Single node on-prem deployment with TGI on Nvidia gpu¶

This section covers single-node on-prem deployment of the ChatQnA example using the TGI LLM service. There are several ways to enable RAG with vectordb and LLM models, but this tutorial will be covering how to build an end-to-end ChatQnA pipeline with the Redis vector database and meta-llama/Meta-Llama-3-8B-Instruct model deployed on NVIDIA GPUs.

Overview¶

The OPEA GenAIComps microservices used to deploy a single node vLLM or TGI megaservice solution for ChatQnA are listed below:

Data Prep
Embedding
Retriever
Reranking
LLM with TGI

This solution is designed to demonstrate the use of Redis vectorDB for RAG and the Meta-Llama-3-8B-Instruct model for LLM inference on NVIDIA GPUs. The steps will involve setting up Docker containers, using a sample Nike dataset in PDF format, and posing a question about Nike to receive a response. Although multiple versions of the UI can be deployed, this tutorial will focus solely on the default version.

Prerequisites¶

Set up a workspace and clone the GenAIExamples GitHub repo.

export WORKSPACE=<Path>
cd $WORKSPACE
git clone https://github.com/opea-project/GenAIExamples.git # GenAIExamples

(Optional) It is recommended to use a stable release version by setting RELEASE_VERSION to a number only (i.e. 1.0, 1.1, etc) and checkout that version using the tag. Otherwise, by default, the main branch with the latest updates will be used.

export RELEASE_VERSION=<Release_Version> # Set desired release version - number only
cd GenAIExamples
git checkout tags/v${RELEASE_VERSION}
cd ..

Set up a HuggingFace account and generate a user access token. Request access to the meta-llama/Meta-Llama-3-8B-Instruct model.

Set the HUGGINGFACEHUB_API_TOKEN environment variable to the value of the Hugging Face token by executing the following command:

export HUGGINGFACEHUB_API_TOKEN="Your_Huggingface_API_Token"

The example requires setting the host_ip to “localhost” to deploy the microservices on endpoints enabled with ports.

export host_ip="localhost"

Set the NGINX port.

# Example: NGINX_PORT=80
export NGINX_PORT=<Nginx_Port>

For machines behind a firewall, set up the proxy environment variables:

export http_proxy="Your_HTTP_Proxy"
export https_proxy="Your_HTTPs_Proxy"
# Example: no_proxy="localhost, 127.0.0.1, 192.168.1.1"
export no_proxy="Your_No_Proxy",chatqna-ui-server,chatqna-backend-server,dataprep-redis-service,tei-embedding-service,retriever,tei-reranking-service,tgi-service

Use Case Setup¶

ChatQnA will utilize the following GenAIComps services and associated tools. The tools and models listed in the table can be configured via environment variables in either the set_env.sh script or the compose.yaml file.

TGI

use case components	Tools	Model	Service Type
Data Prep	LangChain	NA	OPEA Microservice
VectorDB	Redis	NA	Open source service
Embedding	TEI	BAAI/bge-base-en-v1.5	OPEA Microservice
Reranking	TEI	BAAI/bge-reranker-base	OPEA Microservice
LLM	TGI	meta-llama/Meta-Llama-3-8B-Instruct	OPEA Microservice
UI		NA	Gateway Service

Set the necessary environment variables to set up the use case. To swap out models, modify set_env.sh before running it. For example, the environment variable LLM_MODEL_ID can be changed to another model by specifying the HuggingFace model card ID.

cd $WORKSPACE/GenAIExamples/ChatQnA/docker_compose/nvidia/gpu
source ./set_env.sh

Deploy the Use Case¶

Run docker compose with the provided YAML file to start all the services mentioned above as containers.

TGI

docker compose -f compose.yaml up -d

Check Env Variables¶

After running docker compose, check for warning messages for environment variables that are NOT set. Address them if needed.

TGI

ubuntu@nvidia-vm:~/GenAIExamples/ChatQnA/docker_compose/nvidia/gpu$ docker compose -f ./compose.yaml up -d
WARN[0000] The "LANGCHAIN_API_KEY" variable is not set. Defaulting to a blank string.
WARN[0000] The "LANGCHAIN_TRACING_V2" variable is not set. Defaulting to a blank string.
WARN[0000] The "LANGCHAIN_API_KEY" variable is not set. Defaulting to a blank string.
WARN[0000] The "LANGCHAIN_TRACING_V2" variable is not set. Defaulting to a blank string.
WARN[0000] The "LANGCHAIN_API_KEY" variable is not set. Defaulting to a blank string.
WARN[0000] The "LANGCHAIN_TRACING_V2" variable is not set. Defaulting to a blank string.
WARN[0000] The "LANGCHAIN_API_KEY" variable is not set. Defaulting to a blank string.
WARN[0000] The "LANGCHAIN_TRACING_V2" variable is not set. Defaulting to a blank string.
WARN[0000] /home/ubuntu/GenAIExamples/ChatQnA/docker_compose/nvidia/gpu/compose.yaml: `version` is obsolete

Check Container Statuses¶

Check if all the containers launched via docker compose are running i.e. each container’s STATUS is Up and Healthy.

Run this command to see this info:

docker ps -a

Sample output:

TGI

CONTAINER ID   IMAGE                                                   COMMAND                  CREATED        STATUS        PORTS                                                                                  NAMES
3b5fa9a722da   opea/chatqna-ui:latest                                  "docker-entrypoint.s…"   32 hours ago   Up 2 hours   0.0.0.0:5173->5173/tcp, :::5173->5173/tcp                                              chatqna-ui-server
d3b37f3d1faa   opea/chatqna:latest                                     "python chatqna.py"      32 hours ago   Up 2 hours   0.0.0.0:8888->8888/tcp, :::8888->8888/tcp                                              chatqna-backend-server
b3e1388fa2ca   opea/reranking-tei:latest                               "python reranking_te…"   32 hours ago   Up 2 hours   0.0.0.0:8000->8000/tcp, :::8000->8000/tcp                                              reranking-tei-server
24a240f8ad1c   opea/retriever-redis:latest                             "python retriever_re…"   32 hours ago   Up 2 hours   0.0.0.0:7000->7000/tcp, :::7000->7000/tcp                                              retriever-redis-server
9c0d2a2553e8   opea/embedding-tei:latest                               "python embedding_te…"   32 hours ago   Up 2 hours   0.0.0.0:6000->6000/tcp, :::6000->6000/tcp                                              embedding-tei-server
24cae0db1a70   opea/llm-tgi:latest                                    "bash entrypoint.sh"     32 hours ago   Up 2 hours   0.0.0.0:9000->9000/tcp, :::9000->9000/tcp                                              llm-tgi-server
ea3986c3cf82   opea/dataprep-redis:latest                              "python prepare_doc_…"   32 hours ago   Up 2 hours   0.0.0.0:6007->6007/tcp, :::6007->6007/tcp                                              dataprep-redis-server
e10dd14497a8   redis/redis-stack:7.2.0-v9                              "/entrypoint.sh"         32 hours ago   Up 2 hours   0.0.0.0:6379->6379/tcp, :::6379->6379/tcp, 0.0.0.0:8001->8001/tcp, :::8001->8001/tcp   redis-vector-db
79276cf45a47   ghcr.io/huggingface/text-embeddings-inference:cpu-1.2   "text-embeddings-rou…"   32 hours ago   Up 2 hours   0.0.0.0:8090->80/tcp, :::8090->80/tcp                                                  tei-embedding-server
4943e5f6cd80   ghcr.io/huggingface/text-embeddings-inference:cpu-1.2   "text-embeddings-rou…"   32 hours ago   Up 2 hours   0.0.0.0:8808->80/tcp, :::8808->80/tcp                                                  tei-reranking-server

Each docker container’s log can also be checked using:

docker logs <CONTAINER_ID OR CONTAINER_NAME>

Validate Microservices¶

This section will guide through the various methods for interacting with the deployed microservices.

TEI Embedding Service¶

The TEI embedding service takes in a string as input, embeds the string into a vector of a specific length determined by the embedding model, and returns this vector.

curl ${host_ip}:8090/embed \
    -X POST \
    -d '{"inputs":"What is Deep Learning?"}' \
    -H 'Content-Type: application/json'

In this example, the embedding model used is BAAI/bge-base-en-v1.5, which has a vector size of 768. Therefore, the output of the curl command is a vector of length 768.

Retriever Microservice¶

To consume the retriever microservice, generate a mock embedding vector with a Python script. The length of the embedding vector is determined by the embedding model. The model is set with the environment variable EMBEDDING_MODEL_ID=”BAAI/bge-base-en-v1.5”, which has a vector size of 768.

Check the vector dimension of the embedding model used and set your_embedding dimension equal to it.

export your_embedding=$(python3 -c "import random; embedding = [random.uniform(-1, 1) for _ in range(768)]; print(embedding)")

curl http://${host_ip}:7000/v1/retrieval \
  -X POST \
  -d "{\"text\":\"test\",\"embedding\":${your_embedding}}" \
  -H 'Content-Type: application/json'

The output of the retriever microservice comprises of the a unique id for the request, initial query or the input to the retrieval microservice, a list of top n retrieved documents relevant to the input query, and top_n where n refers to the number of documents to be returned.

The output is retrieved text that is relevant to the input data:

{"id":"b16024e140e78e39a60e8678622be630","retrieved_docs":[],"initial_query":"test","top_n":1,"metadata":[]}

TEI Reranking Service¶

The TEI Reranking Service reranks the documents returned by the retrieval service. It consumes the query and list of documents and returns the document index in decreasing order of the similarity score. The document corresponding to the index with the highest score is the most relevant document for the input query.

curl http://${host_ip}:8808/rerank \
    -X POST \
    -d '{"query":"What is Deep Learning?", "texts": ["Deep Learning is not...", "Deep learning is..."]}' \
    -H 'Content-Type: application/json'

Sample output:

[{"index":1,"score":0.94238955},{"index":0,"score":0.120219156}]

TGI Service¶

During the initial startup, this service will take a few minutes to download the model files and complete the warm-up process. Once this is finished, the service will be ready for use.

TGI

Run the command below to check whether the LLM service is ready. The output should be “INFO text_generation_router::server: router/src/server.rs:2311: Connected”

docker logs tgi-service | grep Connected

Run the command below to use the TGI service to generate text for the input prompt. Sample output is also shown.

curl http://${host_ip}:8008/generate \
  -X POST \
  -d '{"inputs":"What is Deep Learning?", \
     "parameters":{"max_new_tokens":17, "do_sample": true}}' \
  -H 'Content-Type: application/json'

{"id":"chatcmpl-cc4300a173af48989cac841f54ebca09","object":"chat.completion","created":1743553002,"model":"meta-llama/Meta-Llama-3-8B-Instruct","choices":[{"index":0,"message":{"role":"assistant","content":"Deep learning is a subfield of machine learning that is inspired by the structure and function","tool_calls":[]},"logprobs":null,"finish_reason":"length","stop_reason":null}],"usage":{"prompt_tokens":15,"total_tokens":32,"completion_tokens":17,"prompt_tokens_details":null},"prompt_logprobs":null}

Dataprep Microservice¶

The knowledge base can be updated using the dataprep microservice, which extracts text from a variety of data sources, chunks the data, and embeds each chunk using the embedding microservice. Finally, the embedded vectors are stored in the Redis vector database.

nke-10k-2023.pdf is Nike’s annual report on a form 10-K. Run this command to download the file:

wget https://github.com/opea-project/GenAIComps/blob/v1.1/comps/retrievers/redis/data/nke-10k-2023.pdf

Upload the file:

curl -X POST "http://${host_ip}:6007/v1/dataprep" \
     -H "Content-Type: multipart/form-data" \
     -F "files=@./nke-10k-2023.pdf"

HTTP links can also be added to the knowledge base. This command adds the opea.dev website.

curl -X POST "http://${host_ip}:6007/v1/dataprep" \
     -H "Content-Type: multipart/form-data" \
     -F 'link_list=["https://opea.dev"]'

The list of uploaded files can be retrieved using this command:

curl -X POST "http://${host_ip}:6007/v1/dataprep/get_file" \
     -H "Content-Type: application/json"

To delete the file or link, use the following commands:

Delete link¶

# The dataprep service will add a .txt postfix for link file

curl -X POST "http://${host_ip}:6007/v1/dataprep/delete_file" \
     -d '{"file_path": "https://opea.dev.txt"}' \
     -H "Content-Type: application/json"

Delete file¶

curl -X POST "http://${host_ip}:6007/v1/dataprep/delete_file" \
     -d '{"file_path": "nke-10k-2023.pdf"}' \
     -H "Content-Type: application/json"

Delete all uploaded files and links¶

curl -X POST "http://${host_ip}:6007/v1/dataprep/delete_file" \
     -d '{"file_path": "all"}' \
     -H "Content-Type: application/json"

ChatQnA MegaService¶

This will ensure the megaservice is working properly.

curl http://${host_ip}:8888/v1/chatqna -H "Content-Type: application/json" -d '{
     "messages": "What is the revenue of Nike in 2023?"
     }'

Here is the output for reference:

data: b'\n'
data: b'An'
data: b'swer'
data: b':'
data: b' In'
data: b' fiscal'
data: b' '
data: b'2'
data: b'0'
data: b'2'
data: b'3'
data: b','
data: b' N'
data: b'I'
data: b'KE'
data: b','
data: b' Inc'
data: b'.'
data: b' achieved'
data: b' record'
data: b' Rev'
data: b'en'
data: b'ues'
data: b' of'
data: b' $'
data: b'5'
data: b'1'
data: b'.'
data: b'2'
data: b' billion'
data: b'.'
data: b'</s>'
data: [DONE]

NGINX Service¶

This will ensure the NGINX ervice is working properly.

curl http://${host_ip}:${NGINX_PORT}/v1/chatqna \
    -H "Content-Type: application/json" \
    -d '{"messages": "What is the revenue of Nike in 2023?"}'

The output will be similar to that of the ChatQnA megaservice.

Launch UI¶

Basic UI¶

To access the frontend, open the following URL in your browser: http://${host_ip}:${NGINX_PORT}. By default, the UI runs on port 5173 internally. If you prefer to use a different to access the frontend by modifying the port mapping in the compose.yaml file as shown below:

  chatqna-ui-server:
    image: opea/chatqna-ui:${TAG:-latest}
    ...
    ports:
      - "YOUR_HOST_PORT:5173" # Change YOUR_HOST_PORT to the desired port

After making this change, rebuild and restart the containers for the change to take effect.

Stop the Services¶

Navigate to the docker compose directory for this hardware platform.

cd $WORKSPACE/GenAIExamples/ChatQnA/docker_compose/nvidia/gpu

To stop and remove all the containers, use the command below:

TGI

docker compose -f compose.yaml down