# HW 7: Solutions

## Deliverables

* Completed Notebook
* Chainlit Application in a Hugging Face Space Powered by Hugging Face Endpoints
* Screenshot of endpoint usage

### Completed notebook

[Located here](./Completed_BazeleyMikiko_Open_Source_RAG_Leveraging_Hugging_Face_Endpoints_through_LangChain.ipynb)
[Also on HF space](https://huggingface.co/spaces/mmbazel/AIE3-Demo-Wk4Day1/blob/main/%5BCompleted%5D%20BazeleyMikiko_Open_Source_RAG_Leveraging_Hugging_Face_Endpoints_through_LangChain.ipynb)

### Chainlit Application

[Link to Chainlit App in HuggingFace Space](https://huggingface.co/spaces/mmbazel/AIE3-Demo-Wk4Day1)


### Screenshots
#### The chat 
![alt text](img/chat.png)

#### The trace
![alt text](img/full_trace.png)

#### The endpoints
![alt text](img/endpoints.png)

#### The LLM model endpoint
![alt text](img/llm-endpoint.png)

#### The embeddings model endpoint
![alt text](img/embedding-endpoint.png)

### The Loom video 
https://www.loom.com/share/162d71e4d445442faa40dba76f4cbf13


### Lessons Learned & Open Questions

#### Lessons
1. Learning how to translate notebook code into scripts.
2. Learning/reminder that HF spaces can be used in dev mode and connected to VSCode.
3. Learning how to setup LCEL RAG Chain.
    - Understand how to deploy open-source LLMs & embedding models to scalable endpoints for production-grade LLM & RAG applications
    - Build a RAG application with LCEL
    - Build a front-end UI for RAG applications with Chainlit

#### Questions
1. What are the challenges using LangChain in production - see lots of folks complaining about it on LinkedIn and Twitter. 
2. What complex RAG looks like.
3. Have a basic understanding of the metrics used to monitor performance but still a novice with regards to LLM evals.