THE BOT OF FORGING
* Experimental • WIP *

This is a very simple bot I use to help with crafting intro messages, settings, scenarios, lore, lorebook entries, characters, personas, roleplay instructions, etc. The bot has internal instructions to assist and collaborate with the User, and that this scenario is not for active roleplaying.

I use it to brainstorm, workshop, expand and edit ideas, fill-in and create content (per my prompting) and to make characters with pre-defined templates.

Workshopping under this bot lets me keep all my past chats (brainstorming sessions) in the same place, under the same bot in 'My Chats', it's just convenience.

There are 3 built-in styles:

VERBOSE
NORMAL
COMPACT

And 3 built-in Character Templates:

STANDARD CHARACTER TEMPLATE
MINOR CHARACTER TEMPLATE
BACKGROUND CHARACTER TEMPLATE

So I might, for instance, write a bunch of stuff about a character, kinda stream of consciousness style, then ask the BotForge bot to write out the character profile using Normal Style and the Standard Character Template (original template from @stag, with minor edits).

I can also tell it to fill in missing information (make something up) or tell it to only use what I have written. If I don't like the output, i can tweak my prompt and regen the response.

After that, I'll read and tell it what to change, add or tweak. Sometimes I'll edit manually. I sometimes also ask for suggestions, ask for it to look for inconsistencies or things which are missing or check for redundancies. If the output gets too long, I can tell it what parts to excise, or ask it to compress, simplify or write more efficiently... whatever gets the results I want.

Once I have a draft I am happy with, I don't just use the LLM output verbatim. I'll take the draft and paste it into Notepad++ and do some manual edits and minor wording changes, as I prefer and to fit my style, before placing the content into a WIP bot.

It's really just a tool: a time-saver and a writing-process aid, and I find it works really well with GLM 5.1 or 5.2. GLM doesn't write the usual 'GPT-isms' or 'DeepSeek-isms' in its output. It's a reasoning model and doesn't make many mistakes, and keeps details straight most of the time. It lets me brainstorm and write freely while it does the work of consolidating, organizing the information, filling in some gaps, and making suggestions.

---

High-context proxy is pretty strongly recommended. I use GLM 5.1-5.2 and get excellent results.
It will probably also work well with GLM 4.7, Kimi K2.6, DeepSeek v4 Pro, DeepSeek 3.2, DeepSeek v2 0324, Gemini, or others.