Dictionaries/JSON files are great for passing the biographical data back and forth between python scripts (or at least for storage of data when the scripts aren't running; once classes are implemented it's probably easier to just pass the data as an attribute of that particular instance of the drone/queen/hive/whatever) and it's the primary format for communicating with the LLMs, but it's probably not a good idea to have the LLM parse the json file itself. that's a lot of extra token usage - biography.json averages around 465 tokens where summary.txt (which parses biography.json for the user) averages 165 tokens.
Everything could be faster and likely better quality if I make a few tweaks like that going forwards.
Dictionaries/JSON files are great for passing the biographical data back and forth between python scripts (or at least for storage of data when the scripts aren't running; once classes are implemented it's probably easier to just pass the data as an attribute of that particular instance of the drone/queen/hive/whatever) and it's the primary format for communicating with the LLMs, but it's probably not a good idea to have the LLM parse the json file itself. that's a lot of extra token usage - biography.json averages around 465 tokens where summary.txt (which parses biography.json for the user) averages 165 tokens.
Everything could be faster and likely better quality if I make a few tweaks like that going forwards.