/ 01 — Competitive strategy

Different axis from bengaluru.rent. On purpose.

They own real rent data and zero friction. We own reasoning and memory. The strategy is to differentiate hard as the decision-support layer, not to clone a map we'd lose at.

/ 02 — Competitor

bengaluru.rent, verified live

Anonymous crowdsourced rent map. No login. Google Maps JS API. AdSense-funded. Stats captured 2026-06-09.

4,764
real rents pinned
₹235 Cr+
value on map
2,285
matches made
114
owner-direct flats
449
active hunters
931
watchlisted areas
97k+
unique visitors
2.5 km
seeker match radius

Their gaps: no AQI, no school data, no commute math, no reasoning, no accounts, no API. Data quality leans on the crowd; long-tail coverage is sparse.

/ 03 — Gap analysis

Comparison matrix

Where they lead, where we lead, and the priority of closing each gap.

Category bengaluru.rent Ooru (current) Opportunity Priority
Discovery Map + pin drop + 2.5km match Landmark → LangGraph → LLM report Add spatial view of ranked localities High
Rent data 4,764 real rents + 114 owner flats None (Apify broken) Lightweight price signal; be honest on confidence Critical
Transport Namma Metro overlay + "Near Metro" filter Foursquare metro, discarded Wire into state + report + ranking High
Schools None ICSE via Foursquare, fabricated in report Real data → core ranking signal (their blind spot) Critical
Environmental Green cover (Sentinel-2, qualitative) AQI via Open-Meteo (real numbers) Fuse AQI + green into a livability score Medium
Intelligence None (raw data + filters) Kimi LLM advisory (unique) Protect & enhance — the core moat Critical
Personalization Anonymous; email watchlist only Auth + HouseRental history Personal house-hunting memory + compare Medium
Visualization Rich Google Maps UX (core) Text / JSON / CLI Mapbox GL view (reuse existing token) High
Matching Seeker↔lister email match (2,285) None Out of scope for an analysis tool Low
Trust Real pins + crowd fact-check Fabricated claims in report Fix data wiring — non-negotiable Critical
Friction Zero (no login), quick tour Registration; sync blocking /chat Anonymous demo + non-blocking /chat Medium

Their strength is data volume. Ours is intelligence and trust. Racing them on the map is the one strategy that loses.

/ 04 — Sequencing

Roadmap — three horizons

Restore honesty, then reach parity on visualization, then build the moat they can't. Horizon 0 was scoped as an in-place v1 repair; the chosen path is the v2 PocketBase rewrite (see Architecture §04 / MIGRATION.md), which resolves 0.1 and 0.2 by construction — read these as "what's broken + what to build," delivered via v2.

Horizon 0 Trust & hygiene

Do first. Nothing new ships before these.

# Item Effort Risk
0.1 Wire schools + metro into state, prompt & persistence — the honesty fix. main.py · workflow.py · services.py · chatbot.py M low
0.2 JWT secret from env, no hardcoded default; rotate. services.py:22 S low
0.3 .gitignore + untrack .env / __pycache__ / skill dirs; rotate leaked keys S low
0.4 Delete dead code — schemas.py, get_travel_times, unused context var S low
0.5 Request timeouts on all six requests.get() calls. main.py S low
0.6 Pin requirements.txt versions for reproducible builds S low
0.7 Non-blocking /chat or progress (LangGraph streaming / async) M med

Horizon 1 Parity + differentiation

2–4 weeks. Catch up on visualization, pull ahead on reasoning.

# Item Effort Impact
1.1 Fix / replace rentals path — dynamic Apify URL or public price ranges M high
1.2 Map view of ranked localities via Mapbox GL (reuse existing token) M high
1.3 Enrich LLM prompt with real structured data — school count/dist, metro, per-locality AQI, rent range S high
1.4 Comparison endpoint for 2+ saved searches S med

Horizon 2 The moat

What an anonymous map structurally can't do.

# Item Effort Impact
2.1 Multi-signal livability score — AQI + school dist + metro time + green-cover proxy M high
2.2 "What-if" commute modeling — revive dead get_travel_times as a real node M high
2.3 Personal knowledge base — notes per locality, saved reports M med
2.4 Opt-in crowdsourced rent signals (authenticated) — only after trust restored L med

Single highest-leverage step: 0.1. It kills the fabrication problem and turns schools + metro into honest differentiators in one move.

/ 05 — Read next