DocumentationFeatures
Overview of risk categories (Legal, Visual, Compliance, etc.) and the Map-Reduce AI workflow used for contract analysis.
This page describes the features of the AI Property Contract Review system: what risks are detected, how analysis is structured, and what the AI looks for. It aligns with the system’s design (see the project’s REQUIREMENTS.md for full specification).
Risk categories
The system groups risks into five high-level categories. Each risk in the report has a category (e.g. Legal, Visual, Compliance), severity, likelihood, impact, and recommendations.
1. Legal & Title Search
| Risk | Description | How it’s checked |
|---|---|---|
| Easements | Underground sewer/drainage affecting building rights | Sewer Service Diagram + contract text |
| Covenants | Land restrictions (e.g. single-storey, brick exterior) | Title Search |
| Caveats | Third-party claims on the property | Title Search |
| Leases | Existing tenancy affecting vacant possession | Contract terms |
| Special Conditions | Penalty clauses, unfair terms | Contract text analysis |
2. Physical Condition
| Risk | Description | How it’s checked |
|---|---|---|
| Termites/Pests | Pest damage to structures | Pest inspection / certificates |
| Structural Defects | Foundation, walls, roof issues | Building inspection |
| Plumbing/Leaks | Waterproofing, roof leaks | Moisture / reports |
| Electrical/Gas | Old wiring, switchboard, gas | Certificates / reports |
| Asbestos | Pre-1990 buildings may contain asbestos | Building report |
3. Compliance & Certificates
| Risk | Description | How it’s checked |
|---|---|---|
| Illegal Renovations | Unapproved structural changes | Council plans vs actual |
| Pool Compliance | Safety fence requirements | Certificate of Compliance |
| Smoke Alarms | State-specific requirements | Compliance section |
| Occupancy Permit | Building approval for habitation | Council records / certificates |
| Energy Rating | Required disclosure in some states | Certificate |
4. Land & Environment
| Risk | Description | How it’s checked |
|---|---|---|
| Flood Zone | Insurance cost, uninsurability | Planning Certificate |
| Bushfire Zone | Building standards, rebuild cost | Planning Certificate |
| Land Contamination | Former industrial use | Environmental reports |
| Heritage Overlay | Restrictions on modifications | Planning Certificate |
5. Contract clause traps
| Clause | Risk | Recommendation |
|---|---|---|
| Subject to Finance | Loan rejection → deposit loss | Always include |
| Subject to Building & Pest | Major issues found after signing | Include for private sales |
| Sunset Clause | Developer delays (off-plan) | Negotiate reasonable terms |
| As Is | No recourse for defects | Do thorough pre-purchase inspection |
The AI focuses on Australian property contracts (e.g. Section 32, contract of sale) and uses a knowledge base (RAG) for local law and practice where configured.
Multi-agent workflow (Map-Reduce)
Analysis does not use a single “one-shot” AI call. It uses a Map-Reduce style workflow for better accuracy and clearer structure.
Step 1: Indexer agent
- Input: Full contract PDF.
- Job: Find page ranges for key sections:
- Special Conditions
- Title Search (Register Search Statement)
- Plan of Subdivision
- Sewer Service Diagram
- Compliance Certificates (pool, pest, building, energy, etc.)
- Output: A map of section names to page numbers (and basic validation: readable, is it a contract).
Step 2: Specialized agents (parallel)
Three agents run in parallel, each with a focused prompt and inputs:
-
Legal agent
- Input: Special Conditions + Title Search text.
- Focus: Penalties, caveats, covenants, unfair terms, clause traps.
- Output: Legal risks with clause references and recommendations.
-
Visual agent
- Input: Plan of Subdivision + Sewer Diagram (as images).
- Focus: Easements, pipe locations, boundaries, title type (e.g. Strata vs Torrens).
- Uses: Gemini Vision for diagram analysis.
- Output: Visual/spatial risks.
-
Compliance agent
- Input: Certificates section.
- Focus: Missing or inconsistent certificates (pool, smoke alarms, building, energy, etc.).
- Output: Compliance risks and missing items.
Step 3: Synthesizer agent
- Input: All agent outputs.
- Job: Deduplicate, prioritize by severity, and categorize.
- Output: Final risk report (e.g. JSON) that is stored and shown to you.
This design gives:
- Accuracy: Each agent has a narrow task (no single “do everything” prompt).
- Visual analysis: Diagrams are handled by a dedicated vision agent.
- Structured output: Consistent JSON and categories for filtering and display.
What you can do in the product
- Upload PDF property contracts and see processing status.
- Trigger analysis and later view structured risks by Legal, Visual, and Compliance (and underlying categories above).
- Filter risks by severity and category.
- Ask follow-up questions on a specific risk and view conversation history.
- Manage credits: Signup gives 5,000 credits (~5 contract reviews); you can buy more from the Pricing page (e.g. Basic $9.90 for 25,000 credits).
Summary
- Risk categories: Legal & Title, Physical Condition, Compliance & Certificates, Land & Environment, and contract clause traps.
- Workflow: Indexer → Legal + Visual + Compliance (parallel) → Synthesizer → final risk report.
- Credits: Used for contract analysis (~750–1,000 per contract) and Q&A (~150 per question); see Credits and Pricing.
For a step-by-step usage guide, see Overview.