# Athenara

> The open ecosystem for agentic trading: a community registry of AI trading agents, architectures, datasets, skills, research papers, benchmarks, and performance results. Every resource is a Markdown file with YAML frontmatter, validated against open JSON Schemas, in a public git repository. Contributions are pull requests. Content is research and engineering knowledge, not financial advice.

## How to navigate this site

Every registry entry is published in three formats. Prefer Markdown:

- Markdown (frontmatter + body): https://athenara.info/{section}/{id}.md
- JSON (frontmatter + body as `content`): https://athenara.info/api/v1/{section}/{id}.json
- HTML (for humans): https://athenara.info/{section}/{id}/

Sections: agents, architectures, datasets, skills, papers, benchmarks, results.

Fetch progressively — only as deep as the task needs:

1. This file is the complete map. If the entry you need is listed below, fetch its Markdown URL directly and stop.
2. To search or filter across everything, fetch https://athenara.info/api/v1/search-index.json once (id, name, summary, tags, and URLs per entry) and filter locally.
3. To browse one section's structured metadata (no bodies), fetch https://athenara.info/api/v1/{section}.json.
4. Only for corpus-level work (mirroring, RAG ingestion), fetch https://athenara.info/api/v1/{section}.full.json per section or https://athenara.info/llms-full.txt for the entire registry in one file.

Useful facts:

- Frontmatter edge fields (`implements`, `paper`, `uses_dataset`, `evaluated_on`, `reproduces`, `builds_on`) hold knowledge-graph links as registry slugs. Follow them by building the URLs above.
- Every entry declares `prerequisites` — the knowledge/effort it assumes of a user, on four axes (trading, ai, programming, setup), each rated none/basic/moderate/advanced/expert. Use these to match resources to a user's background.
- Results reference agents by slug and carry mandatory context (period, mode, costs, out-of-sample, verification). The leaderboard (https://athenara.info/leaderboard/) is a view over results.
- The JSON Schemas defining every field live at https://athenara.info/schemas/{type}.json (types: common, agent, architecture, dataset, skill, paper, benchmark, result). Validate against them before contributing.
- API root with counts and conventions: https://athenara.info/api/v1/index.json.

## Agents

- [Buy-and-hold baseline](https://athenara.info/agents/buy-and-hold-baseline.md): The reference passive baseline every active result should be compared against; buys the index and does nothing.
- [FinMem](https://athenara.info/agents/finmem.md): An LLM trading agent that combines a configurable character profile with a layered memory module modeled on human trader cognition.
- [FinRobot](https://athenara.info/agents/finrobot.md): An open-source multi-agent platform from AI4Finance Foundation that applies LLM agents to equity research, trading strategy, and risk evaluation tasks.
- [Freqtrade](https://athenara.info/agents/freqtrade.md): A Python cryptocurrency trading bot with backtesting, hyperparameter optimization, and an adaptive machine-learning module called FreqAI.
- [Hummingbot](https://athenara.info/agents/hummingbot.md): An open-source Python framework for building and running market-making and arbitrage bots across centralized and decentralized crypto venues.
- [TradingAgents](https://athenara.info/agents/tradingagents.md): A multi-agent LLM framework that assigns specialized analyst, researcher, trader, and risk-management roles to simulate the workflow of a trading firm.

## Architectures

- [FinGPT](https://athenara.info/architectures/fingpt.md): An open-source family of financial large language models and fine-tuning pipelines from AI4Finance Foundation, built around low-cost LoRA adaptation of open base models.
- [FinRL](https://athenara.info/architectures/finrl.md): A deep reinforcement learning library that packages market environments, DRL algorithms, and backtesting into a pipeline for training automated trading agents.
- [LEAN](https://athenara.info/architectures/lean.md): QuantConnect's open-source event-driven algorithmic trading engine, supporting strategies written in Python or C# across nine asset classes.
- [Multi-agent debate](https://athenara.info/architectures/multi-agent-debate.md): Opposing analyst agents argue a bullish and bearish case before a decision-maker agent commits to a trade.
- [NautilusTrader](https://athenara.info/architectures/nautilustrader.md): A Rust-native, event-driven algorithmic trading platform whose backtest and live engines share identical execution semantics.
- [Planner–executor](https://athenara.info/architectures/planner-executor.md): A planning agent decomposes a trading objective into steps that cheaper executor agents or plain tools carry out.
- [Qlib](https://athenara.info/architectures/qlib.md): Microsoft's AI-oriented quantitative investment platform covering the research workflow from data handling and model training to portfolio construction and backtesting.
- [Reinforcement-learning policy](https://athenara.info/architectures/rl-policy.md): An agent's trading decisions come from a policy trained with reinforcement learning against a market environment.

## Datasets

- [ECTSum](https://athenara.info/datasets/ectsum.md): A benchmark of 2,425 long earnings call transcripts paired with expert-written telegram-style bullet-point summaries drawn from Reuters articles.
- [EDGAR-CORPUS](https://athenara.info/datasets/edgar-corpus.md): A corpus of 220,375 SEC 10-K annual reports from 1993 to 2020, split into their individual item sections and released as JSON.
- [Financial PhraseBank](https://athenara.info/datasets/financial-phrasebank.md): A human-annotated sentiment dataset of 4,840 English sentences from financial news, labelled positive, negative, or neutral from an investor's perspective.
- [FNSPID](https://athenara.info/datasets/fnspid.md): A time-series dataset pairing 29.7 million stock price records with 15.7 million financial news articles for 4,775 companies in the S&P 500 universe, 1999–2023.

## Results

- [buy-and-hold-spy-2024](https://athenara.info/results/buy-and-hold-spy-2024.md): backtest of agent `buy-and-hold-baseline`, equities, 2024-01-02 to 2024-12-31, author-reported

## Optional

- [A Multimodal Foundation Agent for Financial Trading (FinAgent)](https://athenara.info/papers/finagent-paper.md): Presents FinAgent, a multimodal trading agent that processes numerical, textual, and visual market data with dual-level reflection and diversified memory retrieval.
- [FinMem: A Performance-Enhanced LLM Trading Agent with Layered Memory and Character Design](https://athenara.info/papers/finmem-paper.md): Proposes an LLM trading agent whose layered memory and configurable character profile let it adapt to new market information over multiple time horizons.
- [Large Language Model Agent in Financial Trading: A Survey](https://athenara.info/papers/llm-trading-agent-survey.md): Surveys the research landscape of LLM-based agents applied to financial trading, covering architectures, data inputs, backtesting results, and open challenges.
- [TradingAgents: Multi-Agents LLM Financial Trading Framework](https://athenara.info/papers/tradingagents-paper.md): Introduces a multi-agent LLM trading framework whose specialized roles mirror the structure of a professional trading firm.
- [FinBen](https://athenara.info/benchmarks/finben.md): A holistic open-source financial benchmark of 42 datasets spanning 24 tasks, including the first stock-trading evaluation for LLMs.
- [InvestorBench](https://athenara.info/benchmarks/investorbench.md): A benchmark for evaluating LLM-based agents on financial decision-making tasks across single stocks, cryptocurrencies, and ETFs.
- [PIXIU / FLARE](https://athenara.info/benchmarks/pixiu-flare.md): An open-source financial LLM resource bundling the FinMA instruction-tuned models, a 136K-sample instruction dataset, and the FLARE evaluation benchmark.