免费在 GitHub 公开

geoai

Route GeoAI geospatial data, inference, training, foundation-model, and integration workflows.

中文简介正在准备中。当前显示原文说明。

VectorSpaceLab/AREX-Skill

查看安装方法

包含的文件（34）

SKILL.md4.0 KB
references/installation-and-environment.md2.7 KB
references/repo-provenance.md1.7 KB
references/repo-routing-metadata.json482 B
references/top-level-api-map.md4.4 KB
references/troubleshooting.md4.4 KB
scripts/check_geoai_env.py6.7 KB
sub-skills/detection-segmentation-inference/references/api-reference.md14.0 KB
sub-skills/detection-segmentation-inference/references/inference-workflows.md7.9 KB
sub-skills/detection-segmentation-inference/references/troubleshooting.md5.4 KB
sub-skills/detection-segmentation-inference/scripts/inference_input_check.py18.4 KB
sub-skills/detection-segmentation-inference/SKILL.md2.4 KB
sub-skills/foundation-models-embeddings-vlms/references/api-reference.md15.3 KB
sub-skills/foundation-models-embeddings-vlms/references/model-and-embedding-workflows.md11.4 KB
sub-skills/foundation-models-embeddings-vlms/references/troubleshooting.md5.6 KB
sub-skills/foundation-models-embeddings-vlms/scripts/list_geoai_models.py9.8 KB
sub-skills/foundation-models-embeddings-vlms/SKILL.md1.9 KB
sub-skills/geospatial-data-pipelines/references/api-reference.md18.7 KB
sub-skills/geospatial-data-pipelines/references/data-and-pipeline-workflows.md10.0 KB
sub-skills/geospatial-data-pipelines/references/troubleshooting.md10.1 KB
sub-skills/geospatial-data-pipelines/scripts/geospatial_io_smoke.py9.6 KB
sub-skills/geospatial-data-pipelines/scripts/validate_pipeline_config.py9.3 KB
sub-skills/geospatial-data-pipelines/SKILL.md4.0 KB
sub-skills/integrations-agents-qgis-mcp/references/api-reference.md7.4 KB
sub-skills/integrations-agents-qgis-mcp/references/qgis-plugin-and-mcp.md5.1 KB
sub-skills/integrations-agents-qgis-mcp/references/troubleshooting.md6.2 KB
sub-skills/integrations-agents-qgis-mcp/scripts/mcp_config_check.py17.8 KB
sub-skills/integrations-agents-qgis-mcp/scripts/qgis_dependency_probe.py9.2 KB
sub-skills/integrations-agents-qgis-mcp/SKILL.md3.2 KB
sub-skills/training-and-finetuning/references/api-reference.md22.0 KB
sub-skills/training-and-finetuning/references/training-recipes.md11.7 KB
sub-skills/training-and-finetuning/references/troubleshooting.md9.9 KB
sub-skills/training-and-finetuning/scripts/check_training_layout.py25.0 KB
sub-skills/training-and-finetuning/SKILL.md2.3 KB

SKILL.md（原文）

安装前可以查看交给 AI 代理的指令内容。

GeoAI

Use this repo skill when the task is about GeoAI: geospatial data inspection and downloads, raster/vector conversion, batch pipelines, model inference, training and fine-tuning, foundation models and embeddings, or the QGIS/MCP/agent integration surfaces.

GeoAI has a broad package surface. Start with the smallest sub-skill that matches the user's intent, then read the linked references for the exact APIs, commands, and failure modes.

First steps

Read installation and environment to choose the right install path and optional extras.
Run scripts/check_geoai_env.py for a safe import/CLI/backend smoke check when you need to confirm the environment before deeper work.
Read top-level API map when you need to see which module family owns a symbol.
Read troubleshooting for cross-cutting install/import/CLI/backend issues.
Read repository provenance if you need to confirm whether this skill still matches the checkout you have.

Route by task

User request	Read next
Inspect raster/vector files, validate CRS/bounds, download STAC/NAIP/Overture data, run `geoai pipeline`, tile or vectorize data	Geospatial data pipelines
Run segmentation, detection, prompt-based segmentation, RF-DETR, water/cloud/super-resolution, ONNX, or HF-style inference	Detection and segmentation inference
Prepare training layouts, fit/fine-tune models, inspect losses or metrics, or publish checkpoints carefully	Training and finetuning
Query the foundation-model registry, work with embeddings, DINOv3, Prithvi, UniverSat, TESSERA, Moondream, vLLM, or captioning	Foundation models, embeddings, and VLMs
Configure the QGIS plugin, the GeoAI MCP server, or optional Strands agent integrations	GeoAI integrations, agents, QGIS, and MCP

Package identity

Distribution name: geoai-py
Import name: geoai
Console entry point: geoai
Public Python support: Python 3.12 or newer

Safe install pattern

For normal use, install the published package: pip install geoai-py.
When working from a local checkout, an editable install is reasonable: pip install -e ..
Install optional extras only when the task needs them. The main extras currently include agents, building, networks, onnx, osd, rfdetr, sr, terratorch, and vllm.
Do not install optional extras just because they exist; route to the sub-skill that owns the workflow first.

Minimal smoke check

A quick package smoke check should be enough before deeper work:

python -c "import geoai; print(geoai.__version__)"
python -m geoai.cli --help

If the user needs backend confirmation for GPU workflows, use the bundled smoke script rather than jumping straight to a heavy model run.

What this root skill does not do

It does not perform downstream geospatial analysis or model execution by itself.
It does not tell future agents to open files from the original repository checkout when a bundled skill reference or script exists.
It does not bundle training, inference, or integration details that belong in the sub-skills.

Follow-up references

Use repository provenance to check staleness before refresh.
Use top-level API map to trace a public symbol to the owning sub-skill.
Use troubleshooting for cross-cutting error recovery before reading the more specialized sub-skill troubleshooting file.

评价

还没有评价。欢迎分享你的使用感受。

无需登录即可发表。欢迎分享使用感受、提问，或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

同一仓库的技能

3d-resnets-pytorch

免费

Routes 3D ResNets PyTorch video action-recognition workflows across training, inference, and data preparation.

中文简介正在准备中。当前显示原文说明。