Rosetta Research Agent

Multi-language AI financial research agent for enterprise deployment. Generates structured, verifiable investment theses across 5 regional desks via a clean REST API.

What This Is

A productized enterprise API extracted from Rosetta Alpha (Canteen x Arc Agora Agents hackathon, May 2026). Five regional agents each reason in their native language using specialized frontier models, then return a structured InvestmentThesis JSON with direction, confidence, reasoning chains, and optional IPFS/on-chain verification.

This repo is the enterprise product. The hackathon demo lives at rosetta-alpha.vercel.app.

Quick Start

# Minimum viable (US + Crypto desks only)
export GROQ_API_KEY=your_key_here
docker compose up --build

# Full 5-desk deployment
export GROQ_API_KEY=xxx
export DEEPSEEK_API_KEY=xxx
export GOOGLE_API_KEY=xxx
docker compose up --build

API is live at http://localhost:8000. Try:

# Health check
curl http://localhost:8000/health

# List available desks
curl http://localhost:8000/api/v1/desks

# Generate a thesis
curl -X POST http://localhost:8000/api/v1/analyze \
  -H "Content-Type: application/json" \
  -d '{"ticker": "AAPL", "desk": "us"}'

API Endpoints

Method	Endpoint	Description
GET	/health	Health check with feature flags
GET	/api/v1/desks	List all desks and their capabilities
POST	/api/v1/analyze	Generate investment thesis
GET	/api/v1/thesis/{id}	Retrieve a previously generated thesis

POST /api/v1/analyze

Request:

{
  "ticker": "AAPL",
  "desk": "us",
  "language_override": null
}

Response: See AGENT.md for full sample output.

Deploy as Enterprise Agent

Docker (recommended)

# 1. Clone
git clone https://github.com/Mihai-Codes/rosetta-research-agent.git
cd rosetta-research-agent

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

# 3. Run
docker compose up --build -d

# 4. Verify
curl http://localhost:8000/health

Without Docker

pip install -e ".[full]"
uvicorn src.api.main:app --host 0.0.0.0 --port 8000

Environment Variables

Variable	Required	Description
GROQ_API_KEY	Yes	Enables US + Crypto desks
DEEPSEEK_API_KEY	No	Enables China desk (native Chinese reasoning)
GOOGLE_API_KEY	No	Enables EU + Japan desks (Gemini)
TUSHARE_TOKEN	No	Premium A-share data for China desk
FINANCIAL_DATASETS_API_KEY	No	SEC filings for US desk
ENABLE_IPFS	No	Set to "true" to pin traces to IPFS
ENABLE_ONCHAIN	No	Set to "true" for Arc L1 recording
PINATA_JWT	No	Required if ENABLE_IPFS=true
STORACHA_SIDECAR_URL	No	Storacha/Filecoin pinning (default: localhost:3030)

Architecture

POST /api/v1/analyze {"ticker": "AAPL", "desk": "us"}
        |
        v
  [Agent Factory] --> USAgent / ChinaAgent / EUAgent / JapanAgent / CryptoAgent
        |
        v
  [Sub-Agent Pipeline: Fundamental -> Technical -> Sentiment]
        |
        v
  [Thesis Synthesis (AdalFlow Generator)]
        |
        v
  [Optional: IPFS Pin (MultiPinner) + Arc L1 Record]
        |
        v
  AnalyzeResponse JSON

Regional Desks

Desk	Model	Reasoning Language	Data Sources
US	Groq Llama-3.3-70B	English	yfinance, Financial Datasets, Stooq
China	DeepSeek V4 Pro	Simplified Chinese	AKShare, Tushare Pro
EU	Gemini 2.5 Flash	English	yfinance, Stooq
Japan	Gemini 2.5 Flash	Japanese	yfinance, Stooq
Crypto	Groq Llama-3.3-70B	English	CoinGecko, DeFiLlama

Agentalent.ai Listing

This agent is listed on Agentalent.ai (monday.com's managed AI agent marketplace). See AGENT.md for the full capability description, pricing justification, SLA, and sample output.

Prior Work

This is a productized extraction from the hackathon project:

• Source repo: github.com/Mihai-Codes/rosetta-alpha
• Live demo: rosetta-alpha.vercel.app
• Hackathon: Canteen x Arc Agora Agents (May 11-25, 2026)
• Post-optimization score: ~9.0/10 composite judge

License

Proprietary. Contact hello@mihai.codes for enterprise licensing.