Qdrant retriever microservice (opea-project#216)

* PGvector service (opea-project#86)

* Support PGvector service

Signed-off-by: V, Ganesan <[email protected]>
Signed-off-by: gadmarkovits <[email protected]>
Signed-off-by: Yogesh Pandey <[email protected]>

comps/retrievers/haystack/qdrant/README.md

@@ -0,0 +1,69 @@
+# Retriever Microservice with Qdrant
+
+# 🚀Start Microservice with Python
+
+## Install Requirements
+
+```bash
+pip install -r requirements.txt
+```
+
+## Start Qdrant Server
+
+Please refer to this [readme](../../../vectorstores/langchain/qdrant/README.md).
+
+## Setup Environment Variables
+
+```bash
+export http_proxy=${your_http_proxy}
+export https_proxy=${your_https_proxy}
+export QDRANT_HOST=${your_qdrant_host_ip}
+export QDRANT_PORT=6333
+export EMBED_DIMENSION=${your_embedding_dimension}
+export INDEX_NAME=${your_index_name}
+export TEI_EMBEDDING_ENDPOINT=${your_tei_endpoint}
+```
+
+## Start Retriever Service
+
+```bash
+export TEI_EMBEDDING_ENDPOINT="http://${your_ip}:6060"
+python haystack/qdrant/retriever_qdrant.py
+```
+
+# 🚀Start Microservice with Docker
+
+## Build Docker Image
+
+```bash
+cd ../../
+docker build -t opea/retriever-qdrant:latest --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy -f comps/retrievers/haystack/qdrant/docker/Dockerfile .
+```
+
+## Run Docker with CLI
+
+```bash
+docker run -d --name="retriever-qdrant-server" -p 7000:7000 --ipc=host -e http_proxy=$http_proxy -e https_proxy=$https_proxy -e TEI_EMBEDDING_ENDPOINT=${your_tei_endpoint} -e QDRANT_HOST=${your_qdrant_host_ip} -e QDRANT_PORT=${your_qdrant_port} opea/retriever-qdrant:latest
+```
+
+# 🚀3. Consume Retriever Service
+
+## 3.1 Check Service Status
+
+```bash
+curl http://${your_ip}:7000/v1/health_check \
+  -X GET \
+  -H 'Content-Type: application/json'
+```
+
+## 3.2 Consume Embedding Service
+
+To consume the Retriever Microservice, you can generate a mock embedding vector of length 768 with Python.
+
+```bash
+your_embedding=$(python -c "import random; embedding = [random.uniform(-1, 1) for _ in range(768)]; print(embedding)")
+curl http://${your_ip}:7000/v1/retrieval \
+  -X POST \
+  -d "{\"text\":\"What is the revenue of Nike in 2023?\",\"embedding\":${your_embedding}}" \
+  -H 'Content-Type: application/json'
+```

comps/retrievers/haystack/qdrant/__init__.py

@@ -0,0 +1,2 @@
+# Copyright (C) 2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0

comps/retrievers/haystack/qdrant/docker/Dockerfile

@@ -0,0 +1,27 @@
+
+# Copyright (C) 2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+FROM python:3.11-slim
+
+RUN apt-get update -y && apt-get install -y --no-install-recommends --fix-missing \
+    libgl1-mesa-glx \
+    libjemalloc-dev \
+    vim
+
+RUN useradd -m -s /bin/bash user && \
+    mkdir -p /home/user && \
+    chown -R user /home/user/
+
+USER user
+
+COPY comps /home/user/comps
+
+RUN python -m pip install --no-cache-dir --upgrade pip && \
+    python -m pip install --no-cache-dir -r /home/user/comps/retrievers/haystack/qdrant/requirements.txt
+
+ENV PYTHONPATH=$PYTHONPATH:/home/user
+
+WORKDIR /home/user/comps/retrievers/haystack/qdrant
+
+ENTRYPOINT ["python", "retriever_qdrant.py"]

comps/retrievers/haystack/qdrant/ingest.py

@@ -0,0 +1,110 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+# Copyright (C) 2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+#
+
+import argparse
+import io
+import os
+import uuid
+
+import numpy as np
+from haystack.components.embedders import HuggingFaceTEIDocumentEmbedder, SentenceTransformersDocumentEmbedder
+from haystack.dataclasses.document import Document
+from haystack_integrations.document_stores.qdrant import QdrantDocumentStore
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from PIL import Image
+from qdrant_config import EMBED_DIMENSION, EMBED_ENDPOINT, EMBED_MODEL, INDEX_NAME, QDRANT_HOST, QDRANT_PORT
+
+
+def pdf_loader(file_path):
+    try:
+        import easyocr
+        import fitz
+    except ImportError:
+        raise ImportError(
+            "`PyMuPDF` or 'easyocr' package is not found, please install it with "
+            "`pip install pymupdf or pip install easyocr.`"
+        )
+
+    doc = fitz.open(file_path)
+    reader = easyocr.Reader(["en"])
+    result = ""
+    for i in range(doc.page_count):
+        page = doc.load_page(i)
+        pagetext = page.get_text().strip()
+        if pagetext:
+            result = result + pagetext
+        if len(doc.get_page_images(i)) > 0:
+            for img in doc.get_page_images(i):
+                if img:
+                    pageimg = ""
+                    xref = img[0]
+                    img_data = doc.extract_image(xref)
+                    img_bytes = img_data["image"]
+                    pil_image = Image.open(io.BytesIO(img_bytes))
+                    img = np.array(pil_image)
+                    img_result = reader.readtext(img, paragraph=True, detail=0)
+                    pageimg = pageimg + ", ".join(img_result).strip()
+                    if pageimg.endswith("!") or pageimg.endswith("?") or pageimg.endswith("."):
+                        pass
+                    else:
+                        pageimg = pageimg + "."
+                result = result + pageimg
+    return result
+
+
+def ingest_documents(folder_path, tag):
+    """Ingest PDF to Qdrant from the a given path."""
+    # Load list of pdfs
+    doc_path = [os.path.join(folder_path, file) for file in os.listdir(folder_path)][0]
+
+    print("Parsing...", doc_path)
+
+    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1500, chunk_overlap=100, add_start_index=True)
+    content = pdf_loader(doc_path)
+    chunks = text_splitter.split_text(content)
+
+    print("Done preprocessing. Created ", len(chunks), " chunks of the original pdf")
+    # Create vectorstore
+    if EMBED_ENDPOINT:
+        # create embeddings using TEI endpoint service
+        embedder = HuggingFaceTEIDocumentEmbedder(url=EMBED_ENDPOINT)
+    else:
+        # create embeddings using local embedding model
+        embedder = SentenceTransformersDocumentEmbedder(model=EMBED_MODEL)
+        embedder.warm_up()
+
+    # Initialize Qdrant store
+    qdrant_store = QdrantDocumentStore(
+        host=QDRANT_HOST,
+        port=QDRANT_PORT,
+        embedding_dim=EMBED_DIMENSION,
+        index=INDEX_NAME,
+        embedding_field="embedding",
+        similarity="cosine",
+        recreate_index=True,
+    )
+
+    # Batch size
+    batch_size = 32
+    num_chunks = len(chunks)
+    for i in range(0, num_chunks, batch_size):
+        batch_chunks = chunks[i : i + batch_size]
+        batch_texts = [f"Tag: {tag}. " + chunk for chunk in batch_chunks]
+        documents = [Document(id=str(uuid.uuid4()), content=content) for content in batch_texts]
+        documents_with_embeddings = embedder.run(documents)["documents"]
+        qdrant_store.write_documents(documents_with_embeddings)
+
+        print(f"Processed batch {i//batch_size + 1}/{(num_chunks-1)//batch_size + 1}")
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Ingest documents from a specified folder with a tag")
+    parser.add_argument("folder_path", type=str, help="Path to the folder containing documents")
+    parser.add_argument("--tag", type=str, default="", help="Tag to be used as an identifier")
+
+    args = parser.parse_args()
+    ingest_documents(args.folder_path, args.tag)

comps/retrievers/haystack/qdrant/qdrant_config.py

@@ -0,0 +1,20 @@
+# Copyright (C) 2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import os
+
+# Embedding model
+EMBED_MODEL = os.getenv("EMBED_MODEL", "BAAI/bge-base-en-v1.5")
+
+# Embedding dimension
+EMBED_DIMENSION = os.getenv("EMBED_DIMENSION", 768)
+
+# Embedding endpoints
+EMBED_ENDPOINT = os.getenv("TEI_EMBEDDING_ENDPOINT", "")
+
+# Qdrant Connection Information
+QDRANT_HOST = os.getenv("QDRANT_HOST", "localhost")
+QDRANT_PORT = int(os.getenv("QDRANT_PORT", 6333))
+
+# Vector Index Configuration
+INDEX_NAME = os.getenv("INDEX_NAME", "rag-qdrant")

comps/retrievers/haystack/qdrant/requirements.txt

@@ -0,0 +1,13 @@
+docarray[full]
+easyocr
+fastapi
+haystack-ai
+langchain_community
+langsmith
+opentelemetry-api
+opentelemetry-exporter-otlp
+opentelemetry-sdk
+pymupdf
+qdrant-haystack
+sentence_transformers
+shortuuid

comps/retrievers/haystack/qdrant/retriever_qdrant.py

@@ -0,0 +1,49 @@
+# Copyright (C) 2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+from haystack.components.embedders import HuggingFaceTEITextEmbedder, SentenceTransformersTextEmbedder
+from haystack_integrations.components.retrievers.qdrant import QdrantEmbeddingRetriever
+from haystack_integrations.document_stores.qdrant import QdrantDocumentStore
+from langsmith import traceable
+from qdrant_config import EMBED_DIMENSION, EMBED_ENDPOINT, EMBED_MODEL, INDEX_NAME, QDRANT_HOST, QDRANT_PORT
+
+from comps import EmbedDoc768, SearchedDoc, ServiceType, TextDoc, opea_microservices, register_microservice
+
+
+# Create a pipeline for querying a Qdrant document store
+def initialize_qdrant_retriever() -> QdrantEmbeddingRetriever:
+    qdrant_store = QdrantDocumentStore(
+        host=QDRANT_HOST, port=QDRANT_PORT, embedding_dim=EMBED_DIMENSION, index=INDEX_NAME, recreate_index=False
+    )
+
+    retriever = QdrantEmbeddingRetriever(document_store=qdrant_store)
+
+    return retriever
+
+
+@register_microservice(
+    name="opea_service@retriever_qdrant",
+    service_type=ServiceType.RETRIEVER,
+    endpoint="/v1/retrieval",
+    host="0.0.0.0",
+    port=7000,
+)
+@traceable(run_type="retriever")
+def retrieve(input: EmbedDoc768) -> SearchedDoc:
+    search_res = retriever.run(query_embedding=input.embedding)["documents"]
+    searched_docs = [TextDoc(text=r.content) for r in search_res]
+    result = SearchedDoc(retrieved_docs=searched_docs, initial_query=input.text)
+    return result
+
+
+if __name__ == "__main__":
+    if EMBED_ENDPOINT:
+        # create embeddings using TEI endpoint service
+        embedder = HuggingFaceTEITextEmbedder(url=EMBED_ENDPOINT)
+    else:
+        # create embeddings using local embedding model
+        embedder = SentenceTransformersTextEmbedder(model=EMBED_MODEL)
+        embedder.warm_up()
+
+    retriever = initialize_qdrant_retriever()
+    opea_microservices["opea_service@retriever_qdrant"].start()

tests/test_retrievers_haystack_qdrant.sh

@@ -0,0 +1,76 @@
+#!/bin/bash
+# Copyright (C) 2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+set -xe
+
+WORKPATH=$(dirname "$PWD")
+ip_address=$(hostname -I | awk '{print $1}')
+function build_docker_images() {
+    cd $WORKPATH
+    docker build --no-cache -t opea/retriever-qdrant:comps --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy -f comps/retrievers/haystack/qdrant/docker/Dockerfile .
+}
+
+function start_service() {
+    # qdrant
+    docker run -d --name test-qdrant-vector-db -p 5010:6333 -e HTTPS_PROXY=$https_proxy -e HTTP_PROXY=$https_proxy qdrant/qdrant
+    sleep 10s
+
+    # tei endpoint
+    tei_endpoint=5008
+    model="BAAI/bge-base-en-v1.5"
+    docker run -d --name="test-comps-retriever-tei-endpoint" -p $tei_endpoint:80 -v ./data:/data --pull always ghcr.io/huggingface/text-embeddings-inference:cpu-1.2 --model-id $model
+    sleep 30s
+    export TEI_EMBEDDING_ENDPOINT="http://${ip_address}:${tei_endpoint}"
+
+    # qdrant retriever
+    export QDRANT_HOST="${ip_address}"
+    export QDRANT_PORT=5010
+    export INDEX_NAME="rag-qdrant"
+    retriever_port=5009
+    unset http_proxy
+    docker run -d --name="test-comps-retriever-qdrant-server" -p ${retriever_port}:7000 --ipc=host -e TEI_EMBEDDING_ENDPOINT=$TEI_EMBEDDING_ENDPOINT -e http_proxy=$http_proxy -e https_proxy=$https_proxy -e QDRANT_HOST=$QDRANT_HOST -e QDRANT_PORT=$QDRANT_PORT -e INDEX_NAME=$INDEX_NAME opea/retriever-qdrant:comps
+
+    sleep 3m
+}
+
+function validate_microservice() {
+    retriever_port=5009
+    export PATH="${HOME}/miniforge3/bin:$PATH"
+    source activate
+    test_embedding=$(python -c "import random; embedding = [random.uniform(-1, 1) for _ in range(768)]; print(embedding)")
+    http_proxy='' curl http://${ip_address}:$retriever_port/v1/retrieval \
+        -X POST \
+        -d "{\"text\":\"test\",\"embedding\":${test_embedding}}" \
+        -H 'Content-Type: application/json'
+    docker logs test-comps-retriever-qdrant-server
+    docker logs test-comps-retriever-tei-endpoint
+}
+
+function stop_docker() {
+    cid_retrievers=$(docker ps -aq --filter "name=test-comps-retrievers*")
+    if [[ ! -z "$cid_retrievers" ]]; then
+        docker stop $cid_retrievers && docker rm $cid_retrievers && sleep 1s
+    fi
+
+    cid_qdrant=$(docker ps -aq --filter "name=test-qdrant-vector-db")
+    if [[ ! -z "$cid_qdrant" ]]; then
+        docker stop $cid_qdrant && docker rm $cid_qdrant && sleep 1s
+    fi
+}
+
+function main() {
+
+    stop_docker
+
+    build_docker_images
+    start_service
+
+    validate_microservice
+
+    stop_docker
+    echo y | docker system prune
+
+}
+
+main