Web UI

If you want to use Trading Strategy Tester web UI, you can find the installation instructions in the Administrator Documentation. The web UI is a separate project that uses Trading Strategy Tester package as a backend. You can find the repository here.

Main Page

When you open the web UI, you will see the main page which looks like this and consists of several sections:

Prompt Section

This is the main text area in the middle of the page where you can insert your prompt. It's located in the red rectangle in the next image:

After you inserted your prompt you can click the button next to the text area to process the prompt, or you can press Enter key on your keyboard.

Select Model Section

In the right bottom corner of the page you can see the Select LLM section. This is where you can select which model and approach you want to use for processing your prompt.

You can choose between: - Llama 3.2:
This is a fine-tuned Llama 3.2 3B model that generates the whole Strategy object from the prompt.

• Llama 3.2 Fields:
  These are fine-tuned Llama 3.2 models that generate the fields of the Strategy object from the prompt. After generating the fields separately, the Strategy object is created from them. All the fields except conditions are generated using the Llama 3.2 1B model. The conditions field is generated using the Llama 3.2 3B model.

• Llama 3.2 FSP:
  This is a Llama 3.2 3B model that uses few-shot prompting to generate the Strategy object from the prompt. This model creates the whole Strategy object directly.

• Llama 3.2 FSP Fields:
  These are Llama 3.2 3B models that use few-shot prompting to generate the fields of the Strategy object from prompts. After generating the fields separately, the Strategy object is created from them.

• Strategy object:
  Here, the user can submit the Strategy object directly as Python code. It should be valid Python code that creates the Strategy object, and all the parameters should be named correctly except for three exceptions:

  • The ticker parameter in TradingSeries objects should not be named. For example: do not use RSI(ticker='AAPL', length=14); the correct way is RSI('AAPL', length=14).

  • The same rule applies to the const_number parameter in CONST TradingSeries. Do not use CONST(const_number=30); the correct way is CONST(30).

  • Do not name year, month, day in the datetime object. For example: do not use datetime(year=2015, month=1, day=1); the correct way is datetime(2015, 1, 1).

Here is an example of how valid Strategy object looks like:

Strategy(
    ticker="AAPL",
    position_type= PositionTypeEnum.LONG_SHORT_COMBINED,
    buy_condition= CrossOverCondition(
        first_series=RSI('AAPL', length=14),
        second_series=CONST(30)
    ),
    sell_condition=CrossUnderCondition(
        first_series=RSI('AAPL'),
        second_series=CONST(70)
    ),
    start_date=datetime(2015, 1, 1),
    end_date=datetime(2020, 1, 1),
    order_size=Contracts(value=1)
)

If you want to learn more about Strategy object structure, you can check the Strategy documentation and if you want to learn more about how to write prompts for LLMs, you can check manual at How to write prompts.

Light/Dark Mode

In the right top corner of the page you can see the Dark Mode button.

This button allows you to switch between light and dark mode.

Once you submit the prompt, the web UI will process it and display the results in the Result section. To learn more about the results, check the Strategy results.