- 60
- 2025
- 25
Hey all,
Big fan of the AI bot chat in the 7.1 repack — one of the coolest things in here. A few ideas to push it further, all building on what's already in
1) Remote / hosted AI backends (Ollama + OpenAI-compatible API)
Today: AI runs local GGUF via embedded llama.cpp only — you can pick a bundled model or drop in your own
Idea: an option to point the bot AI at an Ollama endpoint, or any OpenAI-compatible API (OpenAI, Claude, Gemini, Qwen, Kimi, etc.). Small config: backend / base_url / model / api_key.
Why: the good local models need 7-12 GB RAM (Mistral-7B, Llama-13B) — a hosted/Ollama option gives better chat with zero load on the world server, or lets you run the model on a separate box. Owner picks the cost/quality/privacy tradeoff.
Cheap to build: Ollama and nearly every provider speak the same OpenAI-compatible API, so it's one HTTP client, not five integrations.
2) Custom + per-faction/race bot personas
Today:
Idea: (a) allow a custom free-text persona, and (b) let it vary per faction/race — an Orc gruff and clipped, a Blood Elf haughty, a Tauren calm. Falls back to the global setting.
Why: big immersion win for almost nothing — the bot population stops sounding identical, and owners can theme their realm's voice. Pure extension of the personality system that's already there.
3) Make the AI feature crash-proof
Today: the embedded LLM can take down the whole worldserver on startup if the VC++ runtime is missing.
Idea: bundle the runtime in the installer, and if the AI backend can't start, just disable AI chat and keep the server running (with a log line). An optional feature shouldn't be able to crash the core.
4) Bots reply in the player's language
Today: bot replies are English regardless of the player's client locale.
Idea: have bots answer in the player's language — the server already knows the locale, so with a capable backend from #1 it's basically adding the locale to the prompt.
Why: a big chunk of the community isn't English-first, so this makes the feature feel global.
Even one or two of these would be awesome. Happy to help flesh any of them out. Thanks for the great repack!
Big fan of the AI bot chat in the 7.1 repack — one of the coolest things in here. A few ideas to push it further, all building on what's already in
worldserver.conf under PlayerBots.AI.*. (Credit where due — custom local .gguf models and the 6 personality presets already work well; these are extensions, not "please add the basics.")1) Remote / hosted AI backends (Ollama + OpenAI-compatible API)
Today: AI runs local GGUF via embedded llama.cpp only — you can pick a bundled model or drop in your own
.gguf, but it's all local.Idea: an option to point the bot AI at an Ollama endpoint, or any OpenAI-compatible API (OpenAI, Claude, Gemini, Qwen, Kimi, etc.). Small config: backend / base_url / model / api_key.
Why: the good local models need 7-12 GB RAM (Mistral-7B, Llama-13B) — a hosted/Ollama option gives better chat with zero load on the world server, or lets you run the model on a separate box. Owner picks the cost/quality/privacy tradeoff.
Cheap to build: Ollama and nearly every provider speak the same OpenAI-compatible API, so it's one HTTP client, not five integrations.
2) Custom + per-faction/race bot personas
Today:
PlayerBots.AI.Personality is a single server-wide pick from 6 presets — every bot shares one voice.Idea: (a) allow a custom free-text persona, and (b) let it vary per faction/race — an Orc gruff and clipped, a Blood Elf haughty, a Tauren calm. Falls back to the global setting.
Why: big immersion win for almost nothing — the bot population stops sounding identical, and owners can theme their realm's voice. Pure extension of the personality system that's already there.
3) Make the AI feature crash-proof
Today: the embedded LLM can take down the whole worldserver on startup if the VC++ runtime is missing.
Idea: bundle the runtime in the installer, and if the AI backend can't start, just disable AI chat and keep the server running (with a log line). An optional feature shouldn't be able to crash the core.
4) Bots reply in the player's language
Today: bot replies are English regardless of the player's client locale.
Idea: have bots answer in the player's language — the server already knows the locale, so with a capable backend from #1 it's basically adding the locale to the prompt.
Why: a big chunk of the community isn't English-first, so this makes the feature feel global.
Even one or two of these would be awesome. Happy to help flesh any of them out. Thanks for the great repack!