judgeval 0.4.0__py3-none-any.whl → 0.6.0__py3-none-any.whl

judgeval/__init__.py

@@ -2,6 +2,7 @@
 from judgeval.clients import client, together_client
 from judgeval.judgment_client import JudgmentClient
 from judgeval.version_check import check_latest_version
+from judgeval.local_eval_queue import LocalEvaluationQueue
 
 check_latest_version()
 
@@ -10,4 +11,5 @@ __all__ = [
     "client",
     "together_client",
     "JudgmentClient",
+    "LocalEvaluationQueue",
 ]

judgeval/cli.py

@@ -0,0 +1,65 @@
+#!/usr/bin/env python3
+
+import typer
+from pathlib import Path
+from dotenv import load_dotenv
+from judgeval.common.logger import judgeval_logger
+from judgeval.judgment_client import JudgmentClient
+
+load_dotenv()
+
+app = typer.Typer(
+    no_args_is_help=True,
+    rich_markup_mode=None,
+    rich_help_panel=None,
+    pretty_exceptions_enable=False,
+    pretty_exceptions_show_locals=False,
+    pretty_exceptions_short=False,
+)
+
+
+@app.command("upload_scorer")
+def upload_scorer(
+    scorer_file_path: str,
+    requirements_file_path: str,
+    unique_name: str = typer.Option(
+        None, help="Custom name for the scorer (auto-detected if not provided)"
+    ),
+):
+    # Validate file paths
+    if not Path(scorer_file_path).exists():
+        judgeval_logger.error(f"Scorer file not found: {scorer_file_path}")
+        raise typer.Exit(1)
+
+    if not Path(requirements_file_path).exists():
+        judgeval_logger.error(f"Requirements file not found: {requirements_file_path}")
+        raise typer.Exit(1)
+
+    try:
+        client = JudgmentClient()
+
+        result = client.save_custom_scorer(
+            scorer_file_path=scorer_file_path,
+            requirements_file_path=requirements_file_path,
+            unique_name=unique_name,
+        )
+
+        if not result:
+            judgeval_logger.error("Failed to upload custom scorer")
+            raise typer.Exit(1)
+
+        raise typer.Exit(0)
+    except Exception:
+        raise
+
+
+@app.command()
+def version():
+    """Show version info"""
+    judgeval_logger.info("JudgEval CLI v0.0.0")
+
+
+if __name__ == "__main__":
+    app()
+
+# judgeval upload_scorer /Users/alanzhang/repo/JudgmentLabs/judgeval/src/demo/profile_match_scorer.py /Users/alanzhang/repo/JudgmentLabs/judgeval/src/demo/requirements.txt

judgeval/clients.py

@@ -2,7 +2,6 @@ import os
 from dotenv import load_dotenv
 from openai import OpenAI
 from typing import Optional
-from together import Together, AsyncTogether
 
 PATH_TO_DOTENV = os.path.join(os.path.dirname(__file__), ".env")
 load_dotenv(dotenv_path=PATH_TO_DOTENV)
@@ -28,6 +27,8 @@ async_together_client: Optional["AsyncTogether"] = None
 together_api_key = os.getenv("TOGETHERAI_API_KEY") or os.getenv("TOGETHER_API_KEY")
 if together_api_key:
     try:
+        from together import Together, AsyncTogether
+
         together_client = Together(api_key=together_api_key)
         async_together_client = AsyncTogether(api_key=together_api_key)
     except Exception:

judgeval/common/api/api.py

@@ -20,13 +20,11 @@ from judgeval.common.api.constants import (
     JUDGMENT_EVAL_DELETE_API_URL,
     JUDGMENT_ADD_TO_RUN_EVAL_QUEUE_API_URL,
     JUDGMENT_GET_EVAL_STATUS_API_URL,
-    JUDGMENT_CHECK_EXPERIMENT_TYPE_API_URL,
-    JUDGMENT_EVAL_RUN_NAME_EXISTS_API_URL,
     JUDGMENT_SCORER_SAVE_API_URL,
     JUDGMENT_SCORER_FETCH_API_URL,
     JUDGMENT_SCORER_EXISTS_API_URL,
+    JUDGMENT_CUSTOM_SCORER_UPLOAD_API_URL,
     JUDGMENT_DATASETS_APPEND_TRACES_API_URL,
-    JUDGMENT_CHECK_EXAMPLE_KEYS_API_URL,
 )
 from judgeval.common.api.constants import (
     TraceFetchPayload,
@@ -45,16 +43,14 @@ from judgeval.common.api.constants import (
     DeleteEvalRunRequestBody,
     EvalLogPayload,
     EvalStatusPayload,
-    CheckExperimentTypePayload,
-    EvalRunNameExistsPayload,
     ScorerSavePayload,
     ScorerFetchPayload,
     ScorerExistsPayload,
-    CheckExampleKeysPayload,
+    CustomScorerUploadPayload,
+    CustomScorerTemplateResponse,
 )
 from judgeval.utils.requests import requests
-
-import orjson
+from judgeval.common.api.json_encoder import json_encoder
 
 
 class JudgmentAPIException(exceptions.HTTPError):
@@ -98,22 +94,28 @@ class JudgmentApiClient:
         method: Literal["POST", "PATCH", "GET", "DELETE"],
         url: str,
         payload: Any,
+        timeout: Optional[Union[float, tuple]] = None,
     ) -> Any:
+        # Prepare request kwargs with optional timeout
+        request_kwargs = self._request_kwargs()
+        if timeout is not None:
+            request_kwargs["timeout"] = timeout
+
         if method == "GET":
             r = requests.request(
                 method,
                 url,
                 params=payload,
                 headers=self._headers(),
-                **self._request_kwargs(),
+                **request_kwargs,
             )
         else:
             r = requests.request(
                 method,
                 url,
-                data=self._serialize(payload),
+                json=json_encoder(payload),
                 headers=self._headers(),
-                **self._request_kwargs(),
+                **request_kwargs,
             )
 
         try:
@@ -187,10 +189,10 @@ class JudgmentApiClient:
         payload: EvalLogPayload = {"results": results, "run": run}
         return self._do_request("POST", JUDGMENT_EVAL_LOG_API_URL, payload)
 
-    def fetch_evaluation_results(self, project_name: str, eval_name: str):
+    def fetch_evaluation_results(self, experiment_run_id: str, project_name: str):
         payload: EvalRunRequestBody = {
             "project_name": project_name,
-            "eval_name": eval_name,
+            "experiment_run_id": experiment_run_id,
         }
         return self._do_request("POST", JUDGMENT_EVAL_FETCH_API_URL, payload)
 
@@ -205,43 +207,21 @@ class JudgmentApiClient:
     def add_to_evaluation_queue(self, payload: Dict[str, Any]):
         return self._do_request("POST", JUDGMENT_ADD_TO_RUN_EVAL_QUEUE_API_URL, payload)
 
-    def get_evaluation_status(self, eval_name: str, project_name: str):
+    def get_evaluation_status(self, experiment_run_id: str, project_name: str):
         payload: EvalStatusPayload = {
-            "eval_name": eval_name,
+            "experiment_run_id": experiment_run_id,
             "project_name": project_name,
             "judgment_api_key": self.api_key,
         }
         return self._do_request("GET", JUDGMENT_GET_EVAL_STATUS_API_URL, payload)
 
-    def check_experiment_type(self, eval_name: str, project_name: str, is_trace: bool):
-        payload: CheckExperimentTypePayload = {
-            "eval_name": eval_name,
-            "project_name": project_name,
-            "judgment_api_key": self.api_key,
-            "is_trace": is_trace,
-        }
-        return self._do_request("POST", JUDGMENT_CHECK_EXPERIMENT_TYPE_API_URL, payload)
-
-    def check_eval_run_name_exists(self, eval_name: str, project_name: str):
-        payload: EvalRunNameExistsPayload = {
-            "eval_name": eval_name,
-            "project_name": project_name,
-            "judgment_api_key": self.api_key,
-        }
-        return self._do_request("POST", JUDGMENT_EVAL_RUN_NAME_EXISTS_API_URL, payload)
-
-    def check_example_keys(self, keys: List[str], eval_name: str, project_name: str):
-        payload: CheckExampleKeysPayload = {
-            "keys": keys,
-            "eval_name": eval_name,
-            "project_name": project_name,
-        }
-        return self._do_request("POST", JUDGMENT_CHECK_EXAMPLE_KEYS_API_URL, payload)
-
-    def save_scorer(self, name: str, prompt: str, options: Optional[dict] = None):
+    def save_scorer(
+        self, name: str, prompt: str, threshold: float, options: Optional[dict] = None
+    ):
         payload: ScorerSavePayload = {
             "name": name,
             "prompt": prompt,
+            "threshold": threshold,
             "options": options,
         }
         try:
@@ -293,6 +273,31 @@ class JudgmentApiClient:
                 request=e.request,
             )
 
+    def upload_custom_scorer(
+        self,
+        scorer_name: str,
+        scorer_code: str,
+        requirements_text: str,
+    ) -> CustomScorerTemplateResponse:
+        """Upload custom scorer to backend"""
+        payload: CustomScorerUploadPayload = {
+            "scorer_name": scorer_name,
+            "scorer_code": scorer_code,
+            "requirements_text": requirements_text,
+        }
+
+        try:
+            # Use longer timeout for custom scorer upload (5 minutes)
+            response = self._do_request(
+                "POST",
+                JUDGMENT_CUSTOM_SCORER_UPLOAD_API_URL,
+                payload,
+                timeout=(10, 300),
+            )
+            return response
+        except JudgmentAPIException as e:
+            raise e
+
     def push_dataset(
         self,
         dataset_alias: str,
@@ -368,16 +373,3 @@ class JudgmentApiClient:
             "verify": True,
             "timeout": 30,
         }
-
-    def _serialize(self, data: Any) -> str:
-        def fallback_encoder(obj):
-            try:
-                return repr(obj)
-            except Exception:
-                try:
-                    return str(obj)
-                except Exception as e:
-                    return f"<Unserializable object of type {type(obj).__name__}: {e}>"
-
-        # orjson returns bytes, so we need to decode to str
-        return orjson.dumps(data, default=fallback_encoder).decode("utf-8")

judgeval/common/api/constants.py

@@ -49,9 +49,9 @@ JUDGMENT_EVAL_DELETE_API_URL = (
 JUDGMENT_EVAL_DELETE_PROJECT_API_URL = f"{ROOT_API}/delete_eval_results_by_project/"
 JUDGMENT_ADD_TO_RUN_EVAL_QUEUE_API_URL = f"{ROOT_API}/add_to_run_eval_queue/"
 JUDGMENT_GET_EVAL_STATUS_API_URL = f"{ROOT_API}/get_evaluation_status/"
-JUDGMENT_CHECK_EXPERIMENT_TYPE_API_URL = f"{ROOT_API}/check_experiment_type/"
-JUDGMENT_EVAL_RUN_NAME_EXISTS_API_URL = f"{ROOT_API}/eval-run-name-exists/"
-JUDGMENT_CHECK_EXAMPLE_KEYS_API_URL = f"{ROOT_API}/check_example_keys/"
+
+# Custom Scorers API
+JUDGMENT_CUSTOM_SCORER_UPLOAD_API_URL = f"{ROOT_API}/build_sandbox_template/"
 
 
 # Evaluation API Payloads
@@ -73,9 +73,9 @@ class EvalLogPayload(TypedDict):
 
 
 class EvalStatusPayload(TypedDict):
-    eval_name: str
-    project_name: str
+    experiment_run_id: str
     judgment_api_key: str
+    project_name: str
 
 
 class CheckExperimentTypePayload(TypedDict):
@@ -162,6 +162,7 @@ JUDGMENT_SCORER_EXISTS_API_URL = f"{ROOT_API}/scorer_exists/"
 class ScorerSavePayload(TypedDict):
     name: str
     prompt: str
+    threshold: float
     options: Optional[dict]
 
 
@@ -171,3 +172,15 @@ class ScorerFetchPayload(TypedDict):
 
 class ScorerExistsPayload(TypedDict):
     name: str
+
+
+class CustomScorerUploadPayload(TypedDict):
+    scorer_name: str
+    scorer_code: str
+    requirements_text: str
+
+
+class CustomScorerTemplateResponse(TypedDict):
+    scorer_name: str
+    status: str
+    message: str

judgeval/common/api/json_encoder.py

@@ -0,0 +1,241 @@
+"""
+
+This is a modified version of https://docs.powertools.aws.dev/lambda/python/2.35.1/api/event_handler/openapi/encoders.html
+
+"""
+
+import dataclasses
+import datetime
+from collections import defaultdict, deque
+from decimal import Decimal
+from enum import Enum
+from pathlib import Path, PurePath
+from re import Pattern
+from types import GeneratorType
+from typing import Any, Callable, Dict, List, Literal, Optional, Tuple, Type, Union
+from uuid import UUID
+
+from pydantic import BaseModel
+from pydantic.types import SecretBytes, SecretStr
+
+
+"""
+This module contains the encoders used by jsonable_encoder to convert Python objects to JSON serializable data types.
+"""
+
+
+def _model_dump(
+    model: BaseModel, mode: Literal["json", "python"] = "json", **kwargs: Any
+) -> Any:
+    return model.model_dump(mode=mode, **kwargs)
+
+
+def json_encoder(
+    obj: Any,
+    custom_serializer: Optional[Callable[[Any], str]] = None,
+) -> Any:
+    """
+    JSON encodes an arbitrary Python object into JSON serializable data types.
+
+    This is a modified version of fastapi.encoders.jsonable_encoder that supports
+    encoding of pydantic.BaseModel objects.
+
+    Parameters
+    ----------
+    obj : Any
+        The object to encode
+    custom_serializer : Callable, optional
+        A custom serializer to use for encoding the object, when everything else fails.
+
+    Returns
+    -------
+    Any
+        The JSON serializable data types
+    """
+    # Pydantic models
+    if isinstance(obj, BaseModel):
+        return _dump_base_model(
+            obj=obj,
+        )
+
+    # Dataclasses
+    if dataclasses.is_dataclass(obj):
+        obj_dict = dataclasses.asdict(obj)
+        return json_encoder(
+            obj_dict,
+        )
+
+    # Enums
+    if isinstance(obj, Enum):
+        return obj.value
+
+    # Paths
+    if isinstance(obj, PurePath):
+        return str(obj)
+
+    # Scalars
+    if isinstance(obj, (str, int, float, type(None))):
+        return obj
+
+    # Dictionaries
+    if isinstance(obj, dict):
+        return _dump_dict(
+            obj=obj,
+        )
+
+    # Sequences
+    if isinstance(obj, (list, set, frozenset, tuple, deque)):
+        return _dump_sequence(
+            obj=obj,
+        )
+
+    # Other types
+    if type(obj) in ENCODERS_BY_TYPE:
+        return ENCODERS_BY_TYPE[type(obj)](obj)
+
+    for encoder, classes_tuple in encoders_by_class_tuples.items():
+        if isinstance(obj, classes_tuple):
+            return encoder(obj)
+
+    # Use custom serializer if present
+    if custom_serializer:
+        return custom_serializer(obj)
+
+    # Default
+    return _dump_other(
+        obj=obj,
+    )
+
+
+def _dump_base_model(
+    *,
+    obj: Any,
+):
+    """
+    Dump a BaseModel object to a dict, using the same parameters as jsonable_encoder
+    """
+    obj_dict = _model_dump(
+        obj,
+        mode="json",
+    )
+    if "__root__" in obj_dict:
+        obj_dict = obj_dict["__root__"]
+
+    return json_encoder(
+        obj_dict,
+    )
+
+
+def _dump_dict(
+    *,
+    obj: Any,
+) -> Dict[str, Any]:
+    """
+    Dump a dict to a dict, using the same parameters as jsonable_encoder
+    """
+    encoded_dict = {}
+    allowed_keys = set(obj.keys())
+    for key, value in obj.items():
+        if key in allowed_keys:
+            encoded_key = json_encoder(
+                key,
+            )
+            encoded_value = json_encoder(
+                value,
+            )
+            encoded_dict[encoded_key] = encoded_value
+    return encoded_dict
+
+
+def _dump_sequence(
+    *,
+    obj: Any,
+) -> List[Any]:
+    """
+    Dump a sequence to a list, using the same parameters as jsonable_encoder
+    """
+    encoded_list = []
+    for item in obj:
+        encoded_list.append(
+            json_encoder(
+                item,
+            ),
+        )
+    return encoded_list
+
+
+def _dump_other(
+    *,
+    obj: Any,
+) -> Any:
+    """
+    Dump an object to a representation without iterating it.
+
+    Avoids calling dict(obj) which can consume iterators/generators or
+    invoke user-defined iteration protocols.
+    """
+    try:
+        return repr(obj)
+    except Exception:
+        return str(obj)
+
+
+def iso_format(o: Union[datetime.date, datetime.time]) -> str:
+    """
+    ISO format for date and time
+    """
+    return o.isoformat()
+
+
+def decimal_encoder(dec_value: Decimal) -> Union[int, float]:
+    """
+    Encodes a Decimal as int of there's no exponent, otherwise float
+
+    This is useful when we use ConstrainedDecimal to represent Numeric(x,0)
+    where an integer (but not int typed) is used. Encoding this as a float
+    results in failed round-tripping between encode and parse.
+
+    >>> decimal_encoder(Decimal("1.0"))
+    1.0
+
+    >>> decimal_encoder(Decimal("1"))
+    1
+    """
+    if dec_value.as_tuple().exponent >= 0:  # type: ignore[operator]
+        return int(dec_value)
+    else:
+        return float(dec_value)
+
+
+ENCODERS_BY_TYPE: Dict[Type[Any], Callable[[Any], Any]] = {
+    bytes: lambda o: o.decode(),
+    datetime.date: iso_format,
+    datetime.datetime: iso_format,
+    datetime.time: iso_format,
+    datetime.timedelta: lambda td: td.total_seconds(),
+    Decimal: decimal_encoder,
+    Enum: lambda o: o.value,
+    frozenset: list,
+    deque: list,
+    GeneratorType: repr,
+    Path: str,
+    Pattern: lambda o: o.pattern,
+    SecretBytes: str,
+    SecretStr: str,
+    set: list,
+    UUID: str,
+}
+
+
+# Generates a mapping of encoders to a tuple of classes that they can encode
+def generate_encoders_by_class_tuples(
+    type_encoder_map: Dict[Any, Callable[[Any], Any]],
+) -> Dict[Callable[[Any], Any], Tuple[Any, ...]]:
+    encoders: Dict[Callable[[Any], Any], Tuple[Any, ...]] = defaultdict(tuple)
+    for type_, encoder in type_encoder_map.items():
+        encoders[encoder] += (type_,)
+    return encoders
+
+
+# Mapping of encoders to a tuple of classes that they can encode
+encoders_by_class_tuples = generate_encoders_by_class_tuples(ENCODERS_BY_TYPE)