judgeval 0.2.0__tar.gz → 0.3.1__tar.gz

{judgeval-0.2.0 → judgeval-0.3.1}/.github/workflows/lint.yaml

@@ -10,20 +10,11 @@ jobs:
     steps:
       - uses: actions/checkout@v4
 
-      - name: Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: '3.11'
-
       - name: Install ruff
         uses: astral-sh/ruff-action@v3
         with:
           args: "--version"
 
-      - name: Install mypy and dependencies
-        run: |
-          pip install mypy types-requests types-PyYAML
-
       - name: Run ruff formatter
         if: always()
         run: ruff format --check .
@@ -31,7 +22,3 @@ jobs:
       - name: Run ruff linter
         if: always()
         run: ruff check .
-
-      - name: Run mypy
-        if: always()
-        run: mypy --explicit-package-bases --ignore-missing-imports .

judgeval-0.3.1/.github/workflows/mypy.yaml

@@ -0,0 +1,25 @@
+name: MyPy Check
+
+on:
+  pull_request:
+    branches: [ main, staging ]
+
+jobs:
+  mypy:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+      
+      - name: Install dependencies
+        run: |
+          pip install uv
+          uv sync --dev
+
+      - name: Run mypy
+        if: always()
+        run: uv run mypy ./src/judgeval/

judgeval-0.3.1/.github/workflows/pre-commit-autoupdate.yaml

@@ -0,0 +1,38 @@
+name: Pre-commit auto-update
+on:
+  schedule:
+    - cron: '0 0 * * 1'  # Weekly on Monday at midnight UTC
+  workflow_dispatch:
+
+jobs:
+  auto-update:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          ref: staging
+
+      - name: Set up Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: '3.11'
+
+      - name: Install and update pre-commit
+        run: |
+          pip install pre-commit
+          pre-commit autoupdate
+
+      - name: Create Pull Request
+        uses: peter-evans/create-pull-request@v7
+        with:
+          commit-message: 'chore: update pre-commit hooks'
+          title: 'chore: update pre-commit hooks'
+          body: |
+            Auto-generated PR to update pre-commit hook versions.
+            
+            Please review the changes and merge if everything looks good.
+            
+            Updated by GitHub Actions on {{ date }}.
+          branch: update-pre-commit-hooks
+          base: staging

judgeval-0.3.1/.pre-commit-config.yaml

@@ -0,0 +1,23 @@
+repos:
+  - repo: https://github.com/astral-sh/uv-pre-commit
+    rev: 0.8.0
+    hooks:
+      - id: uv-lock
+
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    rev: v0.12.4
+    hooks:
+      - id: ruff
+        name: ruff (linter)
+        args: [--fix]
+      - id: ruff-format
+        name: ruff (formatter)
+
+  # - repo: https://github.com/pre-commit/mirrors-mypy
+  #   rev: v1.17.0
+  #   hooks:
+  #     - id: mypy
+  #       language: system
+  #       # These next two lines allow commits even if mypy fails, REMOVE once we fix all mypy errors
+  #       verbose: true
+  #       entry: bash -c 'mypy src/judgeval/ || true'

{judgeval-0.2.0 → judgeval-0.3.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: judgeval
-Version: 0.2.0
+Version: 0.3.1
 Summary: Judgeval Package
 Project-URL: Homepage, https://github.com/JudgmentLabs/judgeval
 Project-URL: Issues, https://github.com/JudgmentLabs/judgeval/issues
@@ -14,6 +14,7 @@ Requires-Dist: anthropic
 Requires-Dist: boto3
 Requires-Dist: datamodel-code-generator>=0.31.1
 Requires-Dist: google-genai
+Requires-Dist: groq>=0.30.0
 Requires-Dist: langchain-anthropic
 Requires-Dist: langchain-core
 Requires-Dist: langchain-huggingface
@@ -22,6 +23,9 @@ Requires-Dist: litellm>=1.61.15
 Requires-Dist: matplotlib>=3.10.3
 Requires-Dist: nest-asyncio
 Requires-Dist: openai
+Requires-Dist: opentelemetry-api>=1.34.1
+Requires-Dist: opentelemetry-sdk>=1.34.1
+Requires-Dist: orjson>=3.9.0
 Requires-Dist: pandas
 Requires-Dist: python-dotenv==1.0.1
 Requires-Dist: python-slugify>=8.0.4
@@ -39,7 +43,7 @@ Description-Content-Type: text/markdown
     Enable self-learning agents with traces, evals, and environment data.
 </div>
 
-## [Docs](https://docs.judgmentlabs.ai/)  •  [Judgment Cloud](https://app.judgmentlabs.ai/register)  • [Self-Host](https://docs.judgmentlabs.ai/documentation/self-hosting/get-started)
+## [Docs](https://docs.judgmentlabs.ai/)  •  [Judgment Cloud](https://app.judgmentlabs.ai/register)  • [Self-Host](https://docs.judgmentlabs.ai/documentation/self-hosting/get-started)  • [Landing Page](https://judgmentlabs.ai/)
 
  [Demo](https://www.youtube.com/watch?v=1S4LixpVbcc) • [Bug Reports](https://github.com/JudgmentLabs/judgeval/issues) • [Changelog](https://docs.judgmentlabs.ai/changelog/2025-04-21)
 
@@ -139,7 +143,7 @@ run_agent("What is the capital of the United States?")
 ```
 You'll see your trace exported to the Judgment Platform:
 
-<p align="center"><img src="assets/trace_demo.png" alt="Judgment Platform Trace Example" width="800" /></p>
+<p align="center"><img src="assets/online_eval.png" alt="Judgment Platform Trace Example" width="1500" /></p>
 
 
 [Click here](https://docs.judgmentlabs.ai/documentation/tracing/introduction) for a more detailed explanation.
@@ -152,9 +156,9 @@ You'll see your trace exported to the Judgment Platform:
 
 |  |  |
 |:---|:---:|
-| <h3>🔍 Tracing</h3>Automatic agent tracing integrated with common frameworks (LangGraph, OpenAI, Anthropic). **Tracks inputs/outputs, agent tool calls, latency, cost, and custom metadata** at every step.<br><br>**Useful for:**<br>• 🐛 Debugging agent runs <br>• 📋 Collecting agent environment data <br>• 🔬 Pinpointing performance bottlenecks| <p align="center"><img src="assets/trace_screenshot.png" alt="Tracing visualization" width="1200"/></p> |
-| <h3>🧪 Evals</h3>Build custom evaluators on top of your agents. Judgeval supports LLM-as-a-judge, manual labeling, and code-based evaluators that connect with our metric-tracking infrastructure. <br><br>**Useful for:**<br>• ⚠️ Unit-testing <br>• 🔬 A/B testing <br>• 🛡️ Online guardrails | <p align="center"><img src="assets/experiments_page.png" alt="Evaluation metrics" width="800"/></p> |
-| <h3>📡 Monitoring</h3>Get Slack alerts for agent failures in production. Add custom hooks to address production regressions.<br><br> **Useful for:** <br>• 📉 Identifying degradation early <br>• 📈 Visualizing performance trends across agent versions and time | <p align="center"><img src="assets/error_analysis_dashboard.png" alt="Monitoring Dashboard" width="1200"/></p> |
+| <h3>🔍 Tracing</h3>Automatic agent tracing integrated with common frameworks (LangGraph, OpenAI, Anthropic). **Tracks inputs/outputs, agent tool calls, latency, cost, and custom metadata** at every step.<br><br>**Useful for:**<br>• 🐛 Debugging agent runs <br>• 📋 Collecting agent environment data <br>• 🔬 Pinpointing performance bottlenecks| <p align="center"><img src="assets/agent_trace_example.png" alt="Tracing visualization" width="1200"/></p> |
+| <h3>🧪 Evals</h3>Build custom evaluators on top of your agents. Judgeval supports LLM-as-a-judge, manual labeling, and code-based evaluators that connect with our metric-tracking infrastructure. <br><br>**Useful for:**<br>• ⚠️ Unit-testing <br>• 🔬 A/B testing <br>• 🛡️ Online guardrails | <p align="center"><img src="assets/test.png" alt="Evaluation metrics" width="800"/></p> |
+| <h3>📡 Monitoring</h3>Get Slack alerts for agent failures in production. Add custom hooks to address production regressions.<br><br> **Useful for:** <br>• 📉 Identifying degradation early <br>• 📈 Visualizing performance trends across agent versions and time | <p align="center"><img src="assets/errors.png" alt="Monitoring Dashboard" width="1200"/></p> |
 | <h3>📊 Datasets</h3>Export traces and test cases to datasets for scaled analysis and optimization. Move datasets to/from Parquet, S3, etc. <br><br>Run evals on datasets as unit tests or to A/B test different agent configurations, enabling continuous learning from production interactions. <br><br> **Useful for:**<br>• 🗃️ Agent environment interaction data for optimization<br>• 🔄 Scaled analysis for A/B tests | <p align="center"><img src="assets/datasets_preview_screenshot.png" alt="Dataset management" width="1200"/></p> |
 
 ## 🏢 Self-Hosting

{judgeval-0.2.0 → judgeval-0.3.1}/README.md

@@ -8,7 +8,7 @@
     Enable self-learning agents with traces, evals, and environment data.
 </div>
 
-## [Docs](https://docs.judgmentlabs.ai/)  •  [Judgment Cloud](https://app.judgmentlabs.ai/register)  • [Self-Host](https://docs.judgmentlabs.ai/documentation/self-hosting/get-started)
+## [Docs](https://docs.judgmentlabs.ai/)  •  [Judgment Cloud](https://app.judgmentlabs.ai/register)  • [Self-Host](https://docs.judgmentlabs.ai/documentation/self-hosting/get-started)  • [Landing Page](https://judgmentlabs.ai/)
 
  [Demo](https://www.youtube.com/watch?v=1S4LixpVbcc) • [Bug Reports](https://github.com/JudgmentLabs/judgeval/issues) • [Changelog](https://docs.judgmentlabs.ai/changelog/2025-04-21)
 
@@ -108,7 +108,7 @@ run_agent("What is the capital of the United States?")
 ```
 You'll see your trace exported to the Judgment Platform:
 
-<p align="center"><img src="assets/trace_demo.png" alt="Judgment Platform Trace Example" width="800" /></p>
+<p align="center"><img src="assets/online_eval.png" alt="Judgment Platform Trace Example" width="1500" /></p>
 
 
 [Click here](https://docs.judgmentlabs.ai/documentation/tracing/introduction) for a more detailed explanation.
@@ -121,9 +121,9 @@ You'll see your trace exported to the Judgment Platform:
 
 |  |  |
 |:---|:---:|
-| <h3>🔍 Tracing</h3>Automatic agent tracing integrated with common frameworks (LangGraph, OpenAI, Anthropic). **Tracks inputs/outputs, agent tool calls, latency, cost, and custom metadata** at every step.<br><br>**Useful for:**<br>• 🐛 Debugging agent runs <br>• 📋 Collecting agent environment data <br>• 🔬 Pinpointing performance bottlenecks| <p align="center"><img src="assets/trace_screenshot.png" alt="Tracing visualization" width="1200"/></p> |
-| <h3>🧪 Evals</h3>Build custom evaluators on top of your agents. Judgeval supports LLM-as-a-judge, manual labeling, and code-based evaluators that connect with our metric-tracking infrastructure. <br><br>**Useful for:**<br>• ⚠️ Unit-testing <br>• 🔬 A/B testing <br>• 🛡️ Online guardrails | <p align="center"><img src="assets/experiments_page.png" alt="Evaluation metrics" width="800"/></p> |
-| <h3>📡 Monitoring</h3>Get Slack alerts for agent failures in production. Add custom hooks to address production regressions.<br><br> **Useful for:** <br>• 📉 Identifying degradation early <br>• 📈 Visualizing performance trends across agent versions and time | <p align="center"><img src="assets/error_analysis_dashboard.png" alt="Monitoring Dashboard" width="1200"/></p> |
+| <h3>🔍 Tracing</h3>Automatic agent tracing integrated with common frameworks (LangGraph, OpenAI, Anthropic). **Tracks inputs/outputs, agent tool calls, latency, cost, and custom metadata** at every step.<br><br>**Useful for:**<br>• 🐛 Debugging agent runs <br>• 📋 Collecting agent environment data <br>• 🔬 Pinpointing performance bottlenecks| <p align="center"><img src="assets/agent_trace_example.png" alt="Tracing visualization" width="1200"/></p> |
+| <h3>🧪 Evals</h3>Build custom evaluators on top of your agents. Judgeval supports LLM-as-a-judge, manual labeling, and code-based evaluators that connect with our metric-tracking infrastructure. <br><br>**Useful for:**<br>• ⚠️ Unit-testing <br>• 🔬 A/B testing <br>• 🛡️ Online guardrails | <p align="center"><img src="assets/test.png" alt="Evaluation metrics" width="800"/></p> |
+| <h3>📡 Monitoring</h3>Get Slack alerts for agent failures in production. Add custom hooks to address production regressions.<br><br> **Useful for:** <br>• 📉 Identifying degradation early <br>• 📈 Visualizing performance trends across agent versions and time | <p align="center"><img src="assets/errors.png" alt="Monitoring Dashboard" width="1200"/></p> |
 | <h3>📊 Datasets</h3>Export traces and test cases to datasets for scaled analysis and optimization. Move datasets to/from Parquet, S3, etc. <br><br>Run evals on datasets as unit tests or to A/B test different agent configurations, enabling continuous learning from production interactions. <br><br> **Useful for:**<br>• 🗃️ Agent environment interaction data for optimization<br>• 🔄 Scaled analysis for A/B tests | <p align="center"><img src="assets/datasets_preview_screenshot.png" alt="Dataset management" width="1200"/></p> |
 
 ## 🏢 Self-Hosting

{judgeval-0.2.0 → judgeval-0.3.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "judgeval"
-version = "0.2.0"
+version = "0.3.1"
 authors = [
     { name="Andrew Li", email="andrew@judgmentlabs.ai" },
     { name="Alex Shan", email="alex@judgmentlabs.ai" },
@@ -33,6 +33,10 @@ dependencies = [
     "matplotlib>=3.10.3",
     "python-slugify>=8.0.4",
     "datamodel-code-generator>=0.31.1",
+    "groq>=0.30.0",
+    "opentelemetry-api>=1.34.1",
+    "opentelemetry-sdk>=1.34.1",
+    "orjson>=3.9.0",
 ]
 
 [project.urls]
@@ -62,6 +66,30 @@ dev = [
     "langgraph>=0.4.3",
     "pre-commit>=4.2.0",
     "types-requests>=2.32.4.20250611",
+    "mypy>=1.17.0",
+    "types-pyyaml>=6.0.12.20250516",
+    "pandas-stubs>=2.3.0.250703",
+    "lxml-stubs>=0.5.1",
+    "types-pygments>=2.19.0.20250715",
+    "types-beautifulsoup4>=4.12.0.20250516",
+    "types-cachetools>=6.1.0.20250717",
+    "types-cffi>=1.17.0.20250523",
+    "types-defusedxml>=0.7.0.20250708",
+    "types-greenlet>=3.2.0.20250417",
+    "types-jsonschema>=4.24.0.20250708",
+    "types-objgraph>=3.6.0.20240907",
+    "types-pexpect>=4.9.0.20250516",
+    "types-protobuf>=6.30.2.20250703",
+    "types-psutil>=7.0.0.20250601",
+    "types-pyopenssl>=24.1.0.20240722",
+    "types-pyasn1>=0.6.0.20250516",
+    "types-regex>=2024.11.6.20250403",
+    "types-reportlab>=4.4.1.20250602",
+    "types-simplejson>=3.20.0.20250326",
+    "types-tensorflow>=2.18.0.20250516",
+    "types-tqdm>=4.67.0.20250516",
+    "types-tree-sitter-languages>=1.10.0.20250530",
+    "types-xmltodict>=0.14.0.20241009",
 ]
 
 [tool.hatch.build]

{judgeval-0.2.0 → judgeval-0.3.1}/src/judgeval/common/api/api.py

@@ -1,4 +1,4 @@
-from typing import Literal, List, Dict, Any
+from typing import Literal, List, Dict, Any, Union
 from requests import exceptions
 from judgeval.common.api.constants import (
     JUDGMENT_TRACES_FETCH_API_URL,
@@ -25,6 +25,8 @@ from judgeval.common.api.constants import (
     JUDGMENT_SCORER_SAVE_API_URL,
     JUDGMENT_SCORER_FETCH_API_URL,
     JUDGMENT_SCORER_EXISTS_API_URL,
+    JUDGMENT_DATASETS_APPEND_TRACES_API_URL,
+    JUDGMENT_CHECK_EXAMPLE_KEYS_API_URL,
 )
 from judgeval.common.api.constants import (
     TraceFetchPayload,
@@ -48,9 +50,12 @@ from judgeval.common.api.constants import (
     ScorerSavePayload,
     ScorerFetchPayload,
     ScorerExistsPayload,
+    CheckExampleKeysPayload,
 )
 from judgeval.utils.requests import requests
 
+import orjson
+
 
 class JudgmentAPIException(exceptions.HTTPError):
     """
@@ -65,7 +70,7 @@ class JudgmentAPIException(exceptions.HTTPError):
         self.request = request
 
     @property
-    def status_code(self) -> int | None:
+    def status_code(self) -> Union[int, None]:
         """Get the HTTP status code from the response."""
         return self.response.status_code if self.response else None
 
@@ -114,8 +119,15 @@ class JudgmentApiClient:
         try:
             r.raise_for_status()
         except exceptions.HTTPError as e:
+            try:
+                detail = r.json().get("detail", "")
+            except Exception:
+                detail = r.text
+
             raise JudgmentAPIException(
-                f"HTTP {r.status_code}: {r.reason}", response=r, request=e.request
+                f"HTTP {r.status_code}: {r.reason}, {detail}",
+                response=r,
+                request=e.request,
             )
 
         return r.json()
@@ -218,6 +230,14 @@ class JudgmentApiClient:
         }
         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: dict):
         payload: ScorerSavePayload = {
             "name": name,
@@ -279,7 +299,7 @@ class JudgmentApiClient:
         project_name: str,
         examples: List[Dict[str, Any]],
         traces: List[Dict[str, Any]],
-        overwrite: bool,
+        overwrite: bool = False,
     ):
         payload: DatasetPushPayload = {
             "dataset_alias": dataset_alias,
@@ -302,6 +322,18 @@ class JudgmentApiClient:
             "POST", JUDGMENT_DATASETS_APPEND_EXAMPLES_API_URL, payload
         )
 
+    def append_traces(
+        self, dataset_alias: str, project_name: str, traces: List[Dict[str, Any]]
+    ):
+        payload: DatasetAppendPayload = {
+            "dataset_alias": dataset_alias,
+            "project_name": project_name,
+            "traces": traces,
+        }
+        return self._do_request(
+            "POST", JUDGMENT_DATASETS_APPEND_TRACES_API_URL, payload
+        )
+
     def pull_dataset(self, dataset_alias: str, project_name: str):
         payload: DatasetPullPayload = {
             "dataset_alias": dataset_alias,
@@ -347,6 +379,5 @@ class JudgmentApiClient:
                 except Exception as e:
                     return f"<Unserializable object of type {type(obj).__name__}: {e}>"
 
-        import json
-
-        return json.dumps(data, default=fallback_encoder)
+        # orjson returns bytes, so we need to decode to str
+        return orjson.dumps(data, default=fallback_encoder).decode("utf-8")

{judgeval-0.2.0 → judgeval-0.3.1}/src/judgeval/common/api/constants.py

@@ -51,6 +51,7 @@ 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/"
 
 
 # Evaluation API Payloads
@@ -90,9 +91,16 @@ class EvalRunNameExistsPayload(TypedDict):
     judgment_api_key: str
 
 
+class CheckExampleKeysPayload(TypedDict):
+    keys: List[str]
+    eval_name: str
+    project_name: str
+
+
 # Datasets API
 JUDGMENT_DATASETS_PUSH_API_URL = f"{ROOT_API}/datasets/push/"
 JUDGMENT_DATASETS_APPEND_EXAMPLES_API_URL = f"{ROOT_API}/datasets/insert_examples/"
+JUDGMENT_DATASETS_APPEND_TRACES_API_URL = f"{ROOT_API}/traces/add_to_dataset/"
 JUDGMENT_DATASETS_PULL_API_URL = f"{ROOT_API}/datasets/pull_for_judgeval/"
 JUDGMENT_DATASETS_DELETE_API_URL = f"{ROOT_API}/datasets/delete/"
 JUDGMENT_DATASETS_EXPORT_JSONL_API_URL = f"{ROOT_API}/datasets/export_jsonl/"
@@ -134,7 +142,7 @@ class DatasetStatsPayload(TypedDict):
 
 
 # Projects API
-JUDGMENT_PROJECT_DELETE_API_URL = f"{ROOT_API}/projects/delete/"
+JUDGMENT_PROJECT_DELETE_API_URL = f"{ROOT_API}/projects/delete_from_judgeval/"
 JUDGMENT_PROJECT_CREATE_API_URL = f"{ROOT_API}/projects/add/"
 
 

{judgeval-0.2.0 → judgeval-0.3.1}/src/judgeval/common/storage/s3_storage.py

@@ -1,6 +1,6 @@
 import os
-import json
 import boto3
+import orjson
 from typing import Optional
 from datetime import datetime, UTC
 from botocore.exceptions import ClientError
@@ -85,8 +85,7 @@ class S3Storage:
         timestamp = datetime.now(UTC).strftime("%Y%m%d_%H%M%S")
         s3_key = f"traces/{project_name}/{trace_id}_{timestamp}.json"
 
-        # Convert trace data to JSON string
-        trace_json = json.dumps(trace_data)
+        trace_json = orjson.dumps(trace_data).decode("utf-8")
 
         self.s3_client.put_object(
             Bucket=self.bucket_name,

{judgeval-0.2.0 → judgeval-0.3.1}/src/judgeval/common/tracer/core.py

@@ -32,6 +32,7 @@ from typing import (
 )
 import types
 
+
 from judgeval.common.tracer.constants import _TRACE_FILEPATH_BLOCKLIST
 
 from judgeval.common.tracer.otel_span_processor import JudgmentSpanProcessor
@@ -45,6 +46,7 @@ from openai.types.chat import ParsedChatCompletion
 from together import Together, AsyncTogether
 from anthropic import Anthropic, AsyncAnthropic
 from google import genai
+from groq import Groq, AsyncGroq
 
 from judgeval.data import Example, Trace, TraceSpan, TraceUsage
 from judgeval.scorers import APIScorerConfig, BaseScorer
@@ -67,6 +69,8 @@ ApiClient: TypeAlias = Union[
     AsyncTogether,
     genai.Client,
     genai.client.AsyncClient,
+    Groq,
+    AsyncGroq,
 ]
 SpanType: TypeAlias = str
 
@@ -79,7 +83,7 @@ class TraceClient:
         tracer: Tracer,
         trace_id: Optional[str] = None,
         name: str = "default",
-        project_name: str | None = None,
+        project_name: Union[str, None] = None,
         enable_monitoring: bool = True,
         enable_evaluations: bool = True,
         parent_trace_id: Optional[str] = None,
@@ -414,8 +418,6 @@ class TraceClient:
                 self.start_time or time.time(), timezone.utc
             ).isoformat(),
             "duration": total_duration,
-            "trace_spans": [span.model_dump() for span in self.trace_spans],
-            "evaluation_runs": [run.model_dump() for run in self.evaluation_runs],
             "offline_mode": self.tracer.offline_mode,
             "parent_trace_id": self.parent_trace_id,
             "parent_name": self.parent_name,
@@ -850,9 +852,9 @@ class Tracer:
 
     def __init__(
         self,
-        api_key: str | None = os.getenv("JUDGMENT_API_KEY"),
-        organization_id: str | None = os.getenv("JUDGMENT_ORG_ID"),
-        project_name: str | None = None,
+        api_key: Union[str, None] = os.getenv("JUDGMENT_API_KEY"),
+        organization_id: Union[str, None] = os.getenv("JUDGMENT_ORG_ID"),
+        project_name: Union[str, None] = None,
         deep_tracing: bool = False,  # Deep tracing is disabled by default
         enable_monitoring: bool = os.getenv("JUDGMENT_MONITORING", "true").lower()
         == "true",
@@ -905,8 +907,8 @@ class Tracer:
             self.class_identifiers: Dict[
                 str, str
             ] = {}  # Dictionary to store class identifiers
-            self.span_id_to_previous_span_id: Dict[str, str | None] = {}
-            self.trace_id_to_previous_trace: Dict[str, TraceClient | None] = {}
+            self.span_id_to_previous_span_id: Dict[str, Union[str, None]] = {}
+            self.trace_id_to_previous_trace: Dict[str, Union[TraceClient, None]] = {}
             self.current_span_id: Optional[str] = None
             self.current_trace: Optional[TraceClient] = None
             self.trace_across_async_contexts: bool = trace_across_async_contexts
@@ -958,7 +960,9 @@ class Tracer:
             self.enable_monitoring = False
             self.enable_evaluations = False
 
-    def set_current_span(self, span_id: str) -> Optional[contextvars.Token[str | None]]:
+    def set_current_span(
+        self, span_id: str
+    ) -> Optional[contextvars.Token[Union[str, None]]]:
         self.span_id_to_previous_span_id[span_id] = self.current_span_id
         self.current_span_id = span_id
         Tracer.current_span_id = span_id
@@ -981,7 +985,7 @@ class Tracer:
 
     def reset_current_span(
         self,
-        token: Optional[contextvars.Token[str | None]] = None,
+        token: Optional[contextvars.Token[Union[str, None]]] = None,
         span_id: Optional[str] = None,
     ):
         try:
@@ -997,7 +1001,7 @@ class Tracer:
 
     def set_current_trace(
         self, trace: TraceClient
-    ) -> Optional[contextvars.Token[TraceClient | None]]:
+    ) -> Optional[contextvars.Token[Union[TraceClient, None]]]:
         """
         Set the current trace context in contextvars
         """
@@ -1030,7 +1034,7 @@ class Tracer:
 
     def reset_current_trace(
         self,
-        token: Optional[contextvars.Token[TraceClient | None]] = None,
+        token: Optional[contextvars.Token[Union[TraceClient, None]]] = None,
         trace_id: Optional[str] = None,
     ):
         try:
@@ -1046,7 +1050,7 @@ class Tracer:
 
     @contextmanager
     def trace(
-        self, name: str, project_name: str | None = None
+        self, name: str, project_name: Union[str, None] = None
     ) -> Generator[TraceClient, None, None]:
         """Start a new trace context using a context manager"""
         trace_id = str(uuid.uuid4())
@@ -1692,25 +1696,31 @@ def wrap(
         return wrapper
 
     if isinstance(client, (OpenAI)):
-        client.chat.completions.create = wrapped(original_create)
-        client.responses.create = wrapped(original_responses_create)
-        client.beta.chat.completions.parse = wrapped(original_beta_parse)
+        setattr(client.chat.completions, "create", wrapped(original_create))
+        setattr(client.responses, "create", wrapped(original_responses_create))
+        setattr(client.beta.chat.completions, "parse", wrapped(original_beta_parse))
     elif isinstance(client, (AsyncOpenAI)):
-        client.chat.completions.create = wrapped_async(original_create)
-        client.responses.create = wrapped_async(original_responses_create)
-        client.beta.chat.completions.parse = wrapped_async(original_beta_parse)
+        setattr(client.chat.completions, "create", wrapped_async(original_create))
+        setattr(client.responses, "create", wrapped_async(original_responses_create))
+        setattr(
+            client.beta.chat.completions, "parse", wrapped_async(original_beta_parse)
+        )
     elif isinstance(client, (Together)):
-        client.chat.completions.create = wrapped(original_create)
+        setattr(client.chat.completions, "create", wrapped(original_create))
     elif isinstance(client, (AsyncTogether)):
-        client.chat.completions.create = wrapped_async(original_create)
+        setattr(client.chat.completions, "create", wrapped_async(original_create))
     elif isinstance(client, (Anthropic)):
-        client.messages.create = wrapped(original_create)
+        setattr(client.messages, "create", wrapped(original_create))
     elif isinstance(client, (AsyncAnthropic)):
-        client.messages.create = wrapped_async(original_create)
+        setattr(client.messages, "create", wrapped_async(original_create))
     elif isinstance(client, (genai.Client)):
-        client.models.generate_content = wrapped(original_create)
+        setattr(client.models, "generate_content", wrapped(original_create))
     elif isinstance(client, (genai.client.AsyncClient)):
-        client.models.generate_content = wrapped_async(original_create)
+        setattr(client.models, "generate_content", wrapped_async(original_create))
+    elif isinstance(client, (Groq)):
+        setattr(client.chat.completions, "create", wrapped(original_create))
+    elif isinstance(client, (AsyncGroq)):
+        setattr(client.chat.completions, "create", wrapped_async(original_create))
 
     return client
 
@@ -1745,6 +1755,8 @@ def _get_client_config(
             None,
             client.beta.chat.completions.parse,
         )
+    elif isinstance(client, (Groq, AsyncGroq)):
+        return "GROQ_API_CALL", client.chat.completions.create, None, None, None
     elif isinstance(client, (Together, AsyncTogether)):
         return "TOGETHER_API_CALL", client.chat.completions.create, None, None, None
     elif isinstance(client, (Anthropic, AsyncAnthropic)):
@@ -1783,9 +1795,17 @@ def _format_output_data(
     if isinstance(client, (OpenAI, AsyncOpenAI)):
         if isinstance(response, ChatCompletion):
             model_name = response.model
-            prompt_tokens = response.usage.prompt_tokens
-            completion_tokens = response.usage.completion_tokens
-            cache_read_input_tokens = response.usage.prompt_tokens_details.cached_tokens
+            prompt_tokens = response.usage.prompt_tokens if response.usage else 0
+            completion_tokens = (
+                response.usage.completion_tokens if response.usage else 0
+            )
+            cache_read_input_tokens = (
+                response.usage.prompt_tokens_details.cached_tokens
+                if response.usage
+                and response.usage.prompt_tokens_details
+                and response.usage.prompt_tokens_details.cached_tokens
+                else 0
+            )
 
             if isinstance(response, ParsedChatCompletion):
                 message_content = response.choices[0].message.parsed
@@ -1793,10 +1813,19 @@ def _format_output_data(
                 message_content = response.choices[0].message.content
         elif isinstance(response, Response):
             model_name = response.model
-            prompt_tokens = response.usage.input_tokens
-            completion_tokens = response.usage.output_tokens
-            cache_read_input_tokens = response.usage.input_tokens_details.cached_tokens
-            message_content = "".join(seg.text for seg in response.output[0].content)
+            prompt_tokens = response.usage.input_tokens if response.usage else 0
+            completion_tokens = response.usage.output_tokens if response.usage else 0
+            cache_read_input_tokens = (
+                response.usage.input_tokens_details.cached_tokens
+                if response.usage and response.usage.input_tokens_details
+                else 0
+            )
+            if hasattr(response.output[0], "content"):
+                message_content = "".join(
+                    seg.text
+                    for seg in response.output[0].content
+                    if hasattr(seg, "text")
+                )
 
         # Note: LiteLLM seems to use cache_read_input_tokens to calculate the cost for OpenAI
     elif isinstance(client, (Together, AsyncTogether)):
@@ -1821,6 +1850,11 @@ def _format_output_data(
         cache_read_input_tokens = response.usage.cache_read_input_tokens
         cache_creation_input_tokens = response.usage.cache_creation_input_tokens
         message_content = response.content[0].text
+    elif isinstance(client, (Groq, AsyncGroq)):
+        model_name = "groq/" + response.model
+        prompt_tokens = response.usage.prompt_tokens
+        completion_tokens = response.usage.completion_tokens
+        message_content = response.choices[0].message.content
     else:
         judgeval_logger.warning(f"Unsupported client type: {type(client)}")
         return None, None