Document Summary LLM Microservice

The Document Summary LLM Microservice leverages LangChain to provide advanced text summarization and Large Language Model (LLM) inference using Text Generation Inference (TGI) on Intel Xeon and Gaudi2 processors. The backend can be configured to use either TGI or vLLM.

Table of Contents

  1. Start Microservice

  2. Consume Microservice

  3. Air-Gapped Environment Usage

Start Microservice

Set Environment Variables

export host_ip=${your_host_ip}
export LLM_ENDPOINT_PORT=8008
export DOCSUM_PORT=9000
export HF_TOKEN=${your_hf_api_token}
export LLM_ENDPOINT="http://${host_ip}:${LLM_ENDPOINT_PORT}"
export LLM_MODEL_ID=${your_hf_llm_model}
export MAX_INPUT_TOKENS=2048
export MAX_TOTAL_TOKENS=4096

MAX_TOTAL_TOKENS must be greater than MAX_INPUT_TOKENS + max_new_tokens + 50 (50 tokens reserved for prompt length).

Build Docker Images

Build Backend LLM Image

For vLLM, refer to vLLM Build Instructions.

TGI does not require additional setup.

Build DocSum Microservice Image

cd ../../../../
docker build -t opea/llm-docsum:latest --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy -f comps/llms/src/doc-summarization/Dockerfile .

Run Docker Service

You can start the service using either the CLI or Docker Compose.

Option A: Run with Docker CLI

  1. Start the backend LLM service (TGI or vLLM).

  2. Start DocSum microservice:

export DocSum_COMPONENT_NAME="OpeaDocSumTgi" # or "OpeaDocSumvLLM"
docker run -d \
    --name="llm-docsum-server" \
    -p 9000:9000 \
    --ipc=host \
    -e http_proxy=$http_proxy \
    -e https_proxy=$https_proxy \
    -e LLM_MODEL_ID=$LLM_MODEL_ID \
    -e LLM_ENDPOINT=$LLM_ENDPOINT \
    -e HF_TOKEN=$HF_TOKEN \
    -e DocSum_COMPONENT_NAME=$DocSum_COMPONENT_NAME \
    -e MAX_INPUT_TOKENS=$MAX_INPUT_TOKENS \
    -e MAX_TOTAL_TOKENS=$MAX_TOTAL_TOKENS \
    opea/llm-docsum:latest

Option B: Run with Docker Compose

export service_name="docsum-tgi"
# Alternatives: "docsum-tgi-gaudi", "docsum-vllm", "docsum-vllm-gaudi"

cd ../../deployment/docker_compose/
docker compose -f compose_doc-summarization.yaml up ${service_name} -d

Kubernetes Deployment (Optional)

kubectl apply -f deployment/k8s/docsum-deployment.yaml
kubectl apply -f deployment/k8s/docsum-service.yaml
kubectl apply -f deployment/k8s/docsum-ingress.yaml  # Optional

For details, see Kubernetes Deployment Guide.

Consume Microservice

Check Service Status

curl http://${your_ip}:9000/v1/health_check \
-X GET \
-H 'Content-Type: application/json'

Consume LLM Service

Basic usage:

In DocSum microservice, except for basic LLM parameters, we also support several optimization parameters setting.

  • “language”: specify the language, can be “auto”, “en”, “zh”, default is “auto”

If you want to deal with long context, can select suitable summary type, details in section 3.2.2.

  • “summary_type”: can be “auto”, “stuff”, “truncate”, “map_reduce”, “refine”, default is “auto”

  • “chunk_size”: max token length for each chunk. Set to be different default value according to “summary_type”.

  • “chunk_overlap”: overlap token length between each chunk, default is 0.1*chunk_size

Basic usage

# Enable stream to receive a stream response. By default, this is set to True.
curl http://${your_ip}:9000/v1/docsum \
  -X POST \
  -d '{"messages":"Text Embeddings Inference (TEI) is a toolkit for deploying and serving open source text embeddings and sequence classification models. TEI enables high-performance extraction for the most popular models, including FlagEmbedding, Ember, GTE and E5.", "max_tokens":32, "language":"en"}' \
  -H 'Content-Type: application/json'

# Disable stream to receive a non-stream response.
curl http://${your_ip}:9000/v1/docsum \
  -X POST \
  -d '{"messages":"Text Embeddings Inference (TEI) is a toolkit for deploying and serving open source text embeddings and sequence classification models. TEI enables high-performance extraction for the most popular models, including FlagEmbedding, Ember, GTE and E5.", "max_tokens":32, "language":"en", "stream":false}' \
  -H 'Content-Type: application/json'

# Use Chinese mode
curl http://${your_ip}:9000/v1/docsum \
  -X POST \
  -d '{"messages":"2024年9月26日，北京——今日，英特尔正式发布英特尔® 至强® 6性能核处理器（代号Granite Rapids），为AI、数据分析、科学计算等计算密集型业务提供卓越性能。", "max_tokens":32, "language":"zh", "stream":false}' \
  -H 'Content-Type: application/json'

Long context summarization with “summary_type”

summary_type=auto

“summary_type” is set to be “auto” by default, in this mode we will check input token length, if it exceed MAX_INPUT_TOKENS, summary_type will automatically be set to refine mode, otherwise will be set to stuff mode.

With long contexts, request may get canceled due to its generation taking longer than the default timeout value (120s for TGI). Increase it as needed.

summary_type=stuff

In this mode LLM generate summary based on complete input text. In this case please carefully set MAX_INPUT_TOKENS and MAX_TOTAL_TOKENS according to your model and device memory, otherwise it may exceed LLM context limit and raise error when meet long context.

summary_type=truncate

Truncate mode will truncate the input text and keep only the first chunk, whose length is equal to min(MAX_TOTAL_TOKENS - input.max_tokens - 50, MAX_INPUT_TOKENS)

curl http://${your_ip}:9000/v1/docsum \
  -X POST \
  -d '{"messages":"Text Embeddings Inference (TEI) is a toolkit for deploying and serving open source text embeddings and sequence classification models. TEI enables high-performance extraction for the most popular models, including FlagEmbedding, Ember, GTE and E5.", "max_tokens":32, "language":"en", "summary_type": "truncate", "chunk_size": 2000}' \
  -H 'Content-Type: application/json'

summary_type=map_reduce

Map_reduce mode will split the inputs into multiple chunks, map each document to an individual summary, then consolidate those summaries into a single global summary. stream=True is not allowed here.

In this mode, default chunk_size is set to be min(MAX_TOTAL_TOKENS - input.max_tokens - 50, MAX_INPUT_TOKENS)

curl http://${your_ip}:9000/v1/docsum \
  -X POST \
  -d '{"messages":"Text Embeddings Inference (TEI) is a toolkit for deploying and serving open source text embeddings and sequence classification models. TEI enables high-performance extraction for the most popular models, including FlagEmbedding, Ember, GTE and E5.", "max_tokens":32, "language":"en", "summary_type": "map_reduce", "chunk_size": 2000, "stream":false, "timeout":200}' \
  -H 'Content-Type: application/json'

summary_type=refine

Refine mode will split the inputs into multiple chunks, generate summary for the first one, then combine with the second, loops over every remaining chunks to get the final summary.

In this mode, default chunk_size is set to be min(MAX_TOTAL_TOKENS - 2 * input.max_tokens - 128, MAX_INPUT_TOKENS).

curl http://${your_ip}:9000/v1/docsum \
  -X POST \
  -d '{"messages":"Text Embeddings Inference (TEI) is a toolkit for deploying and serving open source text embeddings and sequence classification models. TEI enables high-performance extraction for the most popular models, including FlagEmbedding, Ember, GTE and E5.", "max_tokens":32, "language":"en", "summary_type": "refine", "chunk_size": 2000, "timeout":200}' \
  -H 'Content-Type: application/json'

Air-Gapped Environment Usage

The following steps are needed for running the opea/llm-docsum microservice in an air gapped environment (a.k.a. environment with no internet access).

  1. Pre download the following models, i.e. huggingface-cli download --cache-dir <model data directory> <model>

  • gpt2

  • the same model as the LLM inference backend

  1. Launch the opea/llm-docsum microservice with the following settings:

  • mount the host <model data directory> as the /data directory within the microservice container

  • leave environment as unset HF_TOKEN as unset

e.g. unset HF_TOKEN; docker run -v <model data directory>:/data ... ...