Skip to content

Latest commit

 

History

History
127 lines (118 loc) · 5.23 KB

README.md

File metadata and controls

127 lines (118 loc) · 5.23 KB
Raw

Prerequisites:

  • Python =3.11

Testing:

It is implementation of inferencing pipeline.

  1. Create, activate virtual environment, and install the dependencies.
  2. Look at config.ini file for STORAGE_ROOT_PATH and dpp_input_config_file_path and create path.
    e.g. Create C:/DPP/infy_libraries_client/STORAGE/
  3. Copy config file from \apps\config\dpp_docwb_infy_search_service_processor_input_config.json to STORAGE_ROOT_PATH+dpp_input_config_file_path
    e.g. C:/DPP/infy_libraries_client/STORAGE/data/config/ dpp_docwb_infy_search_service_processor_input_config.json
  4. If centralised db is being used to query from (i.e. indexed stored via infy_db_service), modify config file of infy_search_service:
    • Enable infy_db_service under vectordb and sparseindex of QueryRetriever and Reader processor config
    • Provide value for db_service_url under storage of QueryRetriever processor.
      • For vectordb http://<hostname>:8005/api/v1/vectordb/getmatches
      • For sparseindex http://<hostname>:8005/api/v1/sparsedb/getmatches.
        (Replace the hostname from URL with your hostname).
      "QueryRetriever": {
            "embedding": {},
            "storage": {
                "vectordb": {
                    "faiss": {},
                    "infy_db_service": {
                        "enabled": true,
                        "configuration": {
                            "db_service_url": "http://localhost:8005/api/v1/vectordb/getmatches",
                            "model_name": "",
                            "index_id": "",
                            "collections": [
                                {
                                    "collection_name": "",
                                    "collection_secret_key": ""
                                }
                            ]
                        }
                    }
                },
                "sparseindex": {
                    "bm25s": {},
                    "infy_db_service": {
                        "enabled": true,
                        "configuration": {
                            "db_service_url": "http://localhost:8005/api/v1/sparsedb/getmatches",
                            "method_name": "",
                            "index_id": "",
                            "collections": [
                                {
                                    "collection_name": "",
                                    "collection_secret_key": ""
                                }
                            ]
                        }
                    }
                }
            },
            "hybrid_search": {
                "rrf": {
                    "enabled": true
                },
            "queries": []
        }
  5. At variables part of the config file provide respective LLM URL and Key to be used, under Reader processor for fetching answer.
  6. When you run the code from local, run this in browser http://localhost:8004/api/v1/docs#
  7. Now use http://localhost:8004/api/v1/inference/search API by passing question and index_id (created at time of indexing pipeline; check processor_response_data.json file after running indexing pipeline for it).

Build Package:

  1. Run BuildPackage.bat.
  2. Package will be available at apps\infy_search_service\target.

Deploy:

Installation

  1. Copy the package from apps\infy_search_service\target to target server machine where you want to deploy.
  2. Create and activate a virtual environment: 
    python -m venv .venv
source ./.venv/bin/activate
  3. Upgrade pip: 
    pip install --upgrade pip
  4. Install required dependencies 
    pip install -r requirements.txt

Firewall settings

  1. Steps to follow for only the first time you are setting up the service.
  • Update Firewall:
    ( replace <port_number> with the port number used for this service in code) 
    sudo firewall-cmd --zone=public --add-port=8004/tcp --permanent
sudo firewall-cmd --reload
  • Start in Foreground (for Verification):
    Make sure you are inside the service folder and run the following commands: 
    export PYTHONPATH=`pwd`/src
cd src
python main.py

Configure as Systemmd Service in Linux

  1. Copy the service file to the systemd directory:
sudo cp infy_search_service.service /etc/systemd/system/
  1. Verify the service file:
sudo cat /etc/systemd/system/infy_search_service.service
  1. Enable and start the service:
sudo systemctl enable infy_search_service.service
sudo systemctl daemon-reload
sudo systemctl start infy_search_service.service
sudo systemctl status infy_search_service.service

Verification

Verify if the URL is working from browser.
http://<hostname>:8004/api/v1/docs