🧠 LLM File Describe

LLM File Describe is a QGIS plugin that uses a locally hosted large language model (via Ollama) to analyse and describe geospatial data files.
It supports both vector formats (such as .shp and .gpkg) and raster formats (such as .tif and .img).
The plugin reads metadata and sample data from the file and asks the language model to infer what the dataset represents (e.g., cities, roads, land use) and summarise its contents.

---

🚀 Features

• Vector file support:
  Loads .shp or .gpkg files, extracts field names, counts features and samples up to 10 rows from the attribute table to provide context.

• Raster file support:
  Loads raster formats (.tif, .tiff, .img), reports image dimensions, pixel size, bounding box coordinates, and basic statistics (minimum, maximum, mean, standard deviation, and pixel count) for each band.

• Language selection:
  Choose the language of the model’s response (English or French) from a drop-down list.

• Follow-up questions:
  After the initial description, you can ask questions about the dataset and the model will answer based on the previous context.

• Information dialog:
  An info button (?) explains the purpose and usage of the plugin directly in QGIS.

---

🧩 Requirements

• QGIS 3.0+ — Tested with QGIS 3.40 LTR, but should work with any modern version.
• Ollama — Must be installed and running locally, with at least one LLM model (e.g. llama3.1).
  The plugin sends HTTP requests to the Ollama API at:
  http://localhost:11434/api/chat

---

⚙️ Installation

Option 1 — From ZIP

1. Download or clone this repository and locate the file:
   llm_file_describe_final.zip
2. In QGIS, open Plugins → Manage and Install Plugins…
3. Click Install from ZIP… and select the downloaded ZIP file.
4. Restart QGIS if necessary, then enable the plugin from the Plugins menu.

Option 2 — Manual Installation

Unzip the folder llm_file_describe directly into your QGIS profile’s plugins directory:
~/.local/share/QGIS/QGIS3/profiles/default/python/plugins/

---

🧭 Usage

1. Open the plugin:
   Once enabled, access it from Plugins → LLM File Describe, or via the toolbar icon.

2. Select a file:
   Use the ... button to browse for a vector or raster file (.shp, .gpkg, .tif, etc.).

3. Choose model and language:

   • Ollama model: defaults to llama3.1 (you can change it if needed).
   • Response language: English or French.

4. Analyze:
   Click Analyze — the plugin reads metadata, queries the LLM, and displays a textual summary.

5. Ask questions:
   After the description appears, type a question in the Question field and click Ask.
   The model will answer using the dataset’s description as context.

6. Info:
   Click the information icon (?) to read about the plugin’s purpose and usage.

---

🤝 Contributing

Contributions are welcome!
If you encounter a bug, have a feature request, or would like to improve the plugin, please open an issue or pull request on this GitHub repository.

---

📄 License

This project is licensed under the GNU General Public License v3 (GPL-3).
See the LICENSE file for details.

---

👤 Author

Developed and maintained by Romain Wenger
📧 romain.wenger@live-cnrs.unistra.fr

---