tei-loop 0.2.0

Target, Evaluate, Improve: A self-improving loop for agentic systems

TEI Loop

Target, Evaluate, Improve — a self-improving loop for agentic systems.

Get Started

pip install tei-loop

python3 -m tei_loop your_agent.py

TEI does the rest:

• Finds your agent function automatically
• Generates a test query automatically
• Evaluates across 4 dimensions
• Applies targeted improvements
• Saves results to tei-results/ folder (created next to your agent file)

If pip is not found, use pip3 instead.

What TEI Creates

your_project/
  your_agent.py
  tei-results/           <-- created automatically
    run_20260218_071500.json
    latest.json

Options

python3 -m tei_loop agent.py                             # Auto everything
python3 -m tei_loop agent.py --function my_func          # Pick specific function
python3 -m tei_loop agent.py --query "custom input"      # Custom test query
python3 -m tei_loop agent.py --retries 5                 # More improvement cycles
python3 -m tei_loop agent.py --verbose                   # Detailed output

Python API

import asyncio
from tei_loop import TEILoop

def my_agent(query: str) -> str:
    return result

async def main():
    loop = TEILoop(agent=my_agent)
    result = await loop.run("test query")
    print(result.summary())

asyncio.run(main())

4 Evaluation Dimensions

Dimension	What it checks
Target Alignment	Did the agent pursue the correct objective?
Reasoning Soundness	Was the reasoning logical?
Execution Accuracy	Were tools called correctly?
Output Integrity	Is the output complete and accurate?

Two Modes

Runtime (default): Per-query improvement in seconds.

Development: Across many queries, permanent prompt improvements.

results = await loop.develop(queries=[...], max_iterations=50)

Works With Any Agent

Any Python callable. No framework lock-in: LangGraph, CrewAI, custom Python, FastAPI, anything.

License

MIT