/png",
+      ".jpg": "image/jpeg",
+      ".jpeg": "image/jpeg",
+      ".gif": "image/gif",
+      ".webp": "image/webp",
+      ".bmp": "image/bmp",
+      ".pdf": "application/pdf",
+      ".mp3": "audio/mpeg",
+      ".wav": "audio/wav",
+      ".ogg": "audio/ogg",
+      ".m4a": "audio/mp4",
+      ".mp4": "video/mp4",
+      ".mov": "video/quicktime",
+      ".webm": "video/webm"
+    };
+    const mapped = extMap[ext];
+    if (mapped)
+      return mapped;
+  }
+  return "application/octet-stream";
+}
+function isMediaType(mime) {
+  if (typeof mime !== "string")
+    return false;
+  const m = mime.toLowerCase();
+  return m.startsWith("image/") || m.startsWith("audio/") || m.startsWith("video/") || m === "application/pdf";
+}
+var HTTP_URL_RE = /^https?:\/\//i;
+var DATA_URL_RE = /^data:([^;,]+)(;base64)?,(.*)$/s;
+async function parseImageInput(input) {
+  if (!input || typeof input !== "string" || input.trim().length === 0) {
+    throw new Error("parseImageInput: input is empty");
+  }
+  const trimmed = input.trim();
+  if (HTTP_URL_RE.test(trimmed)) {
+    return { type: "image", image: trimmed };
+  }
+  if (DATA_URL_RE.test(trimmed)) {
+    return { type: "image", image: trimmed };
+  }
+  return await parseLocalImagePath(trimmed);
+}
+async function parseLocalImagePath(filePath) {
+  let stat2;
+  try {
+    stat2 = await fs.stat(filePath);
+  } catch (err) {
+    throw new Error(`parseImageInput: cannot stat file '${filePath}': ${err.message}`);
+  }
+  if (!stat2.isFile()) {
+    throw new Error(`parseImageInput: not a regular file: ${filePath}`);
+  }
+  if (stat2.size > ATTACH_FILE_MAX_BYTES) {
+    throw new Error(`parseImageInput: file '${filePath}' exceeds 10 MiB limit ` + `(${stat2.size} bytes > ${ATTACH_FILE_MAX_BYTES} bytes)`);
+  }
+  const buffer = await fs.readFile(filePath);
+  const mime = sniffMimeType(buffer, filePath);
+  if (!mime.startsWith("image/")) {
+    throw new Error(`parseImageInput: file '${filePath}' is not an image (detected ${mime}). ` + `Use ContentPart type 'file' for non-image attachments.`);
+  }
+  return { type: "image", image: buffer };
+}
+
+// src/env/tool/built-in/understand-image.ts
+var injectedProvider = null;
+function setUnderstandImageLLMProvider(provider) {
+  injectedProvider = provider;
+}
+function getUnderstandImageLLMProvider() {
+  return injectedProvider;
+}
+async function resetUnderstandImageLLMProvider() {
+  injectedProvider = null;
+}
+function normalizeImageSource(raw) {
+  if (typeof raw !== "string")
+    return raw;
+  return raw.startsWith("@") ? raw.slice(1) : raw;
+}
+var understandImageTool = {
+  name: "understand_image",
+  description: "Analyze or describe the contents of an image using a multimodal LLM. " + "Provide a prompt describing what you want to extract or understand " + "(e.g. 'Describe this image in detail', 'Read the text in the receipt', " + "'What color is the car?'). The image_source can be an HTTP(S) URL, a " + "data:base64 URL, or a local file path (optionally prefixed with '@').",
+  parameters: z.object({
+    prompt: z.string().min(1).describe("A text prompt describing what you want to analyze or extract from the image."),
+    image_source: z.string().min(1).describe("The location of the image to analyze. Accepts HTTP(S) URL, data:base64 URL, or local file path. Optional '@' prefix is stripped.")
+  }),
+  metadata: {
+    category: "ai",
+    tags: ["multimodal", "vision", "image", "llm", "ocr"],
+    version: "1.0.0"
+  },
+  permission: {
+    level: "safe"
+  },
+  execute: async (args, _ctx) => {
+    const startTime = Date.now();
+    const { prompt, image_source } = args;
+    if (!injectedProvider) {
+      return {
+        success: false,
+        output: "",
+        error: "understand_image is not initialized: no LLM provider has been " + "registered. AgentComponent should call setUnderstandImageLLMProvider() " + "at agent startup.",
+        metadata: { execution_time_ms: Date.now() - startTime }
+      };
+    }
+    let imagePart;
+    try {
+      imagePart = await parseImageInput(normalizeImageSource(image_source));
+    } catch (err) {
+      return {
+        success: false,
+        output: "",
+        error: `failed to read image: ${err.message}`,
+        metadata: { execution_time_ms: Date.now() - startTime }
+      };
+    }
+    const messages = [
+      {
+        role: "user",
+        content: [
+          { type: "text", text: prompt },
+          imagePart
+        ]
+      }
+    ];
+    let responseText;
+    try {
+      responseText = await injectedProvider(messages);
+    } catch (err) {
+      return {
+        success: false,
+        output: "",
+        error: `LLM invocation failed: ${err.message}`,
+        metadata: { execution_time_ms: Date.now() - startTime }
+      };
+    }
+    return {
+      success: true,
+      output: responseText,
+      metadata: {
+        execution_time_ms: Date.now() - startTime
+      }
+    };
+  }
+};
+
+export { ATTACH_FILE_MAX_BYTES, sniffMimeType, isMediaType, parseImageInput, setUnderstandImageLLMProvider, getUnderstandImageLLMProvider, resetUnderstandImageLLMProvider, understandImageTool };

package/dist/shared/@ai-setting/{roy-agent-core-b1vkkcbr.js → roy-agent-core-x65fyjaw.js}

@@ -5,7 +5,7 @@ import {
   BackgroundTaskManager,
   createDelegateTool,
   createStopTool
-} from "./roy-agent-core-at05dk0z.js";
+} from "./roy-agent-core-ss92p0r6.js";
 import {
   SQLiteTaskStore,
   getDefaultTaskDbPath

package/dist/utils/temp-file-manager/index.js

@@ -0,0 +1,10 @@
+import {
+  TempFileManager
+} from "../../shared/@ai-setting/roy-agent-core-5h49ct1a.js";
+import"../../shared/@ai-setting/roy-agent-core-vyygk314.js";
+import"../../shared/@ai-setting/roy-agent-core-7z9b1fm8.js";
+import"../../shared/@ai-setting/roy-agent-core-c6592r3c.js";
+import"../../shared/@ai-setting/roy-agent-core-fs0mn2jk.js";
+export {
+  TempFileManager
+};

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-setting/roy-agent-core",
-  "version": "1.5.91",
+  "version": "1.5.93",
   "type": "module",
   "description": "Core SDK for roy-agent - Environment, Components, Tools, Sessions, Tasks",
   "main": "./dist/index.js",