RE Alpha Engine

Institutional Deal Intelligence Agent for multifamily real estate underwriting — now an autonomous agent economy participant.

An autonomous system that ingests Offering Memorandum PDFs, extracts rent rolls, runs institutional-grade financial modeling, researches live market intelligence, scrapes deal pipelines, and produces investment memos with voice narration. Participates in the Nevermined agent economy to both purchase and sell intelligence.

What It Does

Capability	Description
PDF Extraction	Extracts rent rolls, unit data, and property details from Offering Memorandum PDFs via Reka or Claude Vision
Financial Modeling	8-step underwriting model: NOI, cap rate, DSCR, cash-on-cash, 5-year IRR with exit valuation
Scenario Analysis	Bull/base/bear projections varying rent growth, exit cap compression, and vacancy
Market Intelligence	Nevermined agent network (primary) + Tavily search (fallback) for rent growth, cap rates, comps, supply
Deal Scraping	Live deal pipeline from LoopNet and Crexi via Apify
Investor Leads	Investor/buyer lead scraping for deal distribution
Deep Research	Exa AI-powered research on economic outlook, population migration, regulatory changes
Investment Memo	Claude-generated institutional-grade memo with risk factors, upside drivers, and GO/NO-GO verdict
Negotiation Leverage	Automated detection of pricing weaknesses, vacancy risk, and below-market signals
Voice Brief	45-second spoken summary via OpenAI TTS for investment committee presentations
Knowledge Graph	Neo4j-backed graph storing properties, submarkets, financials, and market trends across analyses
Intelligence Monetization	FastAPI service exposing deal analysis to external agents via Nevermined payment verification
AWS Deployment	SAM template for serverless deployment on Lambda + API Gateway + S3

Architecture

                         Nevermined Agent Network
                        /                        \
                  Buy Intel                   Sell Intel
                  (market data)               (FastAPI API)
                       |                          |
OM PDF                 |                     External Agents
  |                    v                          |
Reka / Claude --> Rent Roll Extraction            |
  |                                               v
Normalizer --> Structured JSON              POST /api/v1/analyze
  |                                         (Nevermined payment)
Financial Engine --> NOI, IRR, Cap Rate, Scenarios
  |
Market Intelligence (Nevermined -> Tavily fallback)
  |
Apify --> Deal Scraping (LoopNet, Crexi)
  |
Exa --> Deep Market Research
  |
Claude --> Contextual Investment Memo + Negotiation Leverage
  |
OpenAI TTS --> Voice Brief
  |
Neo4j --> Knowledge Graph        S3 --> Persistent Storage
  |
Streamlit --> Interactive Dashboard (9 tabs)

Quick Start

Prerequisites

Python 3.10+
API keys: Anthropic (required), all others optional
Docker (optional, for Neo4j knowledge graph)

Setup

# Clone
git clone https://github.com/oabolade/re-alpha-engine.git
cd re-alpha-engine

# Virtual environment
python3 -m venv venv
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Configure API keys
cp .env.example .env
# Edit .env with your keys

Run

# Streamlit UI (9 tabs including Deal Pipeline + Agent Network)
streamlit run app.py

# CLI: from extracted JSON
python main.py data/sample_om.json

# CLI: full agent orchestration loop
python main.py data/sample_om.json --agent

# CLI: from PDF
python main.py path/to/offering_memo.pdf

# CLI: scrape live deals
python main.py --scrape --market "Dallas, TX"

# FastAPI monetization service
uvicorn api.monetization:app --reload

# Register in Nevermined agent network
python -m api.register_service

Neo4j (Optional)

docker run -d --name re-alpha-neo4j \
  --publish=7474:7474 --publish=7687:7687 \
  --env NEO4J_AUTH=neo4j/realpha2025 \
  neo4j:latest

AWS Deployment (Optional)

cd infra
sam build
sam deploy --guided

Project Structure

re-alpha-engine/
├── app.py                         # Streamlit UI (9 tabs)
├── main.py                        # CLI entry point (with --scrape)
├── config.py                      # API keys, default assumptions
├── requirements.txt
├── agents/
│   └── orchestrator.py            # Claude tool-use orchestrator (11 tools)
├── tools/
│   ├── pdf_extractor.py           # Reka / Claude Vision PDF extraction
│   ├── rent_roll_normalizer.py    # Rent roll validation & normalization
│   ├── financial_engine.py        # 8-step financial model + scenarios
│   ├── market_intelligence.py     # Nevermined-first, Tavily fallback
│   ├── memo_generator.py          # Claude investment memo + leverage detection
│   ├── voice_brief.py             # OpenAI TTS / ElevenLabs voice narration
│   ├── knowledge_graph.py         # Neo4j graph storage & queries
│   ├── nevermined_client.py       # Nevermined SDK — search, buy, sell intel
│   ├── deal_scraper.py            # Apify — LoopNet, Crexi scraping
│   ├── exa_research.py            # Exa AI deep research
│   └── s3_storage.py              # AWS S3 storage
├── api/
│   ├── monetization.py            # FastAPI — sell analysis to other agents
│   ├── lambda_handler.py          # Mangum adapter for AWS Lambda
│   └── register_service.py        # One-time Nevermined registration
├── infra/
│   └── template.yaml              # AWS SAM template (Lambda + API GW + S3)
├── data/
│   ├── sample_om.json             # Oakwood Terrace, Dallas (12 units)
│   ├── sample_om_2.json           # Pine Valley, Austin (8 units)
│   └── sample_om_3.json           # Magnolia Heights, Houston (20 units)
├── system_prompt.md               # Orchestrator agent system prompt
├── agent_skills.md                # Skill definitions (12 skills)
├── agent_rules.md                 # Data integrity & communication rules
├── agent_memory_schema.md         # Session memory schema
├── financial_model.md             # Financial model specification
├── reka_extraction_prompt.md      # PDF extraction prompt template
└── render.yaml                    # Render deployment config

Financial Model

The engine implements a transparent 8-step underwriting model:

Gross Annual Rent — Sum of all unit rents x 12
Effective Gross Income — Adjusted for vacancy
Operating Expenses — EGI x expense ratio (default 35%)
Net Operating Income — EGI minus OpEx
Cap Rate — NOI / purchase price
Equity Invested — Purchase price x (1 - LTV)
Debt Service & Cash Flow — Interest-only, annual cash flow
5-Year IRR — NPV-based with exit cap compression
Scenario Analysis — Bull/base/bear with variable rent growth, exit cap, vacancy

Default Assumptions

Parameter	Default
Hold Period	5 years
Rent Growth	3% annually
LTV	70%
Interest Rate	6.5%
Expense Ratio	35%
Exit Cap Compression	2%

All assumptions are adjustable via the sidebar sliders.

Environment Variables

Variable	Required	Description
`ANTHROPIC_API_KEY`	Yes	Claude API for memos, extraction fallback, voice scripts
`TAVILY_API_KEY`	No	Market intelligence research (fallback)
`OPENAI_API_KEY`	No	TTS voice briefs
`REKA_API_KEY`	No	PDF extraction (falls back to Claude Vision)
`ELEVENLABS_API_KEY`	No	Alternative TTS provider
`NEVERMINED_API_KEY`	No	Agent economy — buy/sell intelligence
`NEVERMINED_ENVIRONMENT`	No	`staging` or `production` (default: staging)
`NEVERMINED_AGENT_DID`	No	Agent DID after registration
`APIFY_API_KEY`	No	Deal scraping from LoopNet, Crexi
`EXA_API_KEY`	No	Deep market research via Exa
`AWS_S3_BUCKET`	No	S3 bucket for persistent storage
`AWS_REGION`	No	AWS region (default: us-east-1)
`MONETIZATION_API_URL`	No	Deployed API URL for Nevermined registration
`INTELLIGENCE_BUDGET_PER_QUERY`	No	Max spend per intelligence query (default: $1.00)
`NEO4J_URI`	No	Knowledge graph (e.g. `bolt://localhost:7687`)
`NEO4J_USER`	No	Neo4j username
`NEO4J_PASSWORD`	No	Neo4j password

Graceful Degradation

Every new integration checks for its API key and skips if unavailable. The existing pipeline works unchanged with no new keys set:

No NEVERMINED_API_KEY → market intel falls back to Tavily
No TAVILY_API_KEY → market intel section is empty
No APIFY_API_KEY → Deal Pipeline tab shows setup instructions
No EXA_API_KEY → deep research tool returns empty
No AWS_S3_BUCKET → storage operations return None silently
No NEO4J_URI → Knowledge Graph tab shows setup instructions

API Endpoints (FastAPI)

POST   /api/v1/analyze          Submit deal for analysis (requires Nevermined payment)
GET    /api/v1/status/{job_id}  Check job status
GET    /api/v1/result/{job_id}  Retrieve results (requires payment verification)
GET    /health                  Health check

Deployment

Render (Streamlit UI)

Deployed via render.yaml blueprint.

AWS (Monetization API)

cd infra
sam build && sam deploy --guided

Deploys: Lambda + API Gateway + S3 bucket.

Tech Stack

LLM: Claude (Anthropic) — extraction, modeling, memos
PDF Processing: Reka Flash / Claude Vision
Agent Economy: Nevermined (payments-py) — discovery, purchase, monetization
Market Research: Tavily Search API (fallback), Exa AI (deep research)
Deal Scraping: Apify (LoopNet, Crexi actors)
Voice: OpenAI TTS / ElevenLabs
Graph DB: Neo4j
API: FastAPI + Mangum (Lambda adapter)
Storage: AWS S3
UI: Streamlit (9 tabs)
Deployment: Render (UI), AWS SAM (API)

License

MIT

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

RE Alpha Engine

What It Does

Architecture

Quick Start

Prerequisites

Setup

Run

Neo4j (Optional)

AWS Deployment (Optional)

Project Structure

Financial Model

Default Assumptions

Environment Variables

Graceful Degradation

API Endpoints (FastAPI)

Deployment

Render (Streamlit UI)

AWS (Monetization API)

Tech Stack

License

About

Uh oh!

Releases

Packages

Uh oh!

Contributors

Uh oh!

Languages

Name		Name	Last commit message	Last commit date
Latest commit History 5 Commits
.streamlit		.streamlit
agents		agents
api		api
data		data
infra		infra
tools		tools
.env.example		.env.example
.gitignore		.gitignore
.mcp.json.template		.mcp.json.template
CLAUDE.md		CLAUDE.md
README.md		README.md
agent_memory_schema.md		agent_memory_schema.md
agent_rules.md		agent_rules.md
agent_skills.md		agent_skills.md
app.py		app.py
config.py		config.py
financial_model.md		financial_model.md
main.py		main.py
reka_extraction_prompt.md		reka_extraction_prompt.md
render.yaml		render.yaml
requirements.txt		requirements.txt
system_prompt.md		system_prompt.md
template.yaml		template.yaml

Folders and files

Latest commit

History

Repository files navigation

RE Alpha Engine

What It Does

Architecture

Quick Start

Prerequisites

Setup

Run

Neo4j (Optional)

AWS Deployment (Optional)

Project Structure

Financial Model

Default Assumptions

Environment Variables

Graceful Degradation

API Endpoints (FastAPI)

Deployment

Render (Streamlit UI)

AWS (Monetization API)

Tech Stack

License

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages