github icereed/paperless-gpt v0.9.0
v0.9.0 - AI powered Correspondent detection

latest releases: v0.22.0, v0.21.0, v0.20.0...
7 months ago

paperless-gpt v0.9.0

We’re excited to announce paperless-gpt v0.9.0, introducing Automatic Correspondent Management on top of our existing LLM-powered OCR, AI-driven tagging, and document title suggestions. While OCR remains unchanged since our last release, the new Correspondent Management feature is here to simplify your workflow even further.

New in v0.9

1. Correspondent Management

  • Auto-Identify Senders: paperless-gpt now analyzes document data (including OCR text, if enabled) to auto-generate correspondents.
  • Manual Overrides: If the AI guess is off or you prefer custom naming, you can easily edit or override the suggested correspondent.
  • Improved Tagging Synergy: Combine auto-generated correspondents with your manual, automatic, or LLM-based tags for a more holistic organization.

2. Updated Environment Variables

  • AUTO_GENERATE_CORRESPONDENTS: Set to true to opt into the new auto-discovery of senders.
  • CORRESPONDENT_BLACK_LIST: Provide a comma-separated list of names you want to exclude from auto-detection (e.g. John Doe,Unwanted Bot).

Note: The OCR features remain exactly the same as in previous releases. No changes were made to the LLM-based OCR pipeline, page limits, or environment variables relating to OCR in this update.


Why It Matters

  1. Less Data Entry: No need to type in the same sender’s name repeatedly—let the AI do that for you.
  2. Enhanced Organization: Search by sender/correspondent in paperless-ngx to quickly locate relevant documents.
  3. Flexible Workflows: Keep your existing OCR and tagging setup intact, while seamlessly layering on correspondent management—manual or automatic.

OCR Still Rock-Solid

While v0.9 introduces no changes to the OCR functionality, you can continue to leverage:

  • LLM-based OCR from previous versions for context-aware text extraction.
  • Manual or Automatic tag flows, powered by environment variables like MANUAL_TAG, AUTO_TAG, and AUTO_OCR_TAG.

Upgrade Steps

  1. Pull the Latest Docker Image
    docker pull icereed/paperless-gpt:latest
  2. Set Up New Variables (Optional):
    AUTO_GENERATE_CORRESPONDENTS=true # is true by default
    CORRESPONDENT_BLACK_LIST="John Doe,Unwanted Bot"
  3. Restart your container.
    • docker-compose up -d or docker run -d ...

Example Flow

sequenceDiagram
    participant PaperlessGPT as paperless-gpt
    participant LLM as Language Model
    participant PaperlessNGX as paperless-ngx

    PaperlessGPT->>PaperlessNGX: Retrieve Document
    PaperlessNGX-->>PaperlessGPT: Return Document Data
    PaperlessGPT->>LLM: (If OCR is enabled) Perform LLM-based OCR
    LLM-->>PaperlessGPT: Return Extracted Text
    PaperlessGPT->>LLM: Perform LLM-based Metadata Extraction
    LLM-->>PaperlessGPT: Return Extracted Metadata
    PaperlessGPT->>PaperlessNGX: Update Document with Correspondent, Title, Tags
Loading

Poetic Nod to the Future

"An AI that not only sees—but knows who writes,
Correspondent auto-chosen, no more manual fights,
OCR stays sturdy, forging text from scans in lights,
Version 0.9 sets the stage for still bigger heights!"


Enjoy the Correspondent Magic

With v0.9, you’ll see new power in how you manage your docs—no changes to OCR, just brand-new ways to keep track of who sent each document. Let us know what you think; your feedback helps guide what’s next!

Happy Correspondent-Organizing & Document Managing!

Full Changelog: v0.8.4...v0.9.0

Don't miss a new paperless-gpt release

NewReleases is sending notifications on new releases.