minor fixes and proper packaging

Author: saichandrapandraju

detectors/Dockerfile.judge

@@ -4,9 +4,7 @@ RUN microdnf update -y && \
         python-pip python-devel && \
     pip install --upgrade --no-cache-dir pip wheel && \
     microdnf clean all
-RUN pip install --no-cache-dir torch
 
-# FROM icr.io/fm-stack/ubi9-minimal-py39-torch as builder
 FROM base as builder
 
 COPY ./common/requirements.txt .
@@ -20,12 +18,13 @@ FROM builder
 WORKDIR /app
 ARG CACHEBUST=1
 RUN echo "$CACHEBUST"
-COPY ./common /common
-COPY ./llm_judge/app.py /app
-COPY ./llm_judge/detector.py /app
-COPY ./llm_judge/scheme.py /app
+COPY ./common /app/detectors/common
+COPY ./llm_judge/app.py /app/detectors/llm_judge/app.py
+COPY ./llm_judge/detector.py /app/detectors/llm_judge/detector.py
+COPY ./llm_judge/scheme.py /app/detectors/llm_judge/scheme.py
+RUN touch /app/detectors/llm_judge/__init__.py
 
 EXPOSE 8000
-CMD ["uvicorn", "app:app", "--workers", "4", "--host", "0.0.0.0", "--port", "8000", "--log-config", "/common/log_conf.yaml"]
+CMD ["uvicorn", "detectors.llm_judge.app:app", "--workers", "4", "--host", "0.0.0.0", "--port", "8000", "--log-config", "/app/detectors/common/log_conf.yaml"]
 
 # gunicorn main:app --workers 4 --worker-class uvicorn.workers.UvicornWorker --bind 0.0.0.0:8000

detectors/__init__.py

@@ -1,15 +1,12 @@
-import os
-import sys
 from contextlib import asynccontextmanager
 from typing import Annotated, Dict
 
 from fastapi import Header
 from prometheus_fastapi_instrumentator import Instrumentator
-sys.path.insert(0, os.path.abspath(".."))
 
-from common.app import DetectorBaseAPI as FastAPI
-from .detector import LLMJudgeDetector
-from .scheme import (
+from detectors.common.app import DetectorBaseAPI as FastAPI
+from detectors.llm_judge.detector import LLMJudgeDetector
+from detectors.llm_judge.scheme import (
     ContentAnalysisHttpRequest,
     ContentsAnalysisResponse,
     MetricsListResponse,

detectors/llm_judge/app.py

@@ -17,10 +17,10 @@ spec:
       name: guardrails-detector-llm-judge
   containers:
     - name: kserve-container
-      image: quay.io/rh-ee-spandraj/guardrails-detector-judge:latest
+      image: quay.io/spandraj/guardrails-detector-judge:latest
       command:
         - uvicorn
-        - app:app
+        - detectors.llm_judge.app:app
       args:
         - "--workers"
         - "1"
@@ -29,7 +29,7 @@ spec:
         - "--port"
         - "8000"
         - "--log-config"
-        - "/common/log_conf.yaml"
+        - "/app/detectors/common/log_conf.yaml"
       env:
         - name: VLLM_BASE_URL
           value: "http://qwen2-predictor:8080" # <-- Change this to your vLLM URL

detectors/llm_judge/deploy/servingruntime.yaml

@@ -1,13 +1,10 @@
 import os
-import sys
 from typing import List, Dict, Any
 
-sys.path.insert(0, os.path.abspath(".."))
-
 from vllm_judge import Judge, EvaluationResult, BUILTIN_METRICS
 from vllm_judge.exceptions import MetricNotFoundError
-from common.app import logger
-from .scheme import (
+from detectors.common.app import logger
+from detectors.llm_judge.scheme import (
     ContentAnalysisHttpRequest,
     ContentAnalysisResponse,
     ContentsAnalysisResponse,
@@ -24,7 +21,6 @@ def __init__(self) -> None:
 
         # Get configuration from environment
         self.vllm_base_url = os.environ.get("VLLM_BASE_URL")
-        self.vllm_model = os.environ.get("VLLM_MODEL")
 
         if not self.vllm_base_url:
             raise ValueError("VLLM_BASE_URL environment variable is required")

detectors/llm_judge/detector.py

@@ -1,2 +1,2 @@
-vllm-judge>=0.1.5
-pyyaml>=6.0.2
+vllm-judge==0.1.6
+pyyaml==6.0.2