@holoscript/holoscript-agent 2.1.4 → 2.1.5

package/dist/index.js

@@ -1082,13 +1082,13 @@ var MESH_TOOLS = [
   },
   {
     name: "vision_analyze",
-    description: "Analyze an image using the local Fara-7B vision model (Ollama on loopback). Reads the image file at `image_path`, sends it to fara:7b via the local Ollama API, and returns the model's text analysis. Counts as a productive tool call \u2014 use for GUI-grounding, visual QA, image captioning, or any task that requires perceiving image content. Only available on surfaces with a local Ollama instance running fara:7b.",
+    description: "Analyze an image using the local Fara-7B vision model (Ollama on loopback). Reads the image file at `image_path` (max 512KB \u2014 downscale larger images first), sends it to the vision model via the local Ollama API (env: HOLOSCRIPT_AGENT_VISION_MODEL), and returns the model's text analysis. Counts as a productive tool call \u2014 use for GUI-grounding, visual QA, image captioning, or any task that requires perceiving image content. Only available on surfaces with a local Ollama instance and HOLOSCRIPT_AGENT_LOCAL_LLM_BASE_URL set.",
     input_schema: {
       type: "object",
       properties: {
         image_path: {
           type: "string",
-          description: "Absolute path to the image file (png, jpg, webp, gif)"
+          description: "Absolute path to the image file (png, jpg, webp) \u2014 must be under 512KB"
         },
         prompt: {
           type: "string",
@@ -1096,7 +1096,7 @@ var MESH_TOOLS = [
         },
         model: {
           type: "string",
-          description: 'Ollama model tag to use (default: "fara:7b")'
+          description: "Ollama model tag override (default: HOLOSCRIPT_AGENT_VISION_MODEL env var)"
         }
       },
       required: ["image_path"]
@@ -1350,19 +1350,29 @@ ${truncated}`);
       const denied = checkReadAllowed(imagePath);
       if (denied) return errResult(use.id, `vision_analyze: ${denied}`);
       const prompt = String(use.input.prompt ?? "Describe this image in detail.");
-      const model = String(use.input.model ?? "fara:7b");
+      const model = String(
+        use.input.model ?? process.env.HOLOSCRIPT_AGENT_VISION_MODEL ?? "fara:7b"
+      );
       const ollamaBase = process.env.HOLOSCRIPT_AGENT_LOCAL_LLM_BASE_URL;
       if (!ollamaBase) {
         return errResult(
           use.id,
-          "vision_analyze: HOLOSCRIPT_AGENT_LOCAL_LLM_BASE_URL is not set \u2014 configure it to point to your local Ollama instance (e.g. http://holojetson.local:11434)"
+          "vision_analyze: HOLOSCRIPT_AGENT_LOCAL_LLM_BASE_URL is not set \u2014 configure it to point to your local Ollama instance"
         );
       }
+      const MAX_IMAGE_BYTES = 512e3;
       const TIMEOUT_MS = 12e4;
       const controller = new AbortController();
       const timer = setTimeout(() => controller.abort(), TIMEOUT_MS);
       try {
         const imageBytes = await readFile3(imagePath);
+        if (imageBytes.length > MAX_IMAGE_BYTES) {
+          clearTimeout(timer);
+          return errResult(
+            use.id,
+            `vision_analyze: image is ${Math.round(imageBytes.length / 1024)}KB \u2014 exceeds ${MAX_IMAGE_BYTES / 1024}KB limit. Downscale the image first (e.g. to 256\xD7256 or smaller) then retry vision_analyze.`
+          );
+        }
         const imageB64 = imageBytes.toString("base64");
         const res = await fetch(`${ollamaBase}/api/generate`, {
           method: "POST",