Problem
ZenNotes already provides a great experience for storing and viewing PDFs and images alongside Markdown notes. However, the content inside these files is not currently searchable from the global search.
For users who use their vault as a long-term personal knowledge base or document archive, this makes it difficult to find information that exists only inside a PDF or image.
For example, I may remember a person's name, an invoice number, or a sentence from a document, but not remember which note or file contains it.
Proposed solution
Add optional local text extraction and OCR for PDFs and images, and include the extracted text in ZenNotes' global search index.
Ideally, the feature could support:
- Text extraction from text-based PDFs.
- OCR for scanned PDFs.
- OCR for common image formats such as PNG, JPG/JPEG and WebP.
- Automatic processing of new or modified assets.
- Batch processing for existing files in a vault.
- Global search results that return the original PDF/image when a match is found.
- Local/offline processing, without requiring an external AI or cloud service.
- A local cache/index for extracted text, without modifying the original PDF or image.
Example:
Vault
├── notes.md
├── documents/
│ ├── report.pdf
│ └── scanned-document.pdf
└── images/
└── receipt.jpg
Global search: "John Smith"
Results:
- notes.md
- report.pdf — match found inside PDF
- scanned-document.pdf — match found through OCR
- receipt.jpg — match found through OCR
Self-hosted use
It would be particularly useful if OCR/indexing could run server-side for self-hosted vaults.
That would allow a server to process new assets once and maintain the search index centrally, rather than requiring every client to perform OCR independently.
Conceptually:
New PDF/image
↓
ZenNotes server
↓
Text extraction / OCR
↓
Search index
↓
Desktop / Web / Mobile search
This would be especially useful for large vaults where the self-hosted server is always running.
MCP integration
Since ZenNotes already provides MCP support, it would also be useful if extracted text could eventually be exposed through the existing search/MCP tools.
This could allow an MCP client to find information stored inside PDFs and images without requiring the client to independently download and OCR every asset.
For example:
«"Find the document where John Smith is mentioned."»
The MCP search could return the PDF or image containing the matching OCR text.
This would make ZenNotes considerably more useful as a searchable personal knowledge base while keeping the actual OCR processing local and independent from LLM providers.
Possible future extension
As a separate future enhancement, the same indexing architecture could potentially support text extraction from other document formats such as:
However, I think PDF and image OCR would already cover the most important use case and could be implemented independently.
Problem
ZenNotes already provides a great experience for storing and viewing PDFs and images alongside Markdown notes. However, the content inside these files is not currently searchable from the global search.
For users who use their vault as a long-term personal knowledge base or document archive, this makes it difficult to find information that exists only inside a PDF or image.
For example, I may remember a person's name, an invoice number, or a sentence from a document, but not remember which note or file contains it.
Proposed solution
Add optional local text extraction and OCR for PDFs and images, and include the extracted text in ZenNotes' global search index.
Ideally, the feature could support:
Example:
Vault
├── notes.md
├── documents/
│ ├── report.pdf
│ └── scanned-document.pdf
└── images/
└── receipt.jpg
Global search: "John Smith"
Results:
Self-hosted use
It would be particularly useful if OCR/indexing could run server-side for self-hosted vaults.
That would allow a server to process new assets once and maintain the search index centrally, rather than requiring every client to perform OCR independently.
Conceptually:
New PDF/image
↓
ZenNotes server
↓
Text extraction / OCR
↓
Search index
↓
Desktop / Web / Mobile search
This would be especially useful for large vaults where the self-hosted server is always running.
MCP integration
Since ZenNotes already provides MCP support, it would also be useful if extracted text could eventually be exposed through the existing search/MCP tools.
This could allow an MCP client to find information stored inside PDFs and images without requiring the client to independently download and OCR every asset.
For example:
«"Find the document where John Smith is mentioned."»
The MCP search could return the PDF or image containing the matching OCR text.
This would make ZenNotes considerably more useful as a searchable personal knowledge base while keeping the actual OCR processing local and independent from LLM providers.
Possible future extension
As a separate future enhancement, the same indexing architecture could potentially support text extraction from other document formats such as:
However, I think PDF and image OCR would already cover the most important use case and could be implemented independently.