The Edge

Six engineering bets. Each non-trivial. Together — unforkable.

Retailopædia is not a generic model with a retail prompt. It is a domain-specific reasoning system built for one vertical in one market, and the moat is the composition of these six capabilities.

1. 01

   Adversarial council

   Seven specialist agents compose and cross-check every signal against source text, so claims that cannot be supported never publish.

2. 02

   Completeness-proven ingest

   64+ sources crawled via union-fetch (RSS + sitemap + BFS + push) with five-minute completeness oracles and explicit SLAs.

3. 03

   GraphRAG retrieval

   Multi-path retrieval (full-text + vector + entity graph) with citation validation and confidence scoring behind every answer.

4. 04

   Persistent entity graph

   Self-healing entity resolution with alias learning and hot-entity prioritisation down to a 60-second cadence.

5. 05

   Scored scheduling

   One unified scheduler with backpressure keeps coverage fair across sources under load.

6. 06

   India-retail domain model

   Multilingual extraction, statutory-filing parsers, and an 800+ Indian-retail brand taxonomy.

Build vs. buy

Reproducing this in-house runs 14–18 months and ₹6–9 Cr upfront before the first reliable signal. A Retailopædia subscription deploys in under 72 hours at ₹72L–₹1.2 Cr a year — see Pricing.

Why now

Three tailwinds converge: Indian retail consolidation is accelerating, frontier LLMs have matured enough to reason over filings, and there is still no category-native intelligence product for this market.