Junyr integrates French public APIs (DVF, DPE, Cadastre) for automated property analysis and reliable valuations.

AI for Real Estate: Property Analysis, Cadastre, and Valuation with Junyr

The real estate industry generates massive volumes of data: online listings, DVF transactions, energy diagnostics, cadastral records, geolocation data. Junyr offers a dedicated sector module to help real estate professionals harness this data through specialized AI agents.

---

Why a Dedicated Sector Module for Real Estate?

Generic tools do not understand the specifics of the French property market:

• DPE and GES: Energy performance labels directly influence prices (up to -18% for a G-rated property)
• DVF: The Demandes de Valeurs Foncieres (public property transaction records) are a goldmine, but difficult to exploit manually
• Cadastre: Parcel and municipal data is scattered across multiple APIs
• Multiple portals: SeLoger, Leboncoin, BienIci, PAP -- each portal has its own formats and filters

Junyr centralizes these sources into a unified pipeline, accessible by your AI agents.

---

Sector Architecture: DDD Isolation

The real estate module is isolated using the Domain-Driven Design pattern:

backend/sectors/real_estate/
  __init__.py              -- Sector skeleton + get_agent_tools()
  services/
    analysis_service.py    -- Property analysis by URL and images (Gemini Vision)
    cadastre_service.py    -- DVF, DPE, cadastral data (public APIs)
    valuation_service.py   -- Price estimation through market comparison
    geocoding_service.py   -- Geocoding, geographic enrichment
    search_url_builder.py  -- Multi-portal search URL generation

The real estate code never imports from other sectors. The core ERP runs without the sector loaded. Auto-discovery via sector_registry.py loads the module at startup.

---

Property Analysis (Planned)

The AnalysisService extracts structured data from 8 French property portals (SeLoger, Leboncoin, PAP, BienIci, Logic-Immo, FNAIM, Orpi, Century 21).

URL-Based Analysis

1. Fetch the listing page via ScrapingBee (JavaScript rendering, premium proxy)
2. Extract structured data via Gemini (price, surface area, DPE, amenities)
3. Enrich automatically with geographic data (geocoding, neighborhood DVF transactions)
4. Save to database for history and comparison

Extracted data includes: property type, price, surface area, rooms, full location, energy performance (DPE/GES), detected amenities, contact information, and a confidence score (0 to 1).

Image-Based Analysis

The visual analysis service uses Gemini Vision to evaluate a property from photos:

• Room identification and condition assessment (score from 1 to 10)
• Amenity detection: flooring, windows, heating, storage
• Renovation estimate: estimated cost and priority repairs
• Market positioning: target buyer profile and selling points

Property Comparison

Compare up to 5 properties with a comparative table, ranking by criteria (value for money, location, potential) and an overall recommendation with confidence score.

---

Cadastre and Open Data Integration

The CadastreService leverages free French public APIs:

DVF Market Statistics

From DVF transactions, the service computes median price, average price, price per square meter, transaction volume, and min/max range -- over the last 3 years by default.

DPE and Municipal Data

The service queries the ADEME API for DPE/GES labels, energy consumption, and zone statistics (label distribution by postal code). Via the Geo API: INSEE code, population, area, and geographic boundaries.

---

Price Estimation

The ValuationService produces a well-reasoned estimate in 7 steps: geocoding, DVF retrieval, DPE context, base price calculation (median price per sqm x surface area), adjustments, confidence range, and confidence level.

DPE Label Adjustments

Other Factors

The service also identifies comparable properties (similarity score based on surface area, distance, and rooms) and computes market trends over 5 years.

---

Geocoding and Geographic Enrichment

The GeocodingService provides forward and reverse geocoding (Google Maps API), with database caching to reduce costs.

For each property, geographic enrichment analyzes the surrounding area (1 km radius):

An estimated walkability score (0-100) is calculated from the detected amenities.

---

Search URL Generator

The SearchURLBuilder generates optimized search URLs for Leboncoin, SeLoger, and BienIci with 50+ criteria (outdoor features, view, interior, kitchen, heating, etc.), automatic budget expansion (50% below to 25% above), sort by date, and 45+ major cities pre-configured.

---

Integration with the Junyr ERP (Planned)

Planned Pipeline

Listing detected (portal or email)
   -> Automatic analysis (URL + images)
   -> Prospect created (Commercial module)
   -> Valuation generated (DVF + DPE + geo)
   -> Project created (Production module)
   -> Follow-ups and reminders (Commercial module)
   -> Transaction / Invoicing (Finances module)

---

The Role of Junyrs (AI Agents)

Once the module is finalized, a Junyr configured for the real estate sector will be able to:

1. Receive an email from a prospect ("I'm looking for a 3-bedroom apartment in Lyon, budget 300k, DPE C minimum")
2. Generate search URLs on Leboncoin, SeLoger, and BienIci
3. Analyze the listings found (structured data extraction)
4. Estimate prices (DVF comparison, DPE and amenity adjustments)
5. Compare properties and rank by relevance
6. Respond to the prospect with a well-reasoned selection

Planned AI Tools

These tools will be dynamically injected into the agent context via sector_registry.py, only for companies configured in the immobilier (real estate) sector.

---

Current Status and Roadmap

What Exists (Backend)

What Is Missing

Roadmap

1. Phase 1: Expose services via a sector router (/api/sectors/real_estate/)
2. Phase 2: Define LLM tools for injection into AI agents
3. Phase 3: Frontend interface (ERP tabs, map view, dashboard)
4. Phase 4: Email-to-task integration (automatic analysis of received listings)

---

Early Adopter Program

We are looking for real estate professionals to test the module in beta:

• Validate use cases: your needs will guide development priorities
• Test valuations: compare with your local market knowledge
• Configure workflows: define the business processes to automate
• Preferential pricing: access to the sector module included during beta

What You Get Today

Even without the real estate module, Junyr offers professionals:

• Full 8-module ERP (commercial, production, finances, HR, and more)
• AI agents (Junyrs) to automate emails, tasks, and follow-ups
• Universal workspace to centralize emails, documents, and datasets
• Sales pipeline to track prospects and mandates
• Invoicing with PDF generation and payment tracking

---

Summary

Junyr's real estate module rests on solid foundations -- five complete backend services representing over 2,600 lines of code -- powered by French public data. The goal is to transform these services into tools accessible by your AI agents, to automate property analysis, valuation, and tracking.

Interested in the early adopter program? Contact us to learn more.