hermes-financial 0.1.3

Multi-agent financial research framework

Hermes Financial

Multi-agent financial research framework built on LlamaIndex.

Hermes gives AI engineers pre-built financial data tools, document ingestion pipelines, output generation (Excel models, Word/PDF reports), and composable agents for equity research. Use the agents out of the box, swap in your own, or extend with custom tools and data sources.

Quick Start

# Install with uv (recommended)
uv add hermes-financial

# Or with pip
pip install hermes-financial

import asyncio
import hermes

hermes.configure(
    llm_provider="anthropic",
    sec_user_agent="MyApp/1.0 ([e-mail])",  # required by SEC
)

h = hermes.Hermes()

# List available tools and agents
print(h.list_tools())
print(h.list_agents())

What's Included

Data Tools (35 total)

Module	Tools	Source
SEC EDGAR	Company facts (XBRL), filing search, submissions, filing URLs, structured financial tables, labeled text sections (MD&A, Risk Factors, etc.), insider transactions, institutional holdings	edgartools + efts.sec.gov — free, no key
FRED	Economic time series, series search, metadata	api.stlouisfed.org — free key
Market Data	Quotes, historical OHLCV, batch quotes	Yahoo Finance — free, no key
News	Company news, general financial news	Yahoo/Google RSS — free, no key
Excel	Create, write, read, format, formulas, charts, save .xlsx	openpyxl
Documents	Create, headings, paragraphs, tables, images, save .docx, export PDF	python-docx + LibreOffice
Charts	Line, bar, waterfall, scatter, heatmap → PNG	matplotlib

Specialist Agents (7)

Agent	Type	Role
SecFilingsAgent	FunctionAgent	SEC filing retrieval and analysis
MacroAgent	FunctionAgent	FRED macroeconomic data
MarketDataAgent	FunctionAgent	Quotes and historical prices
NewsAgent	FunctionAgent	Financial news and sentiment
ModelingAgent	ReActAgent	Excel financial model construction
ReportAgent	ReActAgent	Word/PDF report generation
ResearchOrchestrator	ReActAgent	Multi-agent coordination

Infrastructure

• File-based cache with per-item TTL (filings cached permanently, quotes never cached)
• Async rate limiter (token bucket) for SEC EDGAR, FRED, Yahoo Finance
• Streaming events for real-time progress reporting
• ChromaDB vector indices for semantic search over financial documents

Configuration

All settings can be passed to configure() or set via environment variables with the HERMES_ prefix:

hermes.configure(
    llm_provider="anthropic",           # or "openai"
    llm_model="claude-sonnet-4-5-20250514",
    sec_user_agent="MyApp/1.0 ([e-mail])",
    fred_api_key="your_key",
    output_dir="./output",
    verbose=True,
)

Environment variables: HERMES_LLM_PROVIDER, HERMES_FRED_API_KEY, etc.

Extending Hermes

Register a Custom Tool

from llama_index.core.tools import FunctionTool

def my_bloomberg_lookup(ticker: str) -> dict:
    """Look up data from Bloomberg."""
    ...

tool = FunctionTool.from_defaults(fn=my_bloomberg_lookup, name="bbg_lookup")

h = hermes.Hermes()
h.register_tool("bbg_lookup", tool, tags=["market_data"])

Register a Custom Agent

from hermes.agents import HermesAgent

class OptionsFlowAgent(HermesAgent):
    name = "options_flow"
    description = "Analyzes unusual options activity and flow data."
    system_prompt = "You are an options flow specialist..."
    agent_type = "function"

    def get_tools(self):
        return [...]

h = hermes.Hermes()
h.register_agent("options_flow", OptionsFlowAgent)

Using Tools Directly

You don't need agents to use the tools. Every tool module exposes async functions that work standalone:

from hermes.tools.sec_edgar import (
    get_company_facts,
    get_filing_urls,
    get_filing_financial_tables,
    get_filing_text,
)
from hermes.tools.fred import get_series
from hermes.tools.excel import excel_create_workbook, excel_write_cells, excel_save

# Fetch Apple's multi-period XBRL financial data
facts = await get_company_facts("AAPL")

# Discover recent 10-K and 10-Q filings
filings = await get_filing_urls("AAPL", filing_types="10-K,10-Q", limit=10)

# Extract XBRL-parsed, pre-classified financial statements from a filing
tables = await get_filing_financial_tables("AAPL", filings[0]["accessionNumber"])

# Extract labeled qualitative sections (MD&A, Risk Factors, Business, etc.)
text = await get_filing_text(filings[0]["url"])

# Pull GDP data from FRED
gdp = await get_series("GDP", start_date="2020-01-01")

# Build an Excel workbook
wb_id = excel_create_workbook("model", sheets=["Assumptions", "Income Statement"])
excel_write_cells(wb_id, "Assumptions", {"A1": "Revenue Growth", "B1": 0.12})
filepath = excel_save(wb_id, "my_model.xlsx")

Project Structure

hermes/
├── __init__.py              # Public API: Hermes, configure
├── config.py                # Pydantic configuration
├── registry.py              # Tool/agent registry
├── core.py                  # Hermes facade class
├── tools/
│   ├── _base.py             # HTTP client, rate limiter, cache helpers
│   ├── sec_edgar.py         # SEC EDGAR (XBRL, EFTS, submissions)
│   ├── fred.py              # FRED macroeconomic data
│   ├── market_data.py       # Yahoo Finance quotes and history
│   ├── news.py              # RSS-based news retrieval
│   ├── excel.py             # Excel workbook manipulation
│   ├── documents.py         # Word document generation + PDF export
│   └── charts.py            # matplotlib chart generation
├── agents/
│   ├── base.py              # HermesAgent abstract base class
│   ├── sec_filings.py       # SEC filings specialist
│   ├── macro.py             # Macroeconomic data specialist
│   ├── market.py            # Market data specialist
│   ├── news.py              # News/sentiment specialist
│   ├── modeling.py          # Financial modeling (Excel)
│   ├── report.py            # Report writing (Word/PDF)
│   └── orchestrator.py      # Multi-agent coordinator
├── ingestion/
│   ├── sec_parser.py        # SEC filing HTML → LlamaIndex nodes
│   ├── transcript_parser.py # Earnings transcript parser
│   └── index_manager.py     # ChromaDB index management
└── infra/
    ├── cache.py             # File-based cache with TTL
    ├── rate_limiter.py      # Async token bucket rate limiter
    └── streaming.py         # Streaming event types

Development

# Clone and install with dev dependencies
git clone https://github.com/hermes-financial/hermes-financial.git
cd hermes-financial
uv sync --all-extras

# Run tests
uv run pytest

# Lint
uv run ruff check hermes/

# Type check
uv run mypy hermes/

Requirements

• Python >= 3.10
• An LLM API key (Anthropic or OpenAI) for agent functionality
• SEC EDGAR requires a User-Agent string identifying your application
• FRED requires a free API key from fred.stlouisfed.org
• LibreOffice for PDF export (optional, available in Docker image)

Professional Services

Need this deployed internally, integrated with your data sources, or extended with custom agents and models? [e-mail]

Acknowledgments

SEC EDGAR data retrieval is powered by edgartools (MIT), an excellent Python library by Dwight Gunning that provides high-level access to SEC EDGAR filings, XBRL-parsed financial statements, and structured document sections. If you find Hermes useful for SEC research, please consider starring the edgartools repo.

License

MIT