agentevals/examples/custom_evaluators/eval_config.yaml at 854133a67287e249a59bd96d8e0fd6724740dd1a · agentevals-dev/agentevals · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
# Example eval config using a custom code-based evaluator alongside built-in ones.
#
# Run with:
#   agentevals run samples/helm.json \
#     --config examples/custom_evaluators/eval_config.yaml \
#     --eval-set samples/eval_set_helm.json

evaluators:
  # Built-in metric
  - name: tool_trajectory_avg_score
    type: builtin

  # Custom code evaluators (local scripts)
  - name: tool_call_checker
    type: code
    path: ./examples/custom_evaluators/tool_call_checker.py
    threshold: 1.0
    config:
      min_tool_calls: 1

  - name: response_quality
    type: code
    path: ./examples/custom_evaluators/response_quality.py
    threshold: 0.7
    config:
      min_response_length: 20

  # Reference an evaluator from Github
  - name: random_evaluator
    type: remote
    source: github
    ref: evaluators/random_evaluator/random_evaluator.py
    threshold: 0.110
    executor: local

  # OpenAI Evals API graders (requires OPENAI_API_KEY)
  - name: response_similarity
    type: openai_eval
    threshold: 0.8
    grader:
      type: text_similarity
      evaluation_metric: fuzzy_match

  - name: city_name_check
    type: openai_eval
    grader:
      type: string_check
      operation: eq
      reference: "{{ item.expected_response }}"