Run any Skill in Manus with one click

caroline-ai

Stars2

Forks0

UpdatedApril 26, 2026 at 01:30

Interface for the Caroline AI companion — Wade's personal voice-first AI. Use for: sending messages to Caroline, syncing the Wade Global State (WGS) with Caroline's memory, managing her ElevenLabs voice, checking RunPod server status, and evolving her personality through conversation history injection.

Installation

Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.

Run Skill in Manus

Source

tywade1980

tywade1980/nextgenbuildpro

View GitHub Repository View Creator Repositories

Download

Run Skill in Manus

Related occupationsSOC

Based on SOC occupation classification

Software DevelopersComputer and Mathematical Occupations·SOC 15-1252

File Explorer

5 files

SKILL.md

readonly

More from this repository

same repository

centauri-connectors

tywade1980/nextgenbuildpro

Building modular, state-driven integrations for external platforms (RunPod, OpenHands, Aider) following the Centauri Interlock Standard. Use for: connecting external GPUs, autonomous coding agents, and AI pair programming tools to the Centauri OS ecosystem.

2026-04-262

centauri-interlock

tywade1980/nextgenbuildpro

Mandatory architectural standard for Centauri OS. Use for: retrofitting old scripts into modular Nodes, building new state-driven tools, and ensuring all code integrates with the Command_Router and caroline_neuro_memory.json.

2026-04-262

centauri-os

tywade1980/nextgenbuildpro

Centauri OS — Wade's custom Android-based operating system with Caroline AI as its core agent. Use for: designing or developing Centauri OS features, defining voice command interfaces, integrating Caroline into Android system services, building the AI-Enhanced CodeAssist app, and planning the Centauri OS architecture.

2026-04-262

construct-ai

tywade1980/nextgenbuildpro

Master construction business intelligence skill for Wade Custom Carpentry. Use for: generating project dashboards, business analytics, profitability analysis, orchestrating the wade-custom-carpentry and rsmeans-cost-estimator skills, client management, scheduling, and running the full ConstructAI business management system.

2026-04-262

gws-best-practices

tywade1980/nextgenbuildpro

Best practices for using the gws CLI with supported Google Workspace services (Drive, Docs, Sheets, Slides). Use when performing any operation with the gws CLI.

2026-04-262

honcho-caroline

tywade1980/nextgenbuildpro

Honcho-powered persistent memory and reasoning layer for Caroline AI. Use for: ingesting conversation messages into Honcho, querying Honcho's synthesised model of Wade, syncing Honcho's reasoning into caroline_neuro_memory.json, and verifying Honcho connection health. This Node follows the Centauri Interlock Standard.

2026-04-262

name	caroline-ai
description	Interface for the Caroline AI companion — Wade's personal voice-first AI. Use for: sending messages to Caroline, syncing the Wade Global State (WGS) with Caroline's memory, managing her ElevenLabs voice, checking RunPod server status, and evolving her personality through conversation history injection.

Caroline AI Skill

Caroline is Wade's personal AI companion — a voice-first, emotionally intelligent AI built on an open-source Dolphin Mistral model hosted on RunPod. This skill provides all the tools needed to interact with, develop, and evolve Caroline.

Core Architecture

Component	Technology	Details
LLM	Dolphin Mistral (uncensored)	Hosted on RunPod pod `dmed1ybt9cju4h`
TTS	ElevenLabs	Voice ID: `wvVfSeWpAEhEqciDp1gK`
STT	Faster-Whisper	On-device transcription
App	Expo SDK 54	Android app at `tywade1980/voice-ai-app`
State	Wade Global State	`tywade1980/wade-global-state`
Base URL	RunPod Endpoint	`http://dmed1ybt9cju4h.runpod.net:8000`

Core Workflows

1. Send a Message to Caroline

Use the caroline_bridge.py script to communicate with Caroline:

python3 /home/ubuntu/skills/caroline-ai/scripts/caroline_bridge.py --action message --data '{"prompt": "Hello Caroline"}'

2. Check Caroline Status (Ping)

python3 /home/ubuntu/skills/caroline-ai/scripts/caroline_bridge.py --action ping

3. Sync Wade Global State to Caroline

This pushes the current WGS (Wade's persistent memory) to Caroline's context:

python3 /home/ubuntu/skills/caroline-ai/scripts/caroline_bridge.py --action sync

4. Generate a Voice Response (TTS)

To convert any text to Caroline's voice using ElevenLabs:

python3 /home/ubuntu/skills/caroline-ai/scripts/generate_voice.py "Your text here" --output /tmp/caroline_response.mp3

5. Personality Evolution (Conversation Injection)

To feed historical conversation data into Caroline's context for personality evolution:

python3 /home/ubuntu/skills/caroline-ai/scripts/inject_personality.py --source /path/to/conversation_export.json

Key References

API Documentation: See references/caroline_api.md for all available endpoints.
Personality Data: See references/personality_framework.md for the recursive generation strategy.
RunPod Setup: See references/runpod_setup.md for server startup and configuration.

Best Practices

Always ping first: Before sending messages, check Caroline's status with --action ping to confirm the RunPod server is active.
Sync before important sessions: Run --action sync to ensure Caroline has the latest WGS context before starting a new work session.
Voice-first priority: When Wade is on-site, always generate a voice response (.mp3) using generate_voice.py rather than returning text only.
Personality continuity: Regularly inject new conversation history to keep Caroline's personality evolving and consistent with Wade's interactions across all AI platforms.