The original TradingAgents repository has 116882 stars. That number is the first thing most people see, and it is also the least interesting thing about the project. What matters more for a developer trying to read the code is that the documentation is in English, and the framework itself is a fairly dense multi-agent system with a lot of moving parts.
A Chinese translation of the repository now exists: yangshun2005/TradingAgents-cn. It is a translation of TauricResearch/TradingAgents, released today. It has no stars yet. This article is about what the project is, what the translation adds, and where the documentation still stops short.
What TradingAgents actually is
The official one-line description is "Multi-Agents LLM Financial Trading Framework." In practice, the framework assigns different roles to LLM-driven agents and lets them discuss a market situation before producing a decision. The roles named in the translated README include a fundamentals analyst, a sentiment specialist, a technical analyst, a trader, and a risk management team.
The structure is meant to mirror how a real trading desk divides labor. Each agent looks at a different slice of the problem, and the discussion between agents is part of the mechanism, not an afterthought. The project is written in Python and licensed under Apache-2.0.
If you want to read the source, the language is not a barrier. If you want to understand the design decisions, the release history, and the data-handling rules, the documentation is where the friction lives.
What the Chinese version adds
The translation is faithful to the original README, and it says so explicitly. The translated body is marked as a translation of the original document. On top of that, the maintainer added a section labeled "中文快速上手" (Chinese quick start), which is clearly marked as supplementary material, not part of the original.
That distinction is worth noting because it is easy to blur. The quick-start section is the part that lowers the barrier for Chinese-speaking developers. It answers questions the original README does not answer in its opening pages: what Python version do I need, how do I install it, what command do I run.
The translation date is recorded as 2026-10-27, and the original English README is kept in the repository as README_en.md. If you suspect a translation error, you can check the source directly. The maintainer invites issues and pull requests for inaccuracies.
Getting it running
The quick-start section is short and concrete. The minimum Python version is 3.11, taken from requires-python in pyproject.toml.
Install the project and its dependencies from the repository root:
pip install .
The README notes that requirements.txt is equivalent to pip install ., so either path works.
If you need Amazon Bedrock support, install the optional dependency instead:
pip install "tradingagents[bedrock]"
The core install does not include langchain-aws. That is a common trip-up: if you configure Bedrock and the import fails, you installed the wrong variant.
After installation, a command-line entry point named tradingagents is registered via [project.scripts] in pyproject.toml:
tradingagents
There is also a Docker path. The repository root contains a Dockerfile and a docker-compose.yml, and the image entry point is tradingagents. The translation does not spell out the exact startup command for Docker, and it says as much rather than inventing one.
Dependencies and environment
The core dependencies listed in pyproject.toml include:
langchain-corelangchain-openailangchain-anthropiclangchain-google-genailanggraphpandasyfinancetyperrich
Bedrock support is the optional langchain-aws extra. The Dockerfile is based on python:3.13-slim, which is Debian-family Linux. The README does not list supported operating systems or hardware requirements, and the translation does not fill those gaps.
One practical note the translation includes: dependencies come from PyPI, so users in mainland China may need a mirror or a proxy. The project mentions proxy support, but the exact configuration is not documented in the excerpt available.
The LLM key problem
The framework is driven by LLMs, so you need API keys for whichever provider you use. The material mentions API-key auto-detection and multiple provider support, along with TRADINGAGENTS_* environment variables. The specific variable names are not spelled out in the available excerpt.
This is the most likely place for a first run to fail. Prepare your keys before you run tradingagents, and expect to consult the full README body and CHANGELOG.md for the details the quick start does not cover.
Reading the release history
The translated README includes the news section, and it is more useful than it looks. Several releases in 2026 changed behavior in ways that affect how you interpret results.
Version 0.7.1 scores decisions only at the final closing price and settles across long holidays. It also fixed a bug where the model could read files committed after the run date. Version 0.7.0 added the ability to run on your own ChatGPT Plus or Pro plan, changed backtesting to score from the first close after each decision, added a new-version prompt at startup, and began shipping a wheel with each GitHub release.
Version 0.6.0 saves reports as a single HTML page, allows per-tier provider configuration so managers and analysts can run on different models, settles historical decisions per ticker while analysts work, and reads company news from Yahoo search when the Yahoo news feed is down.
Earlier releases in September and August 2026 dealt with point-in-time integrity across dated paths, SEC EDGAR fundamentals, backtesting over grids of tickers and dates, portfolio-aware runs, and look-ahead issues in FRED macro data, social sentiment, and decision-log memory.
If you are evaluating the framework for research, the point-in-time work is the part to read closely. A backtest that can see data published after the analysis date is not a backtest.
Where the translation stops
The translated README body is truncated after the "TradingAgents Framework" introduction. The full Installation and CLI sections are not included in the available material. The translation says this directly rather than guessing.
So the quick-start section gets you to a running command, but the finer configuration steps — CLI flags, provider setup, environment variable names — are not covered. For those, the translation points you to the repository README body and CHANGELOG.md.
That is the honest state of the documentation. A translation is not a rewrite, and the maintainer did not pretend otherwise.
Who this is for
If you read English documentation comfortably, the translation adds little beyond the quick-start section. The original repository is the canonical source, and it will always be more current.
If you are a Chinese-speaking developer who wants to understand the agent roles, the release history, and the data-integrity decisions without translating in your head, the translation is a reasonable starting point. The quick-start section in particular saves you from reading the whole README just to get the install right.
The star count belongs to the original project. The translation is new, and its quality will be judged by whether the issues and pull requests it invites actually get addressed.
Repository: yangshun2005/TradingAgents-cn
Original: TauricResearch/TradingAgents
Chinese documentation: https://github.com/yangshun2005/TradingAgents-cn
I'm an independent enterprise AI delivery consultant — private LLM deployment, RAG, agent workflows, AIOps. If your team is stuck somewhere in that stack, my inbox is open.
Top comments (0)