Using TFire with LLMs

TFire is optimized to be effectively used together with Large Language Models (LLMs) like Claude or ChatGPT. TFire comes with a tailor-made documentation file (found under Downloads) specifically made to be attached when using an LLM.

Example Workflow

1. Strategy Development:

# 1. Include documentation
[Paste or attach LLM documentation]

# 2. Describe strategy
"Help me create a momentum strategy that:
- Uses RSI and MACD indicators
- Buys when RSI crosses above 30
- Sells when RSI crosses above 70
- Implements proper error handling"

# 3. Receive and review code

Common Use Cases

1. Creating New Layers

# Example prompt:
"'Help me create a new Layer that implements Relative Strength Index (RSI) calculations."

1. Building Trading Strategies

# Example prompt:
"Implement a strategy that buys when the 50-day moving average crosses above the 200-day moving average."

1. Debugging Code

# Paste code and error message
# Example prompt:
"What might be causing this issue?"

Effective Prompting

Structure your requests like this:

Context: [Paste relevant TFire documentation]

Task: Help me implement [your specific requirement]

Additional requirements:
- [Any specific constraints]
- [Performance considerations]
- [Error handling needs]

Do:

1. Be Specific: Clearly describe your requirements and any constraints
2. Include Context: Provide relevant documentation sections
3. Iterate: Start simple and gradually add complexity
4. Validate: Always test generated code thoroughly
5. Ask for Explanations: Have the LLM explain the code it generates

Don't:

1. Don't assume LLMs have current documentation - always provide it
2. Don't rely on LLMs for numerical accuracy - test thoroughly
3. Avoid asking for complex strategies without providing context