LEAF DSS Documentation¶

LEAF (Landscape Evaluation & Assessment Framework) Decision Support System is a geospatial planning platform built by IWMI for the Assam State Rural Livelihoods Mission (ASRLM). It evaluates the feasibility of agricultural and livestock interventions across Assam and plans village-level livelihood clusters.

Live: https://leaf-asrlm.in

What LEAF DSS Does¶

Geospatial visualization - Interactive Leaflet maps of block-level agriculture, water, infrastructure, livestock, and demographic indicators across Assam, with district-boundary and Protected-Area overlays.
Multi-criteria feasibility scoring - Weighted engine scoring each block against configurable variable ranges and weights to produce feasibility classes ("Very Low" to "Very High").
Cluster planning - Greedy seed-and-grow clustering (ALGO_VERSION 5) groups villages into livelihood clusters for six commodities (Dairy, Goatery, Piggery, Backyard Poultry, Duckery, Fishery). Field cadres edit assignments in the app; edits are protected from automatic regeneration.
AI recommendations - RAG pipeline combining block metrics with policy PDFs to generate citation-backed recommendations via OpenAI.

System Overview¶

┌─────────────────────────────────────────────────────────────┐
│                          Clients                              │
│   ┌──────────┐   ┌───────────┐   ┌───────────────┐           │
│   │ Browser  │   │ API Users │   │ Swagger /docs │           │
│   └────┬─────┘   └─────┬─────┘   └──────┬────────┘           │
└────────┼───────────────┼────────────────┼───────────────────┘
         │               │                │
┌────────▼───────────────▼────────────────▼───────────────────┐
│                 Flask Application (app:app)                   │
│                                                               │
│   blueprints/  (modules: pages, blocks, locations,            │
│   clusters, villages, feasibility, interventions, config,     │
│   export, levels, infrastructure, production_tool, …)         │
│                          │                                    │
│   ┌──────────────────────▼───────────────────────────────┐   │
│   │  Business logic: clustering · feasibility · rag_utils │   │
│   │                   google_sheets · db_backup           │   │
│   └───┬──────────────┬──────────────┬──────────────┬─────┘   │
│       │              │              │              │          │
│  ┌────▼─────┐  ┌─────▼──────┐  ┌────▼─────┐  ┌─────▼──────┐   │
│  │ Supabase │  │ CSV +      │  │ Google   │  │ Chroma +   │   │
│  │ Postgres │  │ Shapefiles │  │ Sheets   │  │ PDFs (RAG) │   │
│  │ (clusters│  │ (villages, │  │ overlay  │  │            │   │
│  │  + POIs) │  │  geodata)  │  │(block    │  │            │   │
│  └──────────┘  └────────────┘  │ values)  │  └────────────┘   │
│                                └──────────┘                   │
│   In-process scheduler thread: daily cluster sweep +          │
│   nightly Supabase Storage backup                             │
└───────────────────────────────────────────────────────────────┘

Coverage¶

LEAF DSS covers all 35 districts and 220 blocks of Assam. Village clustering reads from data/villages.csv (the planner's source of truth); block-level map and feasibility indicators come from a published Google Sheet overlay (block_values), falling back to the committed CSV.

Map Overlays¶

Layer	Description	Default
Feasibility Choropleth	Blocks color-coded by feasibility score	Always on
District Boundaries	Dashed outlines of Assam's districts	Always on
Protected Areas	National parks, sanctuaries, conservation reserves	Off (toggle)

Quick Links¶

Resource	Path	Description
Getting Started	-	Install, run, and make your first API call
API Reference	`/api`	Full endpoint documentation
Architecture	-	System design and module responsibilities
Deployment	-	Deploy to Render or run locally
Swagger UI	`/docs`	Interactive API explorer with try-it-out
Documentation Site	`/documentation/`	This documentation, served from the Flask app

Technology Stack¶

Layer	Technology
Backend	Python 3.11, Flask (blueprints package), Flask-CORS, Gunicorn
Database	Supabase Postgres (via `DATABASE_URL`, psycopg2, raw SQL); manual SQL migrations
Data & Geospatial	Pandas, GeoPandas (Fiona/GDAL, Shapely, PyProj), CSV + Shapefiles, EPSG:4326
External data	Google Sheets published-CSV overlay (block values, intervention config)
Background jobs	In-process daemon scheduler thread (daily cluster sweep + nightly Supabase Storage backup), Postgres advisory locks
AI/RAG	LangChain, ChromaDB, OpenAI (`gpt-4o-mini`)
API Docs	Flasgger (Swagger/OpenAPI), MkDocs Material
Deployment	Render (Gunicorn, `standard` plan), custom domain `leaf-asrlm.in`, GitHub auto-deploy

No authentication

There is no user authentication. Admin/destructive actions are gated only by a ?admin=1 query flag or X-Admin: 1 header. Do not expose write endpoints publicly without adding real auth.