Connect your engine

TradingAgents is a Python system that runs on your own machine. A small bridge exposes it to this dashboard, so nothing about your API keys or data leaves your computer.

Setup steps

1. 1

   Get the engine code and create an environment

   Clone your backend repository, then create and activate a Python virtual environment inside it.

   git clone https://github.com/tsekennykl1/myTradingAgentsBackend.git
   cd myTradingAgentsBackend
   python3 -m venv .venv
   source .venv/bin/activate

   Copy
2. 2

   Install the dependencies

   This installs FastAPI, the worker, and the TradingAgents engine requirements.

   pip install -r requirements.txt

   Copy
3. 3

   Create your settings file

   Build it with the form further down this page, then save it as .env in the engine folder. Or start from the example file and edit it by hand.

   cp .env.example .env
   open .env   # or: nano .env

   Copy
4. 4

   Run the guided configurator

   initialSetup.py is the safest way to create .env. It asks one question at a time — which AI provider you use, its key, which deep-thinking and quick-thinking models to run, how many debate rounds, the temperature, and where to keep the cache and memory log. Press Enter at any prompt to accept the suggested value; it reads your existing .env first, so re-running it only changes what you retype, and keys are shown masked. Model names are checked against the engine's own list and friendly names are converted automatically ("DeepSeek Pro" becomes deepseek-v4-pro), so a typo is caught before anything is saved. Data keys for FRED and Alpha Vantage are your own free keys — create them on the FRED and Alpha Vantage sites and paste them in when asked. Nothing is written until every value passes validation; then it writes .env, refreshes .env.example, and creates sample_run_payload.json you can use to test the engine with one curl command.

   python initialSetup.py

   Copy
5. 5

   Or configure it unattended

   For servers and deployments, skip the questions entirely: put the same answers in a JSON file and run with --non-interactive. Point it at the file with --config, or set CONFIG_JSON instead — a local path, an http(s) address, an s3:// object, or the JSON itself all work. Keys named the friendly way (llm_provider, deep_think_llm, research_depth) are translated for you, and anything under "env" is written to .env exactly as typed. The same validation runs, so a bad model name or a missing provider key stops the deployment instead of failing mid-analysis.

   cat > config.json <<'JSON'
   {
     "llm_provider": "deepseek",
     "deep_think_llm": "deepseek-v4-pro",
     "quick_think_llm": "deepseek-v4-flash",
     "research_depth": 1,
     "temperature": 0.0,
     "env": {
       "DEEPSEEK_API_KEY": "sk-your-key",
       "PUBLIC_BASE_URL": "http://127.0.0.1:8000",
       "FRONTEND_ORIGINS": "http://localhost:8080"
     }
   }
   JSON
   
   python initialSetup.py --non-interactive --config config.json
   # or: CONFIG_JSON=s3://my-bucket/config/config.json python initialSetup.py --non-interactive

   Copy
6. 6

   Choose where the dashboard may call from

   PUBLIC_BASE_URL is the address the engine is reached at; FRONTEND_ORIGINS lists the dashboards allowed to talk to it. Without your dashboard address here, the browser blocks every request.

   PUBLIC_BASE_URL=http://127.0.0.1:8000
   FRONTEND_ORIGINS=http://localhost:8080,https://trading-agent-companion.lovable.app

   Copy
7. 7

   Start the engine

   Keep this terminal open while you use the dashboard.

   source .venv/bin/activate
   uvicorn app.main:app --host 127.0.0.1 --port 8000

   Copy
8. 8

   Point the dashboard at it

   Enter the address above (default http://127.0.0.1:8000) in the box at the top of this page and test the connection.

Build your .env

Pick your AI provider, paste its key, and copy the finished file. Model names come straight from the engine's own list, so they are always valid. Add your own free FRED and Alpha Vantage keys for market and economic data.

# TradingAgents engine settings — generated by the dashboard setup page

# --- Where the engine runs and who may call it ---
PUBLIC_BASE_URL=https://www.myfinance5051.com/api
FRONTEND_ORIGINS=http://localhost:8080,https://trading-agent-companion.lovable.app,https://www.myfinance5051.com

# --- AI provider ---
LLM_PROVIDER=deepseek
DEEP_THINK_LLM=deepseek-v4-pro
QUICK_THINK_LLM=deepseek-v4-flash
DEEPSEEK_API_KEY=""

# --- Market and economic data ---
ALPHA_VANTAGE_API_KEY=""
FRED_API_KEY=""

# --- Performance ---
TRADINGAGENTS_PARALLEL_ANALYSTS=1
TRADINGAGENTS_DATA_CACHE=1
TRADINGAGENTS_DATA_CACHE_TTL=21600
RUN_WORKER_COUNT=2

How to Generate the Large Language Model API_KEY?

Open the provider you want to use, sign in, create an API key, then copy it into your .env file under the variable name shown below. One provider is enough — pick that provider in the New Analysis form afterwards.

DEEPSEEK_API_KEY="sk-…"
# or OPENAI_API_KEY="sk-…"  /  GOOGLE_API_KEY="…"  /  ANTHROPIC_API_KEY="sk-ant-…"

Market and economic data keys

Both are free. Alpha Vantage supplies price and fundamentals data; FRED supplies US economic series used by the fundamentals and macro write-ups.

1. Alpha Vantage

   Open alphavantage.co/support/#api-key, enter your email, and the key appears on screen immediately. Free tier is rate limited, so keep the engine's data cache switched on.

   ALPHA_VANTAGE_API_KEY="your-key"

   Copy

2. FRED (Federal Reserve Economic Data)

   Create a free account at fredaccount.stlouisfed.org/apikeys, then request an API key. It is issued right away.

   FRED_API_KEY="your-key"

   Copy

After editing .env, stop the engine (Ctrl+C) and start it again so the new keys are picked up.

If the browser blocks the connection

Add this dashboard's address to FRONTEND_ORIGINS in .env and restart the engine. If it still refuses, run the engine with --host 0.0.0.0 and use your machine's local IP address instead of 127.0.0.1.