[Bug] Default ChatQnA service gives incorrect output for seed question when testing with curl

[devpramod]

Priority

P2-High

OS type

Ubuntu

Hardware type

Xeon-SPR

Installation method

• Pull docker images from hub.docker.com
• Build docker images from source

Deploy method

• Docker compose
• Docker
• Kubernetes
• Helm

Running nodes

Single Node

What's the version?

https://hub.docker.com/layers/opea/chatqna/latest/images/sha256-b6fe31dd33dc819054b6ae850ef45e1572d62e80305164759bed852638a0c691?context=explore

Description

When using the default compose.yaml file for Xeon on IBM cloud ( bx3d-16x80 instance on IBM Cloud. It has 16 vCPU, 80GB RAM and 100GB boot volume), the output for the question What is the revenue of Nike in 2023? is not as expected, or close to accepted.
Also tried the question what is the capital city of France?

Attaching the output logs:

Question: What is the revenue of Nike in 2023?
default_output.txt

Question: What is the capital of France?
default_output_paris.txt

if you limit the tokens, the output is just truncated and the user doesn't get the answer:

Question: What is the capital of France? (max tokens=10)
truncated.txt

Reproduce steps

git clone https://github.com/opea-project/GenAIExamples.git
cd GenAIExamples/ChatQnA

cd docker_compose/intel/cpu/xeon/
source set_env.sh

docker compose up -d

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

Raw log

No response

[lvliang-intel]

@devpramod,
could you please try again with latest docker image?