Archivist

Organize PDFs with AI.

Personally, I scan documents I'd like to keep to a network drive, where the archivist processes and organizes the PDFs for me.

Installation

Ollama is needed for the archivist's capabilities. Install Ollama somewhere accessible to the archivist.

The archivist is intended to run via Docker:

$ docker run \
  --net=host \
  -v ~/Documents:/archive \
  -v ~/Downloads/inbox:/inbox \
  -e ARCHIVIST_ARCHIVE_DIR="/archive" \
  -e ARCHIVIST_INBOX_DIR="/inbox" \
  ghcr.io/jdav-dev/archivist

ARCHIVIST_INBOX_DIR and ARCHIVIST_ARCHIVE_DIR are required environment variables, specifying the respective source and destination directories for PDFs. Optional environment variables include ARCHIVIST_OLLAMA_BASE_URL to override the URL for Ollama's API (defaults to "http://localhost:11434/api") and ARCHIVIST_OLLAMA_TIMEOUT_SECONDS to override how long to wait on responses from Ollama (defaults to "60").

Development

The archivist uses development containers for local development. After cloning the repository and before opening the project in its development container, a file must be created at .devcontainer/docker-compose.extend.yml. This file customizes the Docker Compose development stack for your local machine. For example:

services:
  archivist:
    volumes:
      - ~/Documents:/archive
      - ~/Downloads:/inbox
  ollama:
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [ gpu ]

This will mount ~/Documents to /archive and ~/Downloads to ~/inbox in your development environment. It will also give the Ollama container access to a Nvidia GPU. Change the directories as needed for your machine, and change or remove the Ollama section if you do not have an Nvidia GPU available.