Chart Library integration for financial research roles

[grahammccain]

Proposal

Add Chart Library as a tool for MetaGPT's financial research roles. Chart Library provides 24M+ historical chart pattern embeddings that enable AI agents to perform visual pattern matching — "find me charts that looked like this and show what happened next."

How it fits MetaGPT's role-based architecture

MetaGPT's multi-agent role system maps naturally to financial research workflows:

Researcher role could use Chart Library to:

• Search for similar historical patterns by ticker + date
• Pull forward return distributions (1/3/5/10 day) from matches
• Detect technical patterns (breakouts, bull flags, wedges)

Analyst role could use regime context:

• Market regime analysis (VIX quartile, sector rotation)
• Correlation shifts from SPY
• Signal crowding indicators

Example action

import httpx

class ChartPatternSearch:
    """Search Chart Library for similar historical chart patterns."""
    
    async def run(self, symbol: str, date: str) -> str:
        async with httpx.AsyncClient() as client:
            resp = await client.get(
                "https://chartlibrary.io/api/v1/search",
                params={"symbol": symbol, "date": date},
                headers={"X-API-Key": "your-key"}
            )
            data = resp.json()
            matches = data.get("matches", [])
            summary = data.get("ai_summary", "")
            return f"Found {len(matches)} similar patterns.\n\nAI Summary: {summary}"

What Chart Library offers

• 24M+ embeddings across 19K symbols, 10 years of data
• Multi-timeframe: RTH, premarket, 5min, 15min, 30min, 1hr, 3-day, 5-day
• MCP server: pip install chartlibrary-mcp (19 tools)
• REST API: 40+ endpoints at chartlibrary.io/developers
• Free tier: 200 API calls/day

Financial research is one of the most requested use cases for multi-agent systems — this would give MetaGPT a concrete data source for that vertical.