Generative AI Use Cases

Introduction

Beside Tensorflow Serving API (/v1) and KServe API (/v2) frontends, the model server supports a range of endpoints for generative use cases (v3). They are extendible using MediaPipe graphs. Currently supported endpoints are:

OpenAI compatible endpoints:

• chat/completions

• completions

• embeddings Cohere Compatible endpoint:

• rerank

OpenAI API Clients

When creating a Python-based client application, you can use OpenAI client library - openai.

Alternatively, it is possible to use just a curl command or requests python library.

Install the Package

pip3 install openai
pip3 install requests
pip3 install cohere

Request chat completions with unary calls

python [OpenAI]

from openai import OpenAI
client = OpenAI(base_url="http://localhost:8000/v3", api_key="unused")
response = client.chat.completions.create(
    model="meta-llama/Llama-2-7b-chat-hf",
    messages=[{"role": "user", "content": "Say this is a test"}],
    stream=False,
)

print(response.choices[0].message)

python [requests]

import requests
payload = {"model": "meta-llama/Llama-2-7b-chat-hf", "messages": [ {"role": "user","content": "Say this is a test" }]}
headers = {"Content-Type": "application/json", "Authorization": "not used"}
response = requests.post("http://localhost:8000/v3/chat/completions", json=payload, headers=headers)
print(response.text)

curl

curl http://localhost:8000/v3/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "meta-llama/Llama-2-7b-chat-hf", "messages": [ {"role": "user","content": "Say this is a test" }]}'

Check LLM quick start and end to end demo of text generation.

Request completions with unary calls

python [OpenAI]

from openai import OpenAI
client = OpenAI(base_url="http://localhost:8000/v3", api_key="unused")
response = client.completions.create(
    model="meta-llama/Llama-2-7b",
    prompt="Say this is a test",
    stream=False,
)

print(response.choices[0].text)

python [requests]

import requests
payload = {"model": "meta-llama/Llama-2-7b", "prompt": "Say this is a test"}
headers = {"Content-Type": "application/json", "Authorization": "not used"}
response = requests.post("http://localhost:8000/v3/completions", json=payload, headers=headers)
print(response.text)

curl

curl http://localhost:8000/v3/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "meta-llama/Llama-2-7b", "prompt": "Say this is a test"}'

Check LLM quick start and end to end demo of text generation.

Request chat completions with streaming

python [OpenAI]

from openai import OpenAI
client = OpenAI(
  base_url="http://localhost:8000/v3",
  api_key="unused"
)

stream = client.chat.completions.create(
    model="meta-llama/Llama-2-7b-chat-hf",
    messages=[{"role": "user", "content": "Say this is a test"}],
    stream=True,
)
for chunk in stream:
    if chunk.choices[0].delta.content is not None:
        print(chunk.choices[0].delta.content, end="")

Check LLM quick start and end to end demo of text generation.

Request completions with streaming

python [OpenAI]

from openai import OpenAI
client = OpenAI(
  base_url="http://localhost:8000/v3",
  api_key="unused"
)

stream = client.completions.create(
    model="meta-llama/Llama-2-7b",
    prompt="Say this is a test",
    stream=True,
)
for chunk in stream:
    if chunk.choices[0].text is not None:
        print(chunk.choices[0].text, end="")

Check LLM quick start and end to end demo of text generation.

Text embeddings

python [OpenAI]

from openai import OpenAI
client = OpenAI(
  base_url="http://localhost:8000/v3",
  api_key="unused"
)
responses = client.embeddings.create(input=['hello world'], model='Alibaba-NLP/gte-large-en-v1.5')
for data in responses.data:
    print(data.embedding)

python [requests]

import requests
payload = {"model": "Alibaba-NLP/gte-large-en-v1.5", "input": "hello world"}
headers = {"Content-Type": "application/json", "Authorization": "not used"}
response = requests.post("http://localhost:8000/v3/embeddings", json=payload, headers=headers)
print(response.text)

curl

curl http://localhost:8000/v3/embeddings \
  -H "Content-Type: application/json" \
  -d '{"model": "Alibaba-NLP/gte-large-en-v1.5", "input": "hello world"}'

Check text embeddings end to end demo.

Cohere Python Client

Clients can use rerank endpoint via cohere python package - cohere.

Just like with openAI endpoints and alternative is in curl command or requests python library.

Install the Package

pip3 install cohere
pip3 install requests

Documents reranking

python [Cohere]

import cohere
client = cohere.Client(base_url='http://localhost:8000/v3', api_key="not_used")
responses = client.rerank(query="Hello",documents=["Welcome","Farewell"], model='BAAI/bge-reranker-large')
for res in responses.results:
    print(res.index, res.relevance_score)

python [requests]

import requests
payload = {"model": "BAAI/bge-reranker-large", "query": "Hello", "documents":["Welcome","Farewell"]}
headers = {"Content-Type": "application/json", "Authorization": "not used"}
response = requests.post("http://localhost:8000/v3/rerank", json=payload, headers=headers)
print(response.text)

curl

curl http://localhost:8000/v3/rerank \
  -H "Content-Type: application/json" \
  -d '{"model": "BAAI/bge-reranker-large", "query": "Hello", "documents":["Welcome","Farewell"]}'

Check documents reranking end to end demo.