Summary

DeepEar is an open-source framework transforming public opinion into actionable investment logic chains for financial professionals. It automates fragmented data analysis from social media and news into quantitative trading signals, bridging the gap between raw information and investment decisions.

How It Works

It employs multi-agent collaboration (Trend, Finance, Report agents) and a novel news-aware Kronos time-series model that injects news event impacts into price predictions. A hybrid RAG engine combines BM25 and vector search for precise information retrieval, outputting visualized reports with interactive logic graphs.

Quick Start & Requirements

• Installation: Clone repo, cd DeepEar, uv sync.
• Prerequisites: Python 3.12+, uv (recommended). API keys for LLMs (OpenAI, Ollama) and optionally Jina.
• Running: Recommended: Interactive dashboard (cd dashboard/frontend, npm install, npm run build, uv run dashboard/server.py). CLI (uv run src/main_flow.py) and Agent Skill (uv run skills/deepear/scripts/server.py) also available.
• Links: Live Demo: https://deepear.vercel.app/.

Highlighted Details

• Multi-Agent Collaboration: Specialized agents for trend spotting, financial analysis, and report writing.
• Agent Skill Ready: Integrates as a plug-and-play skill for AI agents (Antigravity, OpenCode).
• Interactive Dashboard: React-based UI for real-time workflow monitoring.
• News-Aware Time-Series Model: Custom Kronos integration with news-projection layer for price shock prediction.
• Hybrid RAG Engine: Combines BM25 and Vector Search for precise retrieval.
• Data Sources: Supports 15+ sources (Weibo, Cailian Press, WSJ).

Maintenance & Community

A detailed roadmap indicates ongoing development. Contribution guidelines are provided. No direct community channels (Discord/Slack) are listed.

Licensing & Compatibility

Distributed under the MIT License, generally permitting commercial use and closed-source integration.

Limitations & Caveats

Future development includes US market support, LangGraph migration, and prediction market integration. The dashboard requires initial frontend setup and user registration. Core functionality depends on user-provided API keys.