'infinite' character memory storage and retrieval

• "Infinite" character memory is here - currently disabled by default. Enable it in advanced character settings. Your character will create and store memories based on the message history and try to recall them when appropriate. The system can handle millions of memories. It's far from perfect (e.g. character may not always recall the right memories at the right moment), will need more tweaking - please test it out and give feedback!
• Character responses are now streamed (rather than having to wait for the full response to generate before seeing the message)
• Added gpt-4 in the model selector
• Added a Custom Code Helper starter character to help you make your own OpenCharacters plugins
• New format for initial messages: https://github.com/josephrocca/OpenCharacters/blob/main/docs/initial-messages.md
• You can now import .jsonl thread/conversation files from TavernAI
• Wrote a guide to serve it locally: https://github.com/josephrocca/OpenCharacters/blob/main/docs/local-setup.md This is useful if you want to specifically decide when you receive updates, instead of getting updates automatically whenever I push code changes.
• /user <instruction> to get GPT to generate a message for you based on an instruction - e.g. if you're too lazy to write your response and just want to see what happens. E.g. you could write /user take an action, written in first person, and include something about a flower
• Edit the current summary by typing /sum
• Implemented spend tracking. It records data binned hourly, so we'll eventually be able to have some nice charts, but for now it's just aggregated and displayed as a single per-thread value
• Text-to-speech plugin (rudimentary): https://github.com/josephrocca/OpenCharacters/blob/main/plugins/README.md
• New, advanced instruction & reminder message format for pros: https://github.com/josephrocca/OpenCharacters/blob/main/docs/instruction-and-reminder.md
• Doc for explanation on how to run Python code: https://github.com/josephrocca/OpenCharacters/blob/main/docs/running-python-code.md
• There is now a CORS proxy server, so requests made by e.g. the Fire Alarm Bot, Python Coder, and other bots like that will work for any URL on the internet
• You can now 'favorite' chat threads and they'll stay at the top of the thread list
• You can add your own custom memories with /mem - this can be treated kinda like a "lore book". Memories are triggered based on relevant keywords and phrases, so make sure to use good "trigger keyphrases" within your memories. I'll soon add the ability to explicitly define extra triggers that sit "outside" of the actual memory itself.
• You can change your name and avatar picture for a specific thread with /name and /avatar - so you can take on different roles as the user in each thread rather than having one "global" name/role that you need to keep switching when you switch conversations
• Type your response in the input box like normal, and then at the end, on its own line, you can add /ai <instruction> to give the AI an instruction on how it should respond