The Prompt Editor is the core workspace of Prompt Studio. It provides a structured, IDE‑like interface for building and refining your prompts.
Message Blocks
Prompts are composed of structured message blocks. Each block represents a message in the conversation:
| Block Type | Purpose |
|---|
| System | Sets the model’s behavior, personality, and instructions |
| User | Sample user input for testing and demonstration |
| Assistant | Pre-filled assistant responses for few-shot examples |
Template Variables
Use {{variable_name}} syntax anywhere in your message blocks to create dynamic prompts:
You are a helpful assistant for {{company_name}}.
The user's name is {{user_name}} and they need help with {{topic}}.
curl -X POST https://api.traceport.ai/v1/prompt/my-prompt/run \
-H "Authorization: Bearer <API_KEY>" \
-d '{"variables": {"company_name": "Acme", "user_name": "John", "topic": "billing"}}'
Model Parameters
The right panel lets you configure model-specific settings:
| Parameter | Description |
|---|
| Model | Select the provider and model (e.g., Anthropic Claude 3.5 Sonnet) |
| Temperature | Controls randomness (0 = deterministic, 1 = creative) |
| Max Tokens | Maximum number of tokens in the response |
| Top P | Nucleus sampling parameter |
- Run — Execute the prompt and view the output in the right panel
- Save — Save the current state as a draft
- Publish — Promote the draft to a published version
- Fork — Create a copy of the prompt for experimentation
- Optimize — Let Traceport’s AI analyze and suggest improvements to your prompt
Use the Fork feature to experiment with prompt variations without affecting the original.
