This server has two OCR tools with basic definitions but significant quality gaps. Both tools have descriptions and parameter schemas, but descriptions are minimal and lack critical context for LLM tool selection. Parameter descriptions are generic. There is no documented output schema, the return type annotation exists in code but is not described in the tool definition. Error handling is absent: the tools return empty lists on failure (silent failure) rather than actionable error messages. No enum constraints, no validation guidance, and no recovery hints. The tool names lack action verbs ('ocr_by_content', 'ocr_by_path'), better names would be 'recognize_text_from_base64' or 'recognize_text_from_file'. The parameter descriptions do not explain format constraints (e.g., what image formats are accepted, what base64 encoding is expected). Overall, the definitions would not enable an LLM to reliably select and invoke these tools without ambiguity.

• Rename tools to follow verb_noun convention: 'recognize_text_from_base64' and 'recognize_text_from_file' instead of 'ocr_by_content' and 'ocr_by_path'. This makes intent obvious to LLMs before reading the description.
• Add explicit output schema documentation in tool descriptions: 'Returns a list of recognized text blocks, each containing the extracted text string and confidence score (if available).' Include the exact JSON structure or object definition.
• Enhance parameter descriptions with format constraints: for base64_data, specify 'Base64-encoded image (PNG, JPEG, GIF, or WebP). Maximum 50MB after decoding.' For path, add 'Absolute or relative file path. Must be readable by the server process.'
• Implement structured error handling instead of silent empty-list returns. Return error objects with actionable messages: {'error': 'invalid_base64', 'message': 'Base64 string is malformed. Verify the input is valid UTF-8 base64.', 'recovery': 'Retry with a valid base64-encoded image.'} or similar.
• Add a clarifying description hint between the two tools: 'Use ocr_by_content if you have image data as base64; use ocr_by_path if you have a file path on disk.' This guides tool selection.
• Validate and sanitize file paths in ocr_by_path to prevent directory traversal (e.g., reject paths containing '..' or absolute paths outside a configured directory). Return a clear permission error if access is denied.
• Document which image formats are supported and any image size limits. Include this in parameter descriptions and tool description.