dataprep: support air gapped env

comps/dataprep/README.md

@@ -64,3 +64,20 @@ For details, please refer to this [readme](src/README_finance.md)
 ## Dataprep Microservice with MariaDB Vector
 
 For details, please refer to this [readme](src/README_mariadb.md)
+
+## Running in the air gapped environment
+
+The following steps are common for running the dataprep microservice in an air gapped environment (a.k.a. environment with no internet access), for all DB backends.
+
+1. Download the following models, e.g. `huggingface-cli download --cache-dir <model data directory> <model>`
+
+- microsoft/table-transformer-structure-recognition
+- timm/resnet18.a1_in1k
+- unstructuredio/yolo_x_layout
+
+2. launch the `dataprep` microservice with the following settings:
+
+- mount the `model data directory` as the `/data` directory within the `dataprep` container
+- set environment variable `HF_HUB_OFFLINE` to 1 when launching the `dataprep` microservice
+
+e.g. `docker run -d -v <model data directory>:/data -e HF_HUB_OFFLINE=1 ... ...`

comps/dataprep/deployment/docker_compose/compose.yaml

@@ -75,6 +75,26 @@ services:
       minio:
         condition: service_healthy
 
+  dataprep-milvus-offline:
+    extends: dataprep-milvus
+    depends_on:
+      tei-embedding-serving:
+        condition: service_healthy
+      standalone:
+        condition: service_healthy
+      etcd:
+        condition: service_healthy
+      minio:
+        condition: service_healthy
+    environment:
+      HF_HUB_OFFLINE: 1
+      # Use non-existing proxy to mimic air gapped environment
+      no_proxy: localhost,127.0.0.1,${offline_no_proxy}
+      http_proxy: http://localhost:7777
+      https_proxy: http://localhost:7777
+    volumes:
+      - "${DATA_PATH:-./data}:/data"
+
   dataprep-multimodal-milvus:
     image: ${REGISTRY:-opea}/dataprep:${TAG:-latest}
     container_name: dataprep-multimodal-milvus-server
@@ -242,6 +262,22 @@ services:
       retries: 10
     restart: unless-stopped
 
+  dataprep-qdrant-offline:
+    extends: dataprep-qdrant
+    depends_on:
+      qdrant-vector-db:
+        condition: service_healthy
+      tei-embedding-serving:
+        condition: service_healthy
+    environment:
+      HF_HUB_OFFLINE: 1
+      # Use non-existing proxy to mimic air gapped environment
+      no_proxy: localhost,127.0.0.1,${offline_no_proxy}
+      http_proxy: http://localhost:7777
+      https_proxy: http://localhost:7777
+    volumes:
+      - "${DATA_PATH:-./data}:/data"
+
   dataprep-redis:
     image: ${REGISTRY:-opea}/dataprep:${TAG:-latest}
     container_name: dataprep-redis-server
@@ -271,6 +307,22 @@ services:
       retries: 10
     restart: unless-stopped
 
+  dataprep-redis-offline:
+    extends: dataprep-redis
+    depends_on:
+      redis-vector-db:
+        condition: service_healthy
+      tei-embedding-serving:
+        condition: service_healthy
+    environment:
+      HF_HUB_OFFLINE: 1
+      # Use non-existing proxy to mimic air gapped environment
+      no_proxy: localhost,127.0.0.1,${offline_no_proxy}
+      http_proxy: http://localhost:7777
+      https_proxy: http://localhost:7777
+    volumes:
+      - "${DATA_PATH:-./data}:/data"
+
   dataprep-multimodal-redis:
     image: ${REGISTRY:-opea}/dataprep:${TAG:-latest}
     container_name: dataprep-multimodal-redis-server

comps/dataprep/src/Dockerfile

@@ -46,9 +46,14 @@ RUN pip install --no-cache-dir --upgrade pip setuptools && \
 ENV PYTHONPATH=$PYTHONPATH:/home/user
 
 RUN mkdir -p /home/user/comps/dataprep/src/uploaded_files && chown -R user /home/user/comps/dataprep/src/uploaded_files
+RUN mkdir -p /data && chown -R user /data
 
 USER user
 ENV NLTK_DATA=/home/user/nltk_data
+# air gapped support: predownload all needed nltk data
+RUN mkdir -p /home/user/nltk_data && python -m nltk.downloader -d /home/user/nltk_data punkt_tab averaged_perceptron_tagger_eng stopwords
+# air gapped support: set model cache dir
+ENV HF_HUB_CACHE=/data
 
 WORKDIR /home/user/comps/dataprep/src
 

comps/dataprep/src/README_milvus.md

@@ -207,3 +207,7 @@ curl -X POST \
        -F "chunk_size=500" \
        http://localhost:6010/v1/dataprep/ingest
    ```
+
+## Running in the air gapped environment
+
+Please follow the [common guide](../README.md#running-in-the-air-gapped-environment) to run dataprep microservice in the air gapped environment.

comps/dataprep/src/README_qdrant.md

@@ -72,3 +72,7 @@ curl -X POST \
     -F "table_strategy=hq" \
     http://localhost:6007/v1/dataprep/ingest
 ```
+
+## Running in the air gapped environment
+
+Please follow the [common guide](../README.md#running-in-the-air-gapped-environment) to run dataprep microservice in the air gapped environment.

comps/dataprep/src/README_redis.md

@@ -261,3 +261,7 @@ curl -X POST \
     -d '{"file_path": "all", "index_name": "test_redis_1"}' \
     http://localhost:6007/v1/dataprep/delete
 ```
+
+## Running in the air gapped environment
+
+Please follow the [common guide](../README.md#running-in-the-air-gapped-environment) to run dataprep microservice in the air gapped environment.

tests/dataprep/dataprep_utils.sh

@@ -224,3 +224,21 @@ function check_healthy() {
     echo "$container_name did not become healthy in time."
     return 1
 }
+
+DATAPREP_MODELS=(microsoft/table-transformer-structure-recognition timm/resnet18.a1_in1k unstructuredio/yolo_x_layout)
+
+function prepare_dataprep_models() {
+    local model_path=$1
+    mkdir -p ${model_path}
+    python3 -m pip install huggingface_hub[cli] --user
+    # Workaround for huggingface-cli reporting error when set --cache-dir to same as default
+    local extra_args=""
+    local default_model_dir=$(readlink -m ~/.cache/huggingface/hub)
+    local real_model_dir=$(echo ${model_path/#\~/$HOME} | xargs readlink -m )
+    if [[ "${default_model_dir}" != "${real_model_dir}" ]]; then
+        extra_args="--cache-dir ${model_path}"
+    fi
+    for m in ${DATAPREP_MODELS[@]}; do
+      PATH=~/.local/bin:$PATH huggingface-cli download ${extra_args} $m
+    done
+}

tests/dataprep/test_dataprep_milvus.sh

@@ -28,20 +28,28 @@ function build_docker_images() {
 }
 
 function start_service() {
+    local offline=${1:-false}
     export host_ip=${ip_address}
     export TEI_EMBEDDER_PORT=12005
     export EMBEDDING_MODEL_ID="BAAI/bge-base-en-v1.5"
     export MILVUS_HOST=${ip_address}
     export TEI_EMBEDDING_ENDPOINT="http://${host_ip}:${TEI_EMBEDDER_PORT}"
     export LOGFLAG=true
 
+    if [[ "$offline" == "true" ]]; then
+        service_name="dataprep-milvus-offline tei-embedding-serving etcd minio standalone"
+        export offline_no_proxy="${ip_address},${host_ip}"
+    else
+        service_name="dataprep-milvus tei-embedding-serving etcd minio standalone"
+    fi
     cd $WORKPATH/comps/dataprep/deployment/docker_compose/
     docker compose up ${service_name} -d > ${LOG_PATH}/start_services_with_compose.log
 
     check_healthy "dataprep-milvus-server" || exit 1
 }
 
 function validate_microservice() {
+    local offline=${1:-false}
     # test /v1/dataprep/delete
     delete_all ${ip_address} ${DATAPREP_PORT}
     check_result "dataprep - del" '{"status":true}' dataprep-milvus-server ${LOG_PATH}/dataprep_milvus.log
@@ -69,8 +77,10 @@ function validate_microservice() {
     check_result "dataprep - upload - xlsx" "Data preparation succeeded" dataprep-milvus-server ${LOG_PATH}/dataprep_milvus.log
 
     # test /v1/dataprep/ingest upload link
-    ingest_external_link ${ip_address} ${DATAPREP_PORT}
-    check_result "dataprep - upload - link" "Data preparation succeeded" dataprep-milvus-server ${LOG_PATH}/dataprep_milvus.log
+    if [[ "$offline" != "true" ]]; then
+      ingest_external_link ${ip_address} ${DATAPREP_PORT}
+      check_result "dataprep - upload - link" "Data preparation succeeded" dataprep-milvus-server ${LOG_PATH}/dataprep_milvus.log
+    fi
 
     # test /v1/dataprep/get
     get_all ${ip_address} ${DATAPREP_PORT}
@@ -95,11 +105,21 @@ function main() {
     stop_docker
 
     build_docker_images
-    start_service
+    trap stop_docker EXIT
 
+    echo "Test normal env ..."
+    start_service
     validate_microservice
-
     stop_docker
+
+    if [[ -n "${DATA_PATH}" ]]; then
+        echo "Test air gapped env ..."
+        prepare_dataprep_models ${DATA_PATH}
+        start_service true
+        validate_microservice true
+        stop_docker
+    fi
+
     echo y | docker system prune
 
 }

tests/dataprep/test_dataprep_qdrant.sh

@@ -29,6 +29,7 @@ function build_docker_images() {
 }
 
 function start_service() {
+    local offline=${1:-false}
     export host_ip=${ip_address}
     export EMBEDDING_MODEL_ID="BAAI/bge-base-en-v1.5"
     export EMBED_MODEL=${EMBEDDING_MODEL_ID}
@@ -37,14 +38,20 @@ function start_service() {
     export COLLECTION_NAME="rag-qdrant"
     export QDRANT_HOST=$ip_address
     export QDRANT_PORT=6360
-    service_name="qdrant-vector-db tei-embedding-serving dataprep-qdrant"
+    if [[ "$offline" == "true" ]]; then
+        service_name="qdrant-vector-db tei-embedding-serving dataprep-qdrant-offline"
+        export offline_no_proxy="${ip_address}"
+    else
+        service_name="qdrant-vector-db tei-embedding-serving dataprep-qdrant"
+    fi
     cd $WORKPATH/comps/dataprep/deployment/docker_compose/
     docker compose up ${service_name} -d
 
     check_healthy "dataprep-qdrant-server" || exit 1
 }
 
 function validate_microservice() {
+    local offline=${1:-false}
     # test /v1/dataprep/ingest upload file
     ingest_doc ${ip_address} ${DATAPREP_PORT}
     check_result "dataprep - upload - doc" "Data preparation succeeded" dataprep-qdrant-server ${LOG_PATH}/dataprep-qdrant.log
@@ -68,8 +75,10 @@ function validate_microservice() {
     check_result "dataprep - upload - xlsx" "Data preparation succeeded" dataprep-qdrant-server ${LOG_PATH}/dataprep-qdrant.log
 
     # test /v1/dataprep/ingest upload link
-    ingest_external_link ${ip_address} ${DATAPREP_PORT}
-    check_result "dataprep - upload - link" "Data preparation succeeded" dataprep-qdrant-server ${LOG_PATH}/dataprep-qdrant.log
+    if [[ "$offline" != "true" ]]; then
+      ingest_external_link ${ip_address} ${DATAPREP_PORT}
+      check_result "dataprep - upload - link" "Data preparation succeeded" dataprep-qdrant-server ${LOG_PATH}/dataprep-qdrant.log
+    fi
 
 }
 
@@ -78,14 +87,30 @@ function stop_docker() {
     if [[ ! -z "$cid" ]]; then docker stop $cid && docker rm $cid && sleep 1s; fi
 }
 
+function stop_service() {
+    cd $WORKPATH/comps/dataprep/deployment/docker_compose/
+    docker compose down || true
+}
+
 function main() {
 
     stop_docker
 
     build_docker_images
-    start_service
+    trap stop_service EXIT
 
+    echo "Test normal env ..."
+    start_service
     validate_microservice
+    stop_service
+
+    if [[ -n "${DATA_PATH}" ]]; then
+        echo "Test air gapped env ..."
+        prepare_dataprep_models ${DATA_PATH}
+        start_service true
+        validate_microservice true
+        stop_service
+    fi
 
     stop_docker
     echo y | docker system prune

tests/dataprep/test_dataprep_redis.sh

@@ -28,6 +28,7 @@ function build_docker_images() {
 }
 
 function start_service() {
+    local offline=${1:-false}
 
     export host_ip=${ip_address}
     export REDIS_HOST=$ip_address
@@ -38,14 +39,20 @@ function start_service() {
     export EMBEDDING_MODEL_ID="BAAI/bge-base-en-v1.5"
     export TEI_EMBEDDING_ENDPOINT="http://${ip_address}:${TEI_EMBEDDER_PORT}"
     export INDEX_NAME="rag_redis"
-    service_name="redis-vector-db tei-embedding-serving dataprep-redis"
+    if [[ "$offline" == "true" ]]; then
+        service_name="redis-vector-db tei-embedding-serving dataprep-redis-offline"
+        export offline_no_proxy="${ip_address}"
+    else
+        service_name="redis-vector-db tei-embedding-serving dataprep-redis"
+    fi
     cd $WORKPATH/comps/dataprep/deployment/docker_compose/
     docker compose up ${service_name} -d
 
     check_healthy "dataprep-redis-server" || exit 1
 }
 
 function validate_microservice() {
+    local offline=${1:-false}
 
     # test /v1/dataprep/delete
     delete_all ${ip_address} ${DATAPREP_PORT}
@@ -73,12 +80,14 @@ function validate_microservice() {
     ingest_xlsx ${ip_address} ${DATAPREP_PORT} "redis"
     check_result "dataprep - upload - xlsx" "Data preparation succeeded" dataprep-redis-server ${LOG_PATH}/dataprep_upload_file.log
 
-     # test /v1/dataprep/ingest upload link
-    ingest_external_link ${ip_address} ${DATAPREP_PORT}
-    check_result "dataprep - upload - link" "Data preparation succeeded" dataprep-redis-server ${LOG_PATH}/dataprep_upload_file.log
+    # test /v1/dataprep/ingest upload link
+    if [[ "$offline" != "true" ]]; then
+      ingest_external_link ${ip_address} ${DATAPREP_PORT}
+      check_result "dataprep - upload - link" "Data preparation succeeded" dataprep-redis-server ${LOG_PATH}/dataprep_upload_file.log
 
-    ingest_external_link_with_chunk_parameters ${ip_address} ${DATAPREP_PORT} "rag_redis_test_link_params"
-    check_result "dataprep - upload - link" "Data preparation succeeded" dataprep-redis-server ${LOG_PATH}/dataprep_upload_file.log
+      ingest_external_link_with_chunk_parameters ${ip_address} ${DATAPREP_PORT} "rag_redis_test_link_params"
+      check_result "dataprep - upload - link" "Data preparation succeeded" dataprep-redis-server ${LOG_PATH}/dataprep_upload_file.log
+    fi
 
     ingest_txt_with_index_name ${ip_address} ${DATAPREP_PORT} rag_redis_test
     check_result "dataprep - upload with index - txt" "Data preparation succeeded" dataprep-redis-server ${LOG_PATH}/dataprep_upload_file.log
@@ -114,14 +123,30 @@ function stop_docker() {
     if [[ ! -z "$cid" ]]; then docker stop $cid && docker rm $cid && sleep 1s; fi
 }
 
+function stop_service() {
+    cd $WORKPATH/comps/dataprep/deployment/docker_compose/
+    docker compose down || true
+}
+
 function main() {
 
     stop_docker
 
     build_docker_images
-    start_service
+    trap stop_service EXIT
 
+    echo "Test normal env ..."
+    start_service
     validate_microservice
+    stop_service
+
+    if [[ -n "${DATA_PATH}" ]]; then
+        echo "Test air gapped env ..."
+        prepare_dataprep_models ${DATA_PATH}
+        start_service true
+        validate_microservice true
+        stop_service
+    fi
 
     stop_docker
     echo y | docker system prune