Basic Agent as Judge

Basic usage of Agent as Judge evaluation with numeric scoring and failure callbacks

This example demonstrates basic Agent as Judge evaluation with numeric scoring (1-10 scale) and an on_fail callback for handling evaluation failures.

Add the following code to your Python file

1from kern.agent import Agent
2from kern.db.sqlite import SqliteDb
3from kern.eval.agent_as_judge import AgentAsJudgeEval, AgentAsJudgeEvaluation
4from kern.models.openai import OpenAIResponses
5
6
7def on_evaluation_failure(evaluation: AgentAsJudgeEvaluation):
8    """Callback triggered when evaluation fails (score < threshold)."""
9    print(f"Evaluation failed - Score: {evaluation.score}/10")
10    print(f"Reason: {evaluation.reason[:100]}...")
11
12
13# Setup database to persist eval results
14db = SqliteDb(db_file="tmp/agent_as_judge_basic.db")
15
16agent = Agent(
17    model=OpenAIResponses(id="gpt-5.2"),
18    instructions="You are a technical writer. Explain concepts clearly and concisely.",
19    db=db,
20)
21
22response = agent.run("Explain what an API is")
23
24evaluation = AgentAsJudgeEval(
25    name="Explanation Quality",
26    criteria="Explanation should be clear, beginner-friendly, and use simple language",
27    scoring_strategy="numeric",  # Score 1-10
28    threshold=9,  # Pass if score >= 9
29    on_fail=on_evaluation_failure,
30    db=db,
31)
32
33result = evaluation.run(
34    input="Explain what an API is",
35    output=str(response.content),
36    print_results=True,
37    print_summary=True,
38)
39
40# Query database for stored results
41print("Database Results:")
42eval_runs = db.get_eval_runs()
43print(f"Total evaluations stored: {len(eval_runs)}")
44if eval_runs:
45    latest = eval_runs[-1]
46    print(f"Eval ID: {latest.run_id}")
47    print(f"Name: {latest.name}")

Set up your virtual environment

1uv venv --python 3.12
2source .venv/bin/activate

1uv venv --python 3.12
2.venv\Scripts\activate

Install dependencies

1uv pip install -U kern-ai openai

Export your OpenAI API key

1export OPENAI_API_KEY="your_openai_api_key_here"

1$Env:OPENAI_API_KEY="your_openai_api_key_here"

Run the example

1python agent_as_judge_basic.py