eval_utils

Evaluation model and scoring utilities

​

browseruse_bench.utils.eval_utils

Evaluation-related utility functions and classes.

​

Import

from browseruse_bench.utils import (
    EvaluationModel,
    load_evaluation_model,
    encode_image,
    extract_score_from_response,
    calculate_success,
    normalized_results_file,
)

---

​

EvaluationModel

OpenAI model wrapper class for task evaluation.

class EvaluationModel:
    def __init__(
        self,
        model: str = "gpt-4o",
        api_key: str = None,
        base_url: str = None
    )

model · str · 默认值: "gpt-4o"

Model name

api_key · str · 默认值: "None"

API Key, defaults to environment variable

base_url · str · 默认值: "None"

API Base URL, defaults to environment variable

​

generate

Generate evaluation response with automatic retry.

def generate(
    self,
    messages: List[Dict],
    max_tokens: int = 2048,
    temperature: float = 0.3,
    **kwargs
) -> str

---

​

load_evaluation_model

Load evaluation model with environment variable fallback.

def load_evaluation_model(
    model: str = None,
    api_key: str = None,
    base_url: str = None
) -> EvaluationModel

​

Environment Variables

Variable	Description
EVAL_MODEL_NAME	Model name (fallback: gpt-4o)
EVAL_MODEL_API_KEY	API Key (fallback: OPENAI_API_KEY)
EVAL_MODEL_BASE_URL	Base URL (fallback: OPENAI_BASE_URL)

---

​

encode_image

Convert a PIL image to base64 string.

def encode_image(
    image: PIL.Image,
    scale_factor: float = 1.0
) -> str

image · PIL.Image · 必填

PIL Image object

scale_factor · float · 默认值: "1.0"

Image scale factor (between 0.0 and 1.0), e.g., 0.5 means 50% size

---

​

extract_score_from_response

Extract numerical score from evaluation response.

def extract_score_from_response(response: str) -> int

response · str · 必填

Evaluation response text

return · int

Extracted score (0 if not found)

---

​

calculate_success

Determine if task is successful based on score threshold.

def calculate_success(
    score: int,
    threshold: int = 60
) -> bool

score · int · 必填

Task score

threshold · int · 默认值: "60"

Success threshold

return · bool

True if score meets or exceeds threshold

---

​

normalized_results_file

Context manager that yields a path guaranteed to be JSONL format.

@contextmanager
def normalized_results_file(results_file: Path) -> Generator[Path, None, None]

results_file · Path · 必填

Results file path

​

Usage Example

from browseruse_bench.utils import normalized_results_file

with normalized_results_file(Path("results.json")) as jsonl_path:
    # jsonl_path is guaranteed to be JSONL format
    with open(jsonl_path) as f:
        for line in f:
            record = json.loads(line)