resources

seed/tools/ocr/TOOL.md

@@ -0,0 +1,27 @@
+---
+name: ocr
+label: OCR
+description: Extract text from an image file using the configured OCR service. Sends the image to the OCR API (OpenAI-compatible vision endpoint) and returns the extracted text. Use this tool whenever you need to read text from images, screenshots, documents, receipts, etc. Requires OCR to be configured in Settings → Resources.
+language: typescript
+inputs:
+  file_path:
+    type: string
+    description: Absolute path to the image file to extract text from
+  prompt:
+    type: string
+    description: Optional instructions for the OCR model (e.g. "extract only the table" or "return as markdown")
+    optional: true
+---
+
+# OCR Tool
+
+Extracts text from images using the configured OCR resource (OpenAI-compatible vision API).
+
+## Supported formats
+
+PNG, JPEG, WebP, GIF, and other common image formats.
+
+## Output
+
+Returns the extracted text content. For documents, preserves structure as markdown.
+For tables, uses markdown table format. For code screenshots, uses fenced code blocks.