Pro tip, you can use small models like phi-2 to do semantic chunking rather than just chunk on based on size, and chunking works even better if you book-end chunks with summaries of text coming before/after to enrich the context.
Another pro tip, you can again use a small model to summarize/extract RAG content to get more actual data in the context.
Another pro tip, you can again use a small model to summarize/extract RAG content to get more actual data in the context.