ch7

Description

Exercises for Chapter 7. "Deploying Haystack-based applications"

Topics covered

• Transition from exploratory development to deployment-focused engineering.
• Leverage APIs to serve results and create user-facing interfaces.
• Develop custom API endpoints tailored to your application's unique requirements.
• Structure and package your pipeline for smooth, automated deployment using Docker and CI/CD practices.

Material

Method 1: Build a custom endpoint with FastAPI, Bytewax and Docker

1. Indexing dataflow - This script combines a Haystack indexing pipeline with custom components with a Bytewax dataflow to enable filtering data with a streaming approach.
2. Querying pipeline - This script contains a retriever pipeline to answer queries in natural language.
3. API with FastAPI - This is a sample implementation of an API with an endpoint taking as input a stock symbol and a question in natural language.
4. Sample Dockerfile - This is a simple file to dockerize the API.

Method 2: Use pipeline serialization and Hayhooks to deploy a pipeline endpoint

1. Sample pipeline with a prompt template
2. Serialized pipeline