技能创建器
skill-creator
创建、改进和评估自定义技能,把专业流程转化为可重复使用的 AI 能力。
streamlit-geospatial【数据与研究】Guides researchers through the streamlit-geospatial multipage application, including Streamlit composition, Leafmap/Folium maps, remote Earth Engine workflows, and data-backed geospatial dashboards.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Use this repo skill when a task asks to build, adapt, troubleshoot, or deploy a
Streamlit geospatial application using the workflow patterns demonstrated by
streamlit-geospatial. It is a recipe and routing graph, not a requirement to
keep the original checkout available.
When a request combines routes, validate local inputs first, then cross the external Earth Engine or remote-data boundary only after the caller explicitly provides credentials/network access. Keep each sub-skill's ownership intact.
This repository is a Streamlit application rather than an installable Python
distribution. Use Python 3.11 as the conservative inspection/runtime choice,
install the documented Python requirements, and provide the system geospatial
libraries and media tools listed in packages.txt when deploying. A typical
runtime install is:
python -m pip install -r requirements.txt
The repository uses geemap.foliumap and geemap.foliumap.Map. Set the backend
selector before importing that module:
export USE_FOLIUM=1
python -c "import streamlit, leafmap, geemap, geopandas; print('imports ok')"
Run the environment checker for a deterministic import/version preflight; it does not start Streamlit, authenticate Earth Engine, fetch tiles, or download data. Read the installation reference before using the deployment files because the original setup recipe writes a user Streamlit config and assumes host-level package installation.
The app entry point is Home.py; numbered files under pages/ become sidebar
pages automatically. Keep each page importable and call st.set_page_config
before creating the page UI. Use st.cache_data only around deterministic
fetch/parse functions, keep remote URLs and credentials outside source, and
render Folium/Kepler maps through their Streamlit adapter or PyDeck maps with
st.pydeck_chart.
For public runtime use, preserve attribution and validate remote endpoints. Treat WMS/XYZ/COG/housing sources and Earth Engine assets as mutable external services. Do not claim a remote layer, asset, or timelapse exists without a successful bounded request and an output check.
--help and tiny fixture
checks before starting a server.refresh-repo-skill.