Summary

Dexter JP is an autonomous AI research agent for deep financial analysis of Japanese listed companies. It targets technical users by automating complex research tasks—planning, multi-source data acquisition, validation, and report generation—offering significant analytical depth and time savings.

How It Works

This project refactors the US-based Dexter agent for Japan, using a LangChain agent loop for autonomous planning, tool selection, execution, and validation. A key innovation is the get_financials meta-tool, featuring an internal LLM router to dynamically orchestrate and integrate results from various financial sub-tools. It accesses EDINET DB, J-Quants, and web searches, guided by user-defined research rules (.dexter/RULES.md) and employs context compression and session memory for efficiency.

Quick Start & Requirements

Installation involves cloning the repo, running bun install, and configuring API keys in .env.

• Primary Install: bun install
• Run: bun run start
• Prerequisites: Bun runtime, LLM API keys (OpenAI, Anthropic, Google, xAI, OpenRouter), EDINET DB API key.
• Optional: J-Quants API key, Web Search API keys, X/Twitter API token, local LLM setup.
• Setup: Requires obtaining and configuring multiple API keys.
• Docs: README provides setup and usage examples.

Highlighted Details

• Autonomous Research: Handles complex, multi-source financial analysis from a single query without human intervention.
• get_financials Meta-tool: LLM-powered router intelligently selects and integrates financial data sub-tools.
• Customizable Research Rules: Users define analytical styles and investment criteria in .dexter/RULES.md, influencing agent behavior.
• Multi-Channel Integration: Supports Slack, Discord, LINE, and WhatsApp via bun run gateway.
• Context Compression & Memory: Manages large datasets efficiently and retains state across research sessions.

Maintenance & Community

A fork of the original Dexter by @virattt. While it integrates with messaging platforms like Slack and Discord, direct community links (e.g., Discord server) are not detailed in the README.

Licensing & Compatibility

• License: MIT.
• Compatibility: Permissive for commercial use. Users must manage API key terms. LINE bot integration requires a publicly accessible webhook URL.

Limitations & Caveats

Functionality heavily relies on obtaining and configuring numerous third-party API keys, some potentially incurring costs. Initial setup is complex. Deploying the LINE bot requires a publicly accessible webhook endpoint, limiting use in private network environments without additional setup.