Awesome
Azure OpenAI Semantic Search Demo | Document Upload
About
Version: 0.9.7
The ability to conduct semantic search on vector data is a powerful feature that allows you to find relevant content based on a specific natural language query. This demo is helpful for showcasing and comprehending the abstractive responses generated from your own data in PDF or Word format documents.
This solution has been created by drawing inspiration from existing Enterprise Chat GPT and Document Q&A vector search demos, however it uses a simplified architecture pattern and offers following features, most of which are unique to this implementation.
Key Features
- Simplified architecture
- Built-in document cracking (.pdf, .docx)
- Utilise text embeddings
- Upload own document and ask questions
How to deploy?
Run locally from Visual Studio Code or command prompt
- Open VS Code terminal or command prompt.
- Clone this repository and open in VS Code.
- Create a new conda environment
conda create -n openaidemo_py39 python=3.9 anacondaconda activate openaidemo_py39- For more info go to Conda documentation
- Navigate to misc directory and run
pip install -r requirements.txt
- Provision Azure OpenAI service
- Under deployments within Azure OpenAI Studio, deploy 2 models
- Model for text search e.g. text-search-davinci-doc-001
- Model for text generation e.g. text-davinci-003
- Model availability varies by region in which OpenAI service is provisioned in Azure.
- For more info go to Create a resource and deploy a model using Azure OpenAI
- Open modules/utilities.py file and scroll to very bottom
- Update value for "deployment_name" tag for respective model version.
- If you have used the 2 models mentioned in steps above, no changes are required in app/app.py file.
- Otherwise, update values for aoai_embedding_model and aoai_text_model variables at the begining of app/app.py file.
- Provision Redis stack as Azure Container Instance
- Use image = redis/redis-stack-server:latest
- On advanced tab make sure you add
- --requirepass <your_strong_redis_pwd>
- Alternatively, you may run the Redis stack in local docker environment. In this case, you do not require to provision it as ACI service.
- For more info go to Quickstart: Deploy a container instance in Azure using the Azure portal
- Almost there, rename .env_template file to .env
- Renamed file should be placed within same directory as this readme.md file.
- Replace placeholders with correct values.
- Navigate to /app directory and run following command from VS Code terminal. This will open the App UI in a browser window.
streamlit run app.py
Run as a local docker container
- Complete all configuration steps from section above before running any steps below.
docker build -t oaisearch:0.9.7 -f ./docker/dockerfile .docker run -d -p 8501:8501 oaisearch:0.9.7- Open web browser and go to http://localhost:8501 URL to see the app UI.
To Azure
- More info will be added soon.