A research terminal that files its own morning briefs

What it does Trading-R1 is a terminal UI for financial research and analysis. You ask a question — say, NVDA’s last quarterly revenue — and an agent with a finance-specific toolkit (data, valuation, and web tools) produces an answer you can export as PDF and Markdown. It also runs scheduled jobs: a morning brief on your watchlist, delivered to email or Telegram at 08:00, every weekday.

The interesting bit The modes are the quiet cleverness: research goes web-first, analytical goes numbers-first, and auto lets the agent pick. The Telegram channel is bidirectional — with reply on, you can message the bot from your phone and it runs a short research pass and answers you. It’s backed by a paper (arXiv:2509.11420) on LLM trading via reinforcement learning, though the repo itself is the harness, not the model.

Key highlights

• Terminal-native with a command palette (/), modes, and session management
• Pluggable models: Tauric, OpenAI, Anthropic, Gemini, or a local vLLM server (Qwen)
• Reports export to PDF/Markdown and send via Telegram, email, or a folder
• Scheduled jobs with timezone support and multi-channel delivery
• Background daemon via systemd/launchd so jobs and Telegram replies survive terminal close

Caveats

• The default tauric model requires a Tauric account and API key — this is a hosted product’s client as much as an open tool
• The RL-trained “Trading-R1” model from the paper isn’t something you get here; the repo is the terminal and agent harness
• Explicitly research/education only, not investment advice

Verdict Worth a look if you want an agent-shaped financial research workflow with scheduled briefs, especially if you already live in a terminal. If you wanted the RL trading model itself, that’s not what’s in the box.