Feature Request

Add native support for non-text file types including documents, PDFs, and images to Basic Memory.

Problem

Currently, Basic Memory is optimized for plain text and Markdown files. Users working with research papers, documentation, screenshots, diagrams, and other visual/document content need to manually extract or describe the content before it can be incorporated into their knowledge base.

Proposed Solution

Enable Basic Memory to:

1. Accept and store documents (PDF, DOCX, etc.) and images (PNG, JPG, etc.)
2. Extract text content from PDFs and documents for indexing and search
3. Process images with OCR or vision models to make visual content searchable
4. Maintain original files alongside extracted/processed content
5. Create automatic connections between documents/images and related notes

Use Cases

• Research: Import academic papers (PDFs) and automatically extract key concepts, connect to related notes
• Documentation: Save screenshots of code, errors, or UI designs with automatic text extraction
• Learning: Store textbook pages, diagrams, and visual materials as part of knowledge graph
• Work notes: Import meeting notes from various formats, connect to project documentation

Implementation Considerations

• File storage strategy (local vs cloud bucket, size limits)
• Text extraction pipeline (PDF parsing, OCR for images)
• Vision model integration for image understanding
• Metadata extraction (authors, dates, titles from PDFs)
• Search indexing for extracted content
• UI for viewing/previewing non-text files
• Cost implications for processing (OCR, vision models)

Benefits

• Complete knowledge capture (text + visual + documents)
• Reduced friction in knowledge building workflow
• Better support for academic/research use cases
• Richer semantic connections across all content types

Originally filed by @nellins in basicmachines-co/basicmemory.com#6