How musashi-99/sih works
SatQuery AI is a web-based satellite-image analysis assistant. A user uploads one image or a pair and asks a natural-language question; the server validates and analyzes the imagery, plans and checks tools for the request, then returns a measured answer with confidence and evidence, optionally rendered as an overlay. The README also describes optical–SAR fusion, audit reports, and reproducible run history. Some app and core source was only sampled, so the graph omits uncertain internal wiring rather than inferring a pipeline for it.
Main entry points: app.py Web API · index.html Workspace UI · ingestion.py Image Ingestion · validation.py Input Validation
Components 17
User Interface
- Workspace UI
static/index.html - Web API
app.py
Image Analysis
- Image Ingestion
core/ingestion.py - Input Validation
core/validation.py - Sensor Mapping
core/sensors.py - Feature Extraction
core/features.py - Land Cover
core/landcover.py - Object Detection
core/objects.py - Change Detection
core/change.py - SAR Fusion
core/sar.py
Query Orchestration
- Query Planner and Answerer
core/query.py - Tool Registry
core/registry.py - Overlay Renderer
core/render.py
Evidence and Audit
- Evidence Graph
core/evidence.py - Audit Reports
core/report.py - Run History
core/history.py
Other components
- Analyst
Connections 17
- Analyst → Workspace UI: asks and uploads
- Workspace UI → Web API: sends requests
- Web API → Image Ingestion: ingests imagery
- Web API → Input Validation: validates inputs
- Image Ingestion → Sensor Mapping: resolves bands
- Web API → Feature Extraction: extracts features
- Web API → Land Cover: segments scene
- Web API → Object Detection: detects objects
- Web API → Change Detection: compares scenes
- Web API → SAR Fusion: processes SAR
- Web API → Query Planner and Answerer: plans and answers
- Web API → Tool Registry: checks tools
- Web API → Overlay Renderer: renders overlays
- Web API → Evidence Graph: records evidence
- Web API → Audit Reports: generates reports
- Web API → Run History: records runs
- Web API → Workspace UI: returns answers