Since the first release of our open-source GIS Copilot for QGIS last November, we have received significant interest, feedback, and suggestions from the geospatial community. Thank you to everyone who engaged with the earlier post and tested the tool (https://lnkd.in/e6Zs846M). The copilot allows users to perform spatial analysis in QGIS using natural language for both vector and raster data.
Led by Temitope Ezekiel Akinboyewa, our team has continued to improve the tool. I’m pleased to share that version v1.0 is now available on the official QGIS plugin page (https://lnkd.in/eGjBzS-w). In addition to improved usability, this version increases the success rate and accuracy when handling more complex spatial tasks in our testing.
Key updates include:
1. Integration of GPT-5 as the default model, with configurable reasoning levels
2. Enhanced GUI with improved interactivity, including real-time operational status updates
3. Refined prompt engineering for better geoprocessing workflow generation and map projection handling
4. Added task analysis before generating workflow
5. One-click access to task breakdowns, data overview, and tool documentation
Support for local LLMs will be added in the next version – stay tuned.
Download: https://lnkd.in/eGjBzS-w
GitHub: https://lnkd.in/eFrdmjBk
Video demos (including three new videos): https://lnkd.in/e_95YcqB
Over 100 test cases (benchmarking tasks and datasets): https://lnkd.in/eKUXWA8h
For more details on the GIS Copilot works, see our recent paper at https://lnkd.in/eFa7c3ss
