Understanding Document Metadata and Analysis
When you upload documents to Andri, the platform automatically extracts metadata and performs intelligent analysis to make your files searchable and usable by the AI.
Viewing Document Metadata
From your Knowledge Base, click the dropdown menu next to any document and select View Metadata.
The metadata panel displays all extracted information about the document.
What Metadata Andri Extracts
For every document, Andri captures basic file information including filename, file type, size, and upload date. The AI also extracts content-specific metadata depending on file type.
PDF and Word documents: Full text extraction, page count, creation date, author (if embedded), and OCR for scanned documents.
Images: Image description generated by AI, text extraction via OCR if the image contains readable text, and technical details like dimensions and format.
Audio files: Complete transcription with speaker diarization, audio duration, format details, and speaker identification that you can customize.
Excel files: Sheet names, data structure, and key information extracted from cells.
Email files (.eml): Sender, recipient, subject line, date, body text, and all attachments processed separately.
AI-Generated Summaries
Andri automatically generates a summary for each document, identifying key topics, parties mentioned, dates, and relevant legal concepts. These summaries help you quickly understand document content and improve AI search accuracy.
Review AI-generated summaries for complex documents. If the summary misses important details, you can guide the AI by asking specific questions about the document.
How Metadata Improves AI Performance
The metadata and analysis Andri extracts allows the AI to quickly find relevant information, understand document context, cite sources accurately with page numbers, and distinguish between document types when answering questions.
All metadata extraction and analysis happens automatically when you upload files. No manual tagging or processing is required.