Platform Features
Everything needed to move from documents and data to accountable policy insight.
Knowledge for a Sustainable India
Research repository & dataset catalogue
- Versioned uploads with checksum, MIME sniffing and malware-scan hook
- Asynchronous processing: validation → extraction → OCR → chunking → embedding → indexing
- Draft → review → approval → publication workflow with reviewer comments
- Tabular profiling, observation ingestion linked to administrative units
- Provenance, licensing, quality notes and related-document graph
Hybrid search
- Keyword search (OpenSearch or PostgreSQL full-text) fused with pgvector semantic search using reciprocal rank fusion
- Faceted filters: type, topic, geography, year, organisation, language, source
- Autocomplete, saved searches and search history
- Relevance explanation per result (keyword rank, vector rank, matched-by)
Evidence-based AI assistant
- Query analysis → hybrid retrieval → re-ranking → evidence selection → citation-only generation
- Refuses to answer without sufficient evidence; never fabricates sources
- Confidence score with stated basis and limitations
- Every answer stored as a recommendation with evidence items and a provenance graph
GIS & remote sensing
- Administrative hierarchy (country → state → district → sub-district → village) with real PostGIS geometry
- Thematic layers with legends, temporal validity and per-unit statistics
- Operations: intersection, geodesic area, proximity, aggregation, hotspot, boundary select, affected-area buffer
- STAC connector for Sentinel/Landsat scenes; NDVI/NDWI/NDBI, rule-based land cover and change detection
Analytics engine
- Descriptive statistics, linear trend, geographic comparison, correlation, anomaly detection, forecasting with intervals, composite risk scores, time-series profiles and spatial roll-ups
- Model registry with algorithms, parameters, evaluation notes and limitations
- Every run produces a recommendation with dataset-version evidence
Policy simulation lab
- Scenario baselines derived from observed land-cover layers or declared assumptions
- Interventions: reallocation, yield improvement, irrigation expansion, rainfall shifts
- Monte Carlo propagation of coefficient uncertainty with p10–p90 bands
- Outputs labelled OBSERVED / MODELLED / ASSUMPTION / SCENARIO / UNCERTAINTY, risk flags, scenario comparison and policy-maker approval
Workspaces & innovation portal
- Workspaces with members, linked resources, tasks, comments, activity and a review-to-publish lifecycle
- Challenges with structured evaluation criteria, weighted scoring, shortlisting and pilot tracking
Security, administration & integrations
- JWT access tokens, rotating refresh tokens with reuse detection, lockout, password policy
- Thirteen roles, ~45 permissions, approvals queue, user lifecycle and system settings
- Append-only audit log with request IDs; security headers; rate limiting
- Connectors: geoBoundaries, data.gov.in, STAC imagery, OpenSearch, S3/MinIO, OpenAI-compatible LLMs, SMTP