depwl9992 / faith-n-tech_d7d Goto Github PK

Add a telnet client that can log into MUDs and offer to DM for live players.

Should have Diku, MUX and custom login string capability. This would be like a Discord bot where credentials would be stored somewhere untracked. It would also have to trigger on say and emote commands from within the MU* and parse them properly, so a configurable string would be necessary.

The bot player should have color turned OFF so we don't have to parse xterm or ansi color codes.

As you can see, the keywords and character sheet are optional, but numeric values must be explicitly provided by the user. Would be nice if a blank (or invalid) string just assumed a default for Number of Stages, Minimum Substages, and Maximum Substages.

Offer the bot a commandset that it can call back to the script to generate automatic responses like dice rolls, requests for brief character details, etc.

If I try to go very off topic or make a choice that is not what the bot was expecting (e.g. my character goes off to smell flowers or brainlessly continues escaping the trap without trying something different), the bot will freeze and grind for minutes or many minutes, and never (relatively speaking) come back with a response. Ctrl-C/Ctrl-Brk did not cancel the process and let me try something else, or exit cleanly, or even exit the venv at all without force-closing the terminal.

Generate a JSON structure which represents an adventure module.

• #12
• #13

Framework

Tree structure rather than "MUD Grid"

Details

• NPCs that we can interact with
• Places we can go
• Choices we can take

Know what we did, where we are, and how to loop us until we complete a node. How does it know when to navigate us to a new node? We may need it to present us with a numbered list of options.

Simulate the API controlling the program - a reverse API that the bot calls to your script.

1. The program knows the 3 options that a player can pick from to move forward.
2. The bot presents the question, 'what do you want to do (in plain text)?' or 'what item from this numbered list do you pick?'
3. The player responds with what they want to do.
4. If the response is plaintext, present the BOT (hidden) with 'what prompt that you have stored to move us to the next node CLOSEST matches what the player wants to do?' using the reverse DM-roller routine.

Linking

Every action we take should dump the adventure to a log, and then we should have a subroutine that builds the 'what we did last time' setting and allows us to continue an adventure where we left off from session end/crash/etc.

The current quest state should have a premise/summary, chapter/node id, and description. When you enter any node, the DM provides the entry point into the node.

• #9
• #10
• #11

From a terminal works, but is not always global. Add an option to load in as a Discord bot.

Remove D&D specific verbiage from the project so players can play any type of RPGs (including D&D games). Why limit ourselves? :)

1. For each step and user interaction made, dump to a text log.
2. To avoid hitting the max token length error or exponentially increasing transaction sizes of longer adventures, remove oldest entries from the dialog object (they will still be stored in the local log).
3. On Session Start, read the last entries of the log, stepping backward in time until the setting can be recalled.

Current DM fumbles on rules and facilitating play. We need to pre-train the GPT model with D&D 5e rule books to ensure adherence to the rules.