Updated the Bedrock rag template (#12462)

Updates the bedrock rag template.
- Removes pinecone and replaces with FAISS as the vector store
- Fixes the environment variables, setting defaults
- Adds a `main.py` test file quick sanity testing
- Updates README.md with correct instructions

templates/rag-aws-bedrock/README.md

@@ -4,26 +4,24 @@ AWS Bedrock is a managed serve that offers a set of foundation models.
 
 Here we will use `Anthropic Claude` for text generation and `Amazon Titan` for text embedding.
 
-We will use Pinecode as our vectorstore.
+We will use FAISS as our vectorstore.
 
 (See [this notebook](https://github.com/aws-samples/amazon-bedrock-workshop/blob/main/03_QuestionAnswering/01_qa_w_rag_claude.ipynb) for additional context on the RAG pipeline.)
 
-(See [this notebook](https://github.com/aws-samples/amazon-bedrock-workshop/blob/58f238a183e7e629c9ae11dd970393af4e64ec44/00_Intro/bedrock_boto3_setup.ipynb#Prerequisites) for additional context on setup.)
+Code here uses the `boto3` library to connect with the Bedrock service. See [this page](https://boto3.amazonaws.com/v1/documentation/api/latest/guide/quickstart.html#configuration) for setting up and configuring boto3 to work with an AWS account. 
 
-##  Pinecone
+##  FAISS
 
-This connects to a hosted Pinecone vectorstore.
+You need to install the `faiss-cpu` package to work with the FAISS vector store.
 
-Be sure that you have set a few env variables in `chain.py`:
+```bash
+pip install faiss-cpu
+```
 
-* `PINECONE_API_KEY`
-* `PINECONE_ENV`
-* `index_name`
 
 ##  LLM and Embeddings
 
-Be sure to set AWS enviorment variables:
+The code assumes that you are working with the `default` AWS profile and `us-east-1` region. If not, specify these environment variables to reflect the correct region and AWS profile.
 
 * `AWS_DEFAULT_REGION`
 * `AWS_PROFILE`
-* `BEDROCK_ASSUME_ROLE`

templates/rag-aws-bedrock/main.py

@@ -0,0 +1,6 @@
+from rag_aws_bedrock.chain import chain
+
+if __name__ == "__main__":
+    query = "What is this data about?"
+
+    print(chain.invoke(query))

templates/rag-aws-bedrock/pyproject.toml

@@ -10,10 +10,9 @@ python = ">=3.8.1,<4.0"
 langchain = ">=0.0.313, <0.1"
 openai = ">=0.28.1"
 tiktoken = ">=0.5.1"
-pinecone-client = ">=2.2.4"
+faiss-cpu = ">=1.7.4"
 boto3 = ">=1.28.57" 
 awscli = ">=1.29.57"
-botocore = ">=1.31.57"
 
 [tool.langserve]
 export_module = "rag_aws_bedrock"

templates/rag-aws-bedrock/rag_aws_bedrock/chain.py

@@ -5,56 +5,31 @@ from langchain.llms.bedrock import Bedrock
 from langchain.prompts import ChatPromptTemplate
 from langchain.schema.output_parser import StrOutputParser
 from langchain.schema.runnable import RunnableParallel, RunnablePassthrough
-from langchain.vectorstores import Pinecone
+from langchain.vectorstores import FAISS
-from utils import bedrock
 
-if os.environ.get("PINECONE_API_KEY", None) is None:
+# Get region and profile from env
-    raise Exception("Missing `PINECONE_API_KEY` environment variable.")
+region = os.environ.get("AWS_DEFAULT_REGION", "us-east-1")
-
+profile = os.environ.get("AWS_PROFILE", "default")
-if os.environ.get("PINECONE_ENVIRONMENT", None) is None:
-    raise Exception("Missing `PINECONE_ENVIRONMENT` environment variable.")
-
-if os.environ.get("AWS_DEFAULT_REGION", None) is None:
-    raise Exception("Missing `AWS_DEFAULT_REGION` environment variable.")
-
-if os.environ.get("AWS_PROFILE", None) is None:
-    raise Exception("Missing `AWS_PROFILE` environment variable.")
-
-if os.environ.get("BEDROCK_ASSUME_ROLE", None) is None:
-    raise Exception("Missing `BEDROCK_ASSUME_ROLE` environment variable.")
-
-PINECONE_INDEX_NAME = os.environ.get("PINECONE_INDEX", "langchain-test")
-
-### Ingest code - you may need to run this the first time
-# Load
-# from langchain.document_loaders import WebBaseLoader
-# loader = WebBaseLoader("https://lilianweng.github.io/posts/2023-06-23-agent/")
-# data = loader.load()
-
-# # Split
-# from langchain.text_splitter import RecursiveCharacterTextSplitter
-# text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=0)
-# all_splits = text_splitter.split_documents(data)
-
-# # Add to vectorDB
-# vectorstore = Pinecone.from_documents(
-#     documents=all_splits, embedding=OpenAIEmbeddings(), index_name=PINECONE_INDEX_NAME
-# )
-# retriever = vectorstore.as_retriever()
 
 # Set LLM and embeddings
-boto3_bedrock = bedrock.get_bedrock_client(
+model = Bedrock(
-    assumed_role=os.environ.get("BEDROCK_ASSUME_ROLE", None),
+    model_id="anthropic.claude-v2",
-    region=os.environ.get("AWS_DEFAULT_REGION", None)
+    region_name=region,
+    credentials_profile_name=profile,
+    model_kwargs={'max_tokens_to_sample':200}
+)
+bedrock_embeddings = BedrockEmbeddings(
+    model_id="amazon.titan-embed-text-v1"
 )
-model = Bedrock(model_id="anthropic.claude-v2", 
-                client=boto3_bedrock, 
-                model_kwargs={'max_tokens_to_sample':200})
-bedrock_embeddings = BedrockEmbeddings(model_id="amazon.titan-embed-text-v1", 
-                                       client=boto3_bedrock)
 
-# Set vectostore
+# Add to vectorDB
-vectorstore = Pinecone.from_existing_index(PINECONE_INDEX_NAME, bedrock_embeddings)
+vectorstore = FAISS.from_texts(
+    ["harrison worked at kensho"],
+    embedding=bedrock_embeddings
+)
+retriever = vectorstore.as_retriever()
+
+# Get retriever from vectorstore
 retriever = vectorstore.as_retriever()
 
 # RAG prompt
@@ -64,6 +39,7 @@ Question: {question}
 """
 prompt = ChatPromptTemplate.from_template(template)
 
+
 # RAG
 chain = (
     RunnableParallel({"context": retriever, "question": RunnablePassthrough()})