[general]
name=LibreGeoLens
qgisMinimumVersion=3.30
description=Experiment with MLLMs to analyze remote sensing imagery
about=LibreGeoLens is an experimental QGIS plugin that integrates Multimodal Large Language Models (MLLMs) to analyze remote sensing imagery.

 It enables users to interactively explore local or remote georeferenced images, extract image chips, and engage in AI-assisted conversations about their content.

 The plugin provides tools to select, extract, and track image regions, maintaining a structured record of interactions in GeoJSON format.

 Whether you're a remote sensing analyst, geospatial researcher, or GIS enthusiast, LibreGeoLens empowers you to explore how to leverage frontier AI models for geospatial imagery directly within QGIS.

 See a demo here: https://github.com/ampsight/LibreGeoLens?tab=readme-ov-file#demo

 After you install this plugin, you might get an error about external Python dependencies. Try restarting QGIS and if that does not solve the issue, you'll need to install them manually by following https://github.com/ampsight/LibreGeoLens?tab=readme-ov-file#troubleshooting-python-dependencies.
version=0.6.1
author=Ampsight
email=pedro.uria@ampsight.com

tags=python,ai,analysis
icon=resources/icons/icon.png
experimental=True

homepage=https://github.com/ampsight/LibreGeoLens
tracker=https://github.com/ampsight/LibreGeoLens/issues
repository=https://github.com/ampsight/LibreGeoLens

changelog=
      0.6.1
        * Fixed bug in some Windows machines when saving interactions with locally-loaded imagery chips as inputs
        * Documentation improvements
      0.6
        * Added LiteLLM support, unlocking 100s of MLLM services and models
        * Added support for reasoning models
        * Added code to use a non-reasoning LLM for chat summaries / titles when available
        * Added a button to cancel the "Send to MLLM" operation
        * Allow users to scroll up & down while streaming is in progress
        * Other misc UI/UX/logging improvements
      0.5
        * Replaced Groq llama models with newer LLama versions since the older ones were deprecated.
      0.4
        * Added optimizations around how chips are handled in the chats, how they are sent to the MLLM service, as well as database indexes.
        * Added option to export chats to self-contained htmls, including the chips and a `.geojson` subset of the chips' features.
        * Added option to open the local logs directory.
        * Added a lot of help / info buttons and tooltips, including clarifications about screen vs raw chips and MLLM services resolution limits.
        * Now showing original and downsampled resolution next to each chip in the chat.
        * Added warning when trying to extract large raw chips.
        * Improved chat list with keyboard navigation and better spacing.
        * Fixed chat deletion bug and added logic to still have the option to keep and use features/chips that only belong to the deleted chat.
      0.3
        * Added option to delete chats.
      0.2
        * Fixed a bug when extracting a raw chip for the first time without having previously extracted a screen chip.