Haystack demonstration used for discussions during a DataScience Meetup: Meetup: Building LLM applications - inovex GmbH.
It showcases:
- Index & Query Pipelines
- Crawling website content
- LLM integration & multi-turn conversations
Additionaly it contains a CustomPreprocessor which handles crawled websites.
Install the dependencies inside requirements.txt.
Execute the first cell in main.py to automatically install jupyter-notebook into the same environment.
An Elasticsearch Documentstore is required to run the haystack demo. Deployment via docker is the easiest way.
docker network create elastic
docker run --name es01 --net elastic -p 9200:9200 -it -e "discovery.type=single-node" -e "xpack.security.enabled=false" -m 4GB docker.elastic.co/elasticsearch/elasticsearch:8.12.1LLM API credentials are required, to run later stages of the notebook.
An .env file provides a credentials template.
(We used gpt3.5 from azure. If you plan to use another model read the haystack-docs, since the initialization format changes.)