Add output evaluation for guardrails (#332)

* add single query input/output guardrails Signed-off-by: Tyler Wilbers <[email protected]> * removed comment Signed-off-by: Tyler Wilbers <[email protected]> --------- Signed-off-by: Tyler Wilbers <[email protected]>
opea-project · Aug 13, 2024 · 62ca5bc · 62ca5bc
1 parent ab8ebc4
commit 62ca5bc
Show file tree

Hide file tree

Showing 4 changed files with 13 additions and 8 deletions.
diff --git a/comps/guardrails/llama_guard/README.md b/comps/guardrails/llama_guard/README.md
@@ -36,7 +36,7 @@ pip install -r requirements.txt
 export HF_TOKEN=${your_hf_api_token}
 export LANGCHAIN_TRACING_V2=true
 export LANGCHAIN_API_KEY=${your_langchain_api_key}
-export LANGCHAIN_PROJECT="opea/gaurdrails"
+export LANGCHAIN_PROJECT="opea/guardrails"
 volume=$PWD/data
 model_id="meta-llama/Meta-Llama-Guard-2-8B"
 docker pull ghcr.io/huggingface/tgi-gaudi:2.0.1

diff --git a/comps/guardrails/llama_guard/guardrails_tgi.py b/comps/guardrails/llama_guard/guardrails_tgi.py
@@ -2,13 +2,14 @@
 # SPDX-License-Identifier: Apache-2.0
 
 import os
+from typing import List, Union
 
 from langchain_community.utilities.requests import JsonRequestsWrapper
 from langchain_huggingface import ChatHuggingFace
 from langchain_huggingface.llms import HuggingFaceEndpoint
 from langsmith import traceable
 
-from comps import ServiceType, TextDoc, opea_microservices, register_microservice
+from comps import GeneratedDoc, ServiceType, TextDoc, opea_microservices, register_microservice
 
 DEFAULT_MODEL = "meta-llama/LlamaGuard-7b"
 
@@ -59,12 +60,17 @@ def get_tgi_service_model_id(endpoint_url, default=DEFAULT_MODEL):
     endpoint="/v1/guardrails",
     host="0.0.0.0",
     port=9090,
-    input_datatype=TextDoc,
+    input_datatype=Union[GeneratedDoc, TextDoc],
     output_datatype=TextDoc,
 )
 @traceable(run_type="llm")
-def safety_guard(input: TextDoc) -> TextDoc:
-    response_input_guard = llm_engine_hf.invoke([{"role": "user", "content": input.text}]).content
+def safety_guard(input: Union[GeneratedDoc, TextDoc]) -> TextDoc:
+    if isinstance(input, GeneratedDoc):
+        messages = [{"role": "user", "content": input.prompt}, {"role": "assistant", "content": input.text}]
+    else:
+        messages = [{"role": "user", "content": input.text}]
+    response_input_guard = llm_engine_hf.invoke(messages).content
+
     if "unsafe" in response_input_guard:
         unsafe_dict = get_unsafe_dict(llm_engine_hf.model_id)
         policy_violation_level = response_input_guard.split("\n")[1].strip()
@@ -75,7 +81,6 @@ def safety_guard(input: TextDoc) -> TextDoc:
         )
     else:
         res = TextDoc(text=input.text)
-
     return res
 
 

diff --git a/comps/guardrails/llama_guard/requirements.txt b/comps/guardrails/llama_guard/requirements.txt
@@ -1,6 +1,7 @@
 docarray[full]
 fastapi
-huggingface_hub
+# Fix for issue with langchain-huggingface not using InferenceClient `base_url` kwarg
+huggingface-hub<=0.24.0
 langchain-community
 langchain-huggingface
 langsmith

diff --git a/tests/test_guardrails_llama_guard.sh b/tests/test_guardrails_llama_guard.sh
@@ -25,7 +25,6 @@ function start_service() {
     sleep 4m
     docker run -d --name="test-comps-guardrails-langchain-service" -p 9090:9090 --ipc=host -e http_proxy=$http_proxy -e https_proxy=$https_proxy -e no_proxy=$no_proxy -e SAFETY_GUARD_MODEL_ID=$SAFETY_GUARD_MODEL_ID -e SAFETY_GUARD_ENDPOINT=$SAFETY_GUARD_ENDPOINT -e HUGGINGFACEHUB_API_TOKEN=$HF_TOKEN opea/guardrails-tgi:comps
     sleep 10s
-
     echo "Microservice started"
 }