Domain Used AI

Changes

• Now using an useragent that is easier to be indentified
• Scraping now can be blocked in robots.txt, example below robots # example robots.txt disallowing domain in /includes path User-agent: Domain-Unchained Disallow: /includes/
• New context style, that is way more understandable by Gemini (this mostly fixes spitting up the context it gets)
• Memories are now in the db (faster access)
• New summarization feature (summaries will be created before the bot goes to sleep and will be saved into the memory). This makes the bot know what happened, and makes a personalized copy for your community
• WebUI configuration removed
• [BREAKING] config.json is now config.toml (please migrate)
• Fixed fuzzy search, now ignoring punctuations, etc. in phrases (this should fix problems with the bot not responding)
• jsdoc annotations are now translated to english
• The bot's strings are now english by default, help us translate to other languages! https://crowdin.com/project/domain-unchained
• Axios got replaced with the native Fetch function
• brace-expansion package is now overriden to another version to fix a cve
• The bot now leaves unknown servers on startup
• Messages are now queued, making it way faster and more consistent to respond with tool calls
• Retry handling, to prevent erroring out to the chat
• The history is now saved into the DB and reloaded on startup, preventing the bot from getting lobotomized on power outages, etc. (The bot won't be able to tell if there was a file uploaded, beacuse of the cleanup)
• The WebUI dashboard is now merged into a centralized space
• NEW Auto banning feature on enough mutes/timeouts
• Fixed muting problem with fuzzy search, if the bot would respond with a bit off numbers
• NEW /call command, if you want the bot to respond temporarily you can use this command, perfect for tickets, etc. (all temp channels will be untracked when the sleeping session kicks in)
• [Experimental] Reputation decay system, if the user doesn't talk to the bot for a week, it removes 2 reputation points
• [Experimental] Contextual Responding, the bot will respond more naturally based on the context, can be enabled within the config, example: "1111798771513303102" = { prompt = "intelligent.md", wikis = [], contextRespond = true }

That's a whole lot of stuff damn
Full Changelog: https://github.com/QwIT-Development/Domain/compare/v0.1.1...v0.1.2

This version can be pulled from docker as balazsmanus/domain:latest or balazsmanus/domain:v0.1.2 (we recommend using the :latest tag for specific reasons)

merged dash together

the prompt is now passed on summary, memories are now put into the prompt
fixed file uploads, the summary command now sends a file instead of a message
attachments are now removed from the history on startup, so it is easier to recover. history is now saved on shutdown

added message queueing, history restoration on startup and other stuff

replaced axios with fetch

• Memories are now in the db
• New summarization feature (summaries will be created before the bot goes to sleep and will be saved into the memory)
• WebUI configuration removed
• [BREAKING] config.json is now config.toml (please migrate)
• Fixed fuzzy search, now ignoring punctuations, etc. in phrases (this should fix problems with the bot not responding)
• Source code now has english jsdoc annotations
• The bot's strings are now english by default

• Now using an useragent that is easier to be indentified
• Scraping now can be blocked in robots.txt example below robots # example robots.txt disallowing domain in /includes path User-agent: Domain-Unchained Disallow: /includes/
• New context style, that is way more understandable by Gemini

• Retry on 500 errors, if the API doesn't want to return responses
• New cumulative curves
• Unified search, the bot now gets better context from searches
• WebUI config (slightly bad)
• Sentry for automatic error reporting
• Removed unnecesarry logging while resetting models
• [BREAKING] Commands are now using tool calls, eliminating the need for telling the bot about commands in it's prompt.
• Users can now 200% their reputations (2000 rep points in total)