goblin-agents 2.2.3 → 2.3.0

package/lib/llm/aiAgent.js

@@ -183,6 +183,8 @@ const defaultValues = {
 
 class AiAgentLogic extends Elf.Archetype {
   static db = 'agents';
+  static indices = ['role'];
+
   state = new AiAgentState({
     id: undefined,
     version: 1,
@@ -433,13 +435,13 @@ class AiAgent extends Elf {
     }
   }
 
-  async readPDFpages(pdfPath) {
+  async readPDFpages(pdfPath, pages) {
     await this._mutex.lock(this.id);
     this.quest.defer(() => this._mutex.unlock(this.id));
 
     await this._loadProvider();
 
-    const images = await getPDFImages(pdfPath);
+    const images = await getPDFImages(pdfPath, pages);
     const descriptions = [];
     for (const image of images) {
       const messages = [
@@ -448,7 +450,7 @@ class AiAgent extends Elf {
           content: [
             {
               type: 'text',
-              text: OCR_SYSTEM_PROMPT,
+              text: this.state.prompt || OCR_SYSTEM_PROMPT,
             },
             {
               type: 'image_url',
@@ -469,6 +471,37 @@ class AiAgent extends Elf {
     return descriptions;
   }
 
+  async vision(context, image) {
+    await this._mutex.lock(this.id);
+    this.quest.defer(() => this._mutex.unlock(this.id));
+
+    await this._loadProvider();
+    const messages = [
+      {role: 'system', content: this.state.prompt},
+      {
+        role: 'user',
+        content: [
+          {
+            type: 'image_url',
+            image_url: {
+              url: image,
+            },
+          },
+          {
+            type: 'text',
+            text: context,
+          },
+        ],
+      },
+    ];
+    const result = await this._provider.chat({
+      model: this.state.model,
+      reasoning: this.state.reasoning,
+      messages,
+    });
+    return result.message.content;
+  }
+
   async embed(rawText) {
     if (!rawText) {
       return null;
@@ -522,7 +555,7 @@ class AiAgent extends Elf {
     await this._loadProvider();
     const contextMessages = this.state.messages[contextId] || [];
     const messages = [
-      {role: 'system', content: this.state.prompt},
+      {role: 'system', content: this.state.prompt || ''},
       ...Array.from(contextMessages).map((m) => m._state.toJS()), //history
     ];
 

package/lib/llm/utils.js

@@ -24,7 +24,7 @@ Ignore les artefacts visuels inutiles.
 Corrige les erreurs évidentes et restitue un texte propre et fidèle à l'original.
 Ta réponse contient uniquement le texte lisible.`;
 
-const getPDFImages = async (pdfPath) => {
+const getPDFImages = async (pdfPath, pagesToProcess) => {
   const {pdfToPng} = loadPdfToPngConverter();
   // The function accepts PDF file path or a Buffer
   const pages = await pdfToPng(pdfPath, {
@@ -35,7 +35,7 @@ const getPDFImages = async (pdfPath) => {
     //outputFolder: 'output/folder', // Folder to write output PNG files. If not specified, PNG output will be available only as a Buffer content, without saving to a file.
     //outputFileMaskFunc: (pageNumber) => `page_${pageNumber}.png`, // Output filename mask function. Example: (pageNumber) => `page_${pageNumber}.png`
     //pdfFilePassword: password, // Password for encrypted PDF.
-    //pagesToProcess: [1, 3, 11], // Subset of pages to convert (first page = 1), other pages will be skipped if specified.
+    pagesToProcess, // Subset of pages to convert (first page = 1), other pages will be skipped if specified.
     //strictPagesToProcess: false, // When `true`, will throw an error if specified page number in pagesToProcess is invalid, otherwise will skip invalid page. Default value is false.
     verbosityLevel: 0, // Verbosity level. ERRORS: 0, WARNINGS: 1, INFOS: 5. Default value is 0.
   });

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "goblin-agents",
-  "version": "2.2.3",
+  "version": "2.3.0",
   "description": "AI Agents framework",
   "author": "Epsitec SA",
   "contributors": [