Philately Collection Management System

Save days or weeks of tedious data entry with tweezers and a magnifying glass.

Overview

This project is a Python-based command-line system for managing a philatelic (stamp) collection. It leverages modern AI to process entire directories of stamp album images, extract detailed metadata, and generate a comprehensive, queryable inventory. By using litellm, it supports multiple AI model providers (e.g., Google Gemini, xAI Grok) for maximum flexibility and cost-effectiveness.

Key features include:

• Multi-Model AI Processing: Analyzes stamp images to extract details like country, year, and condition using a two-pass system with configurable "low-cost" and "high-cost" vision models.
• Data Enrichment: Uses powerful text models to enrich the initial data with estimated values, historical context, and philatelic remarks.
• False Positive Detection: Includes a dedicated phase to re-examine high-value items and automatically flag illustrations or other non-stamp entities.
• Persistent, Auditable Storage: Maintains a master inventory in master_inventory.csv that includes all processed stamps, deacquired items, and verification results.
• Comprehensive Reporting: Generates detailed JSON summaries, high-value reports, and content-ready CSVs for platforms like Substack.
• Modular, Phase-Based Execution: Allows you to run the entire pipeline or specific phases (e.g., analysis, enrichment, reporting) independently.
• Command-Line and GUI Interfaces: Provides both a command-line tool (philately) for automated processing and a Streamlit-based GUI (philately-ui) for interactive use.

Prerequisites

• Python: Version 3.13 or higher.
• API Keys: At least one API key for a supported provider (e.g., Google, xAI). These should be set in a .env file.
• System Dependencies:
  • On Ubuntu/Debian: sudo apt-get install libopencv-dev.
  • On macOS: brew install opencv.

Installation

1. Clone the Repository:

   git clone <repository-url>
   cd <repository-directory>
   

2. Create a Virtual Environment:

   python3 -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   

3. Install the Package:

   pip install .
   

4. Set Up Environment Variables:

   Create a .env file in the project root and add your API key(s):

   echo "GOOGLE_API_KEY=your-google-api-key" > .env
   echo "XAI_API_KEY=your-xai-api-key" >> .env
   

5. Prepare Directory Structure:

   • Place stamp images in a directory (e.g., stamps/), organized into subdirectories for each album (e.g., stamps/Isle of Man/).
   • The output directory will be created automatically to store all generated files.

Usage

This package provides two primary entry points: a command-line interface (CLI) and a graphical user interface (GUI).

Command-Line Interface (CLI)

The philately command allows you to run the entire pipeline or specific phases using command-line flags.

Command-Line Flags

The command-line flags are the same as described in the table below.

Graphical User Interface (GUI)

The philately-ui command launches a Streamlit-based web interface that allows you to configure and run the processing pipeline interactively.

philately-ui

Example Commands (CLI)

1. Run the full pipeline on all images:

philately --image-dir ./stamps --output-dir ./output

2. Run only the image analysis phase on the first 10 images:

philately --run-analysis --max-images 10

3. Run the false-positive check on the top 3 most valuable stamps with debug logging:

philately --run-false-positive-check --false-positive-check-limit 3 --debug

4. Generate a Substack export with the top 20 most valuable items:

philately --run-substack-export --substack-items 20

5. Re-run only the enrichment and summary phases:

philately --run-enrichment --run-summaries

Command-Line Flags

Flag	Default	Description
--image-dir	stamps	Directory containing stamp images organized in album folders.
--output-dir	output	Directory to save all outputs.
--confidence-threshold	5	Confidence score (1-7) below which to trigger re-analysis with a high-cost model.
--max-images	None	Limit the number of images to process for testing.
--high-value-threshold	1000	USD threshold to consider a stamp as high-value for reporting.
--debug	False	Enable debug-level logging for verbose output, including API payloads.
--low-cost-model	gemini/gemini-1.5-flash-latest	The vision model for the initial, low-cost pass.
--high-cost-model	gemini/gemini-1.5-pro-latest	The vision model for the high-confidence re-analysis pass.
--narrative-model	gemini/gemini-1.5-pro-latest	The text model for enrichment and summaries.
--collection-summary-model	gemini/gemini-1.5-pro-latest	The high-context model for the final collection-wide summary.
--run-analysis	False	Run only the image analysis phase.
--run-enrichment	False	Run only the philatelic enrichment phase.
--run-summaries	False	Run the full clustering and summary phase.
--run-high-value-report	False	Run only the high-value stamp report generation phase.
--run-collection-summary-only	False	Run only the final collection-wide summary generation.
--run-false-positive-check	False	Run a re-examination of high-value stamps to find false positives.
--false-positive-check-limit	5	Limit the number of stamps to check in the false-positive phase (0 for all).
--run-substack-export	False	Generate a CSV export formatted for Substack posts.
--substack-items	10	Number of top items to include in the Substack export (0 for all).

Output Files

All outputs are saved to the directory specified by --output-dir.

• master_inventory.csv: The master database of all stamps, including detailed analysis and verification data.
• stamp_inventory.json: A structured JSON file containing all data, including collection-wide statistics and narrative summaries.
• false_positive_check_report.csv: A summary of high-value items that were checked for authenticity.
• high_value_summary.csv: A CSV listing all stamps identified as high-value.
• substack_export.csv: A CSV formatted for easy import into content platforms like Substack.
• cropped_entities/: Directory of cropped images for each identified stamp.
• thumbnails/: Directory of 100x100px thumbnails for each stamp.
• high_value_reports/: Individual Markdown reports for each high-value stamp.

Example Data Records

1. Master Inventory Record (master_inventory.csv)

A single row contains the complete data for one stamp.

stamp_id	album	page_filename	common_name	nationality	year	face_value	condition	confidence	estimated_value_high	is_verified_real	verification_reason
a1b2c3d4-...	Isle of Man	IMG_1172.jpeg	1973 Manx Cat	Isle of Man	1973	3p	Mint	7	15	True	This appears to be a genuine, mounted stamp with clear perforations and color.

2. Cluster Summary (stamp_inventory.json)

Summaries provide statistics and a narrative for a specific group of stamps (e.g., an album).

{
    "album_Isle_of_Man": {
        "statistics": {
            "item_count": 58,
            "album_count": 1,
            "countries_represented": 1,
            "year_range": "1973 - 1998",
            "total_value_low": 150,
            "total_value_high": 450,
            "condition_distribution": {
                "Mint": 45,
                "Used": 13
            }
        },
        "narrative_summary": "This cluster from the 'Isle of Man' album represents a strong collection of modern issues, primarily from the 1970s and 1980s. The thematic focus is on local culture, transportation, and fauna, with the 'Manx Cat' and 'TT Races' series being prominent highlights. The overall condition is excellent, with a majority of items in mint condition. A notable gap is the absence of earlier Victorian-era issues."
    }
}

3. False Positive Check Report (false_positive_check_report.csv)

This report provides a clear audit trail for the verification process.

stamp_id	common_name	estimated_value_high	page_filename	is_verified_real	cropped_image_path	verification_reason	action_taken
e5f6g7h8-...	Penny Black	2500	IMG_1245.JPG	False	cropped_entities/e5f6g7h8-..._cropped.jpg	The image is a black and white printed illustration, lacking color and physical depth.	Marked as deacquired (illustration)