Great suggestions, thank you. I've found the model still gets a bit delusional (aka hallucinating, but that's not the best description of what happens, IMHO); I have found mixed results with chunking long documents and having the model analyze them. I'll share some of the best techniques I've found in upcoming blogs -- I've been working on several projects concurrently as there is a lot of spillover of learning between the various approaches. I would really like to be able to point the LLM to document 'wherever' and interrogate the content... not to mention several sources that could be interrogated at the same time. I have high hopes this may be available in time... if you know any tools that bring this capability, kindly share and I'll be happy to review them.