Trellis

Trellis is an AI-augmented pricing platform for quantitative finance. Ask for a price in natural language, work directly from Python when you need control, and drop down into sessions, payoffs, and numerical methods when the workflow calls for it.

Trellis currently has three public faces:

• trellis.ask(...) and Session.ask(...) for agent-first pricing
• package-level Python APIs for reproducible library workflows
• trellis-ui as an experimental companion interface

Installation

pip install trellis

# Optional runtime dependencies are installed separately today
pip install openai      # or: pip install anthropic
pip install requests fredapi

Install external comparison libraries such as QuantLib or FinancePy separately when you need cross-validation coverage.

Quick Examples

Ask for a price

import trellis

# Requires an installed provider client plus OPENAI_API_KEY or ANTHROPIC_API_KEY
result = trellis.ask("Price a 5Y SOFR cap at 4% on $10M")
print(result.price)
print(result.payoff_class)
print(result.matched_existing)

Work offline with deterministic sample data

import trellis

s = trellis.quickstart()
bond = trellis.sample_bond_10y()
result = s.price(bond)

print(result.clean_price)
print(result.greeks["dv01"])

Documentation

• Start with docs/quickstart.rst
• Follow the tutorials in docs/tutorials/index.rst
• Use the high-level workflows in docs/user_guide
• Use docs/quant/index.rst for pricing constructs, extension patterns, and knowledge maintenance
• Use docs/developer/index.rst for hosting, agents, traces, and task/eval operations
• Treat experimental features accordingly:
  • agent-built payoffs
  • trellis-ui
  • live market-data auto-resolution beyond the mock provider

Documentation Policy

The public docs are maintained around the package-level surface first. Examples default to reproducible mock data unless a section is explicitly marked as live-data or API-key dependent.