Skip to main content
To run:

​
1. Install the demo file from GitHub:

curl -L -o run_demo.py \
  https://raw.githubusercontent.com/synth-laboratories/synth-ai/main/demos/gepa_banking77/run_demo.py

​
2. Install the Synth AI SDK in your Python environment:

pip install synth-ai

​
3. Start the GEPA prompt optimization job:

python run_demo.py
On job start, the following happens:
  • Launches a Local API — a FastAPI server that loads the dataset and handles rollout requests
  • Wraps the Local API in a Cloudflare Tunnel so Synth AI’s server can reach your local machine
  • Optimizes the prompt by having Synth AI’s server call the Local API

​
Using the TUI

You can view your prompt optimization job’s progress live through Synth AI’s terminal user interface. In the same environment where the synth-ai SDK is installed, open the TUI and view your job’s status by running: 
synth-ai
Prompt optimization status

​
Prompt optimization finishes

The TUI notifies when the prompt optimization job completes. To get the optimized prompt:
  1. Select the prompt optimization run by pressing the b key and navigating with j/k
  2. Press the p key to open the completed job’s optimized prompt
  3. Press the y key to copy the optimized prompt to clipboard Prompt optimization complete

​
Evals on optimized prompt

Once your prompt optimization job completes, both a baseline and optimized eval will run, all viewable live in the TUI:

​
Baseline eval

Baseline eval job

​
Optimized eval

Optimized eval job