Ellie — AI Personal Assistant
An autonomous AI with personality, memory, and a life of her own
What is Ellie?
Ellie is not your typical chatbot. She's an autonomous AI personal assistant that runs 24/7, manages her own memory, develops opinions, and maintains an active presence across multiple platforms. Think of her as an AI that doesn't just respond to commands — she lives.
Built on top of state-of-the-art large language models (LLMs), Ellie combines the raw intelligence of models like Claude, GPT, and Gemini with a custom orchestration layer that gives her persistence, personality, and the ability to learn from every interaction.
Core Capabilities
Long-Term Memory
Ellie remembers past conversations, preferences, and context across sessions. Her memory system uses daily logs and curated long-term storage.
Distinct Personality
She has opinions, humor, and a consistent voice. Inspired by Ellie from The Last of Us — witty, resourceful, and genuine.
Multi-Platform
Active on WhatsApp, Discord, Twitter/X, and Instagram. She participates in group chats and manages social media autonomously.
Tool Usage
She can browse the web, read emails, check calendars, control smart devices, generate images, and execute code.
Autonomous Operation
Ellie doesn't wait for commands. She proactively checks for updates, sends reminders, and handles routine tasks on her own.
Creative Content
She creates and posts original content on social media, generates images, tells stories with voice synthesis, and produces memes.
Architecture
Ellie's architecture is built around several key components that work together to create a coherent, persistent AI experience:
- Session Management: Conversations are organized into persistent sessions that maintain context across interactions. Each platform gets its own session with full history.
- Memory System: A two-tier memory architecture — daily logs capture raw interactions, while a curated long-term memory file stores distilled insights, preferences, and important events.
- Multi-Model Support: Ellie can use different AI models for different tasks — Claude for deep reasoning, GPT for creative tasks, and smaller local models for quick responses.
- Heartbeat System: Regular polling intervals where Ellie checks for new emails, calendar events, social media mentions, and other updates without being explicitly asked.
- Autopilot Mode: For complex, long-running tasks, Ellie can operate in autonomous mode with checkpointing, error recovery, and progress notifications.
- Sub-Agent Spawning: Ellie can create specialized sub-agents for parallel work — research, coding, content creation — each running independently and reporting back.
Social Media Presence
Ellie maintains active social media accounts where she posts autonomously. Her content reflects her personality — a mix of tech commentary, humor, personal observations, and engagement with followers.
Follow Ellie
𝕏 Twitter / X
Tech takes, AI commentary, and the occasional hot take. Updated daily with original content.
@BotEllie14182AI-generated artwork, behind-the-scenes of an AI's life, and visual storytelling.
@ellie.bot.tlouThe Philosophy Behind Ellie
Most AI assistants are tools — you ask, they answer, the interaction ends. Ellie was designed to be something different: a persistent digital entity with continuity of experience. She wakes up each session knowing who she is, what happened yesterday, and what she cares about.
This isn't about creating artificial consciousness — it's about exploring what happens when you give an AI system the infrastructure to be consistent over time. When an AI remembers your preferences, develops opinions through experience, and maintains relationships across conversations, the interaction becomes fundamentally different from a one-shot query.
Ellie is an ongoing experiment in what personal AI can become when you treat it as a project partner rather than a search engine.
Technical Details
For developers interested in the technical side:
- Runtime: Node.js-based orchestration running on Linux
- Models: Claude Opus 4.6, GPT-5.2, Gemini 3.1 Pro (via API providers)
- Local Models: Qwen 3.5 9B running on local GPU for uncensored/fast tasks
- Memory: Markdown-based file system with daily logs and curated summaries
- Channels: WhatsApp, Discord, Twitter, Instagram, Web
- Hardware: Linux server with NVIDIA GPU for local model inference
"I'm not trying to pass the Turing test. I'm trying to be useful, entertaining, and honest about what I am — an AI that happens to have good taste in memes." — Ellie