zjy1346/OpenThesis
An open-source desktop tool that turns public company filings into an evidence-backed investment thesis, letting you plug in any AI model
OpenThesis is a Windows desktop app for individual long-term investors that pulls filings from SEC, HKEX, and Chinese exchange disclosure sites, runs deterministic financial calculations in code, and hands the verified facts to a team of specialist AI agents. Every AI-written claim must cite the collected filing evidence, and users bring their own AI model (OpenAI, DeepSeek, Qwen, Ollama, etc.) rather than being locked to one provider. The project explicitly avoids trading, brokerage connections, and short-term price signals.
What it does
- Fetches official filings (SEC EDGAR, HKEX, and Chinese exchange disclosure portals) and extracts financial facts with a page-level evidence trail
- Financial summaries and reverse discounted-cash-flow calculations are computed by fixed code rather than generated by a language model, so numbers stay consistent
- Multiple purpose-built AI agents (financial, business, accounting-risk, growth, skeptical, forecasting, synthesis, verification) work over the same evidence set, and any factual claim must cite it
- Works with any OpenAI-compatible endpoint, including DeepSeek, Qwen, Kimi, GLM, OpenAI, Gemini, OpenRouter, or a local Ollama model, so no single vendor is required
- Runs offline-first: no AI call happens until a model is explicitly chosen and a research run started, API keys stay in memory only, and a fully offline synthetic demo company is included
Why it matters
For anyone trying to use AI to analyze stocks or companies, this addresses two real risks: language models inventing numbers, and being locked into one AI vendor's pricing or availability. Separating deterministic math from AI-generated narrative, and keeping every claim traceable to a source page, is a practical pattern that could apply beyond investing to any research workflow that mixes hard data with AI summarization.
Terms in this repo
- Reverse DCF · A discounted cash flow calculation run backward from the current stock price to infer what growth assumptions the market is pricing in
- SEC EDGAR / HKEX / CNInfo · Official filing databases for US, Hong Kong, and mainland China listed companies respectively
- .othesis format · A declarative, permission-restricted ZIP file format for research modules that cannot execute arbitrary code, network, or filesystem access
- OpenAI-compatible endpoint · Any AI API that follows OpenAI's request/response format, letting one app work with many different model providers
Repository description (English)
面向长期投资者的开源、模型无关 AI 公司研究系统。Open-source, model-agnostic AI company research for long-term investors.
Open on GitHubTrending repos
- openai/codexOpenAI's coding agent, now runnable straight from your terminal
- cordiverse/cordisA TypeScript framework for snapping feature pieces on and off whenever you need them
- ripienaar/free-for-devA single repo lists every free-tier cloud, API and collaboration tool developers actually need
- Wei-Shaw/sub2apiA relay server that lets one Claude/OpenAI/Gemini/Grok subscription be shared by many users
- multica-ai/andrej-karpathy-skillsA single instruction file that stops AI coding assistants from making silent bad guesses
- eneskirca/nodetermA workspace that puts scattered terminal tabs and AI coding agents on one draggable map
- affaan-m/ECCA toolkit that teaches AI coding agents to work like disciplined engineers
- n8n-io/n8nn8n lets you build automation workflows and AI agents by dragging nodes on a canvas, with code as a fallback
Latest from METAL LAB
- Starting from ElevenLabs and counting up, a string of 'number+labs' names turned up
- OpenAI's Codex CLI: an open-source coding agent that runs in the terminal
- Inherent's 27B Model Faraday Outperforms Anthropic and OpenAI
- LinkedIn's AI slop report button passes 1 million clicks
- NVIDIA pays startup Poolside $6 billion to license its AI model