logicmatters06/agrimitra-dataset

AgriMitra Agriculture Corpus

This dataset backs the AgriMitra learning platform: a searchable agriculture library with structured course topics, a crop knowledge base, and a retrieval- augmented chatbot.

The repository contains 36 source PDFs at the root and 20 generated JSON exports under json/. The source Markdown used by the web app lives in the GitHub repository.

Structure

.
├── *.pdf                       # 36 agriculture course PDFs
└── json/
    ├── index.json              # course index / metadata
    ├── crops.json              # 22 crops × 15 guide sections
    ├── topics.json             # structured reader topics
    ├── chatbot-index.json      # RAG/keyword search chunks
    └── *_topic_level_*.json    # 16 normalized topic exports

Source and pipeline

The PDFs are agriculture course material collected for this educational project. The JSON exports are generated from those PDFs with the scripts in scripts/pdf-to-json/ and scripts/build-*.mjs:

1. extract text/headings from the PDFs;
2. normalize topic exports into topics.json;
3. build the 22-crop guide (crops.json);
4. chunk the corpus into chatbot-index.json for RAG and keyword fallback.

The crop classifier is published separately at logicmatters06/agrimitra-crop-model. Its 6,596-row training CSV lives in the GitHub repository under crop-ml/.

Rights and licence

The source PDFs remain the property of their original authors and publishers. This project does not claim ownership of them and does not relicense them. They are shared here for educational and research access only. If you are a rights holder and want a file corrected or removed, please open a discussion on this repository and it will be actioned promptly.

Project-created metadata and JSON annotations are released under CC BY 4.0; the source PDFs are excluded from that grant. Verify the rights of each original document before redistributing it elsewhere.

Usage

# PDFs only
hf download logicmatters06/agrimitra-dataset \
  --type dataset --include "*.pdf" --local-dir agri-data

# JSON exports only
hf download logicmatters06/agrimitra-dataset \
  --type dataset --include "json/*" --local-dir /tmp/agrimitra-json

See the AgriMitra README for the full setup and rebuild instructions.

Attribution

If you use the corpus, cite the AgriMitra project and the original authors of the individual course documents. The 22 base crop classes in the merged classifier dataset correspond to the widely used public Kaggle Crop Recommendation dataset; the remaining 29 classes are additional Indian crop records compiled for this project.