optional.txt

1. [X] Implement streaming responses.

?? Uses dynamic few-shot prompting, where the best examples are selected according to the user query.

2. [] There’s code for RAG evaluation in the folder, and the README contains the evaluation results. 
The folder must also contain the evaluation dataset and the evaluation scripts.
https://academy.towardsai.net/courses/take/beginner-to-advanced-llm-dev/multimedia/59791119-evaluating-your-rag-pipeline

3. [X] The app is designed for a specific goal/domain that is not a tutor about AI. For example, it could be about finance, healthcare, etc.

Use live search results. In this case, the user must also input the API keys for the relevant APIs (e.g., Perplexity, Bing search, etc.).

4. [X] You have shown evidence of collecting at least two data sources beyond those provided in our course. 

Your data collection and curation process leverages structured JSON outputs, which are used for advanced RAG functionalities in your app.

Your data collection and curation process leverages images and/or PDFs. The parsed data should be useful for some use cases in your app.

The app can generate images on the fly. They must be pertinent to the scope of your app.

Use a reranker in your RAG pipeline. It can be a fine-tuned version (your choice).

Use hybrid search in your RAG pipeline.

5. [ ] Use metadata filtering.

Use a fine-tuned LLM in your app.

Use a fine-tuned embedding model in your app. 

Your RAG pipeline includes query routing.
https://academy.towardsai.net/courses/take/beginner-to-advanced-llm-dev/multimedia/59791736-adding-question-validation-and-routing

Your query pipeline includes function calling.

Your app manages speech inputs.

Your app manages speech outputs.

Your app appropriately leverages context caching (also known as prompt caching). Explain in the README how you designed your prompts to be adequate for context caching.
https://academy.towardsai.net/courses/take/beginner-to-advanced-llm-dev/multimedia/59791739-long-context-llms-context-caching-vs-rag