07-project-example

7. End-to-End Project Example

Links:

• Project alexeygrigorev/fitness-assistant
• Project criteria

Note: check the final result, it's a bit different from what we showed in the videos: we further improved it by doing some small things here and there, like improved README, code readability, etc.

7.1. Fitness assistant project

• Generating data for the project
• Setting up the project
• Implementing the initial version of the RAG flow

7.2. Evaluating retrieval

• Preparing the README file
• Generating gold standard evaluation data
• Evaluting retrieval
• Findning the best boosting coefficients

7.3 Evaluating RAG

• Using LLM-as-a-Judge (type 2)
• Comparing gpt-4o-mini with gpt-4o

7.4 Interface and ingestion pipeline

• Turnining the jupyter notebook into a script
• Creating the ingestion pipeline
• Creating the API interface with Flask
• Improving README

7.5 Monitoring and containerization

• Creating a Docker image for our application
• Putting everything in docker compose
• Logging all the information for monitoring purposes

7.6 Summary and closing remarks

• Changes between 7.5 and 7.6 (postres logging, grafara, cli.py, etc)
• README file improvements
• Total cost of the project (~$2) and how to lower it
• Using generated data for real-life projects

7.7 Chunking for longer texts

• Different chunking strategies
• Use cases: multiple articles, one article, slide decks

Links:

• https://chatgpt.com/share/a4616f6b-43f4-4225-9d03-bb69c723c210
• https://chatgpt.com/share/74217c02-95e6-46ae-b5a5-ca79f9a07084
• https://chatgpt.com/share/8cf0ebde-c53f-4c6f-82ae-c6cc52b2fd0b

Notes

• First link goes here
• Did you take notes? Add them above this line (Send a PR with links to your notes)