Structured Data Retrieval with Sparrow using OCR and Vision LLM [Improved Accuracy]

I explain improvements I'm adding into Sparrow to achieve better accuracy for structured data. I'm using a method, where I run OCR step first, then construct advanced prompt with injected OCR data. This prompt is sent along with image to Vision LLM for structured data retrieval. All this happens as part of a single pipeline.