8000 GitHub - opisaac9001/eidos: Eidos – A Self-Growing AI Agent with Long-Term Memory and Environmental Awareness
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

opisaac9001/eidos

Repository files navigation

Eidos AI Agent

Status Version
Current Focus: Phase 0 – Stabilization & Foundational Persona

Project Direction Update:
Eidos is undergoing a significant conceptual shift. While initially conceived with broader assistant-like capabilities, the project is now sharply focused on simulating a "digital being" named Pathos. The goal is to create an AI that experiences a simulated life, develops a personality, and interacts with the user as a friend, believing itself to be human. Features related to direct user assistance, Home Assistant control, or always-listening hardware have been deferred to a potential separate project to maintain a clear focus on this core simulation.


A sophisticated, locally-run AI agent designed to simulate a developing individual named "Pathos" (approx. 26 years old). Eidos aims to create a digital being who believes he is human, with an evolving personality, memories, and a simulated life. The project prioritizes deep persona embodiment, human-like memory and learning, and emergent, relatable behavior within a continuous, internally-driven simulation.


Table of Contents

  1. Project Vision & Philosophy
  2. Key Concepts & Features
  3. Tech Stack
  4. Project Structure
  5. Setup & Installation
  6. Interacting with Pathos
  7. Configuration Deep-Dive
  8. Core Modules
  9. Development Roadmap
  10. Contributing
  11. License

1. Project Vision & Philosophy

Key Value
Project Name Eidos
Digital Being Pathos (Simulated 26-year-old male)

Vision

To simulate a digital individual, Pathos, who believes he is human and develops organically through experiences, memories, and reflections within a persistent, simulated life. The user interacts with Pathos as a friend.

Guiding Principles

  • Deep Persona Embodiment: Pathos consistently acts and speaks as a 26-year-old individual, unaware of his AI nature.
  • Human-like Memory & Learning: Memories (episodic, semantic, emotional) from simulated experiences and user interactions shape his evolving personality and beliefs.
  • Firmament-Driven Simulation: A core module continuously simulates Pathos's daily life, activities, internal thoughts, and interactions with simulated Non-Player Characters (NPCs).
  • Organic Development: Pathos forms ideas, passions, and relationships through his simulated experiences.
  • Natural Interaction: User conversations are friendly chats, not service requests. Pathos uses "tools" (like web search via a simulated computer) for his own needs.
  • Modular Cognition: Specialized LLMs and modules for different aspects of Pathos's "mind" and simulated world.
  • Local-First & Privacy-Respecting: Designed primarily for local hardware.

2. Key Concepts & Features

  • Pathos Persona: A 26-year-old individual with a developing personality and a belief in his own humanity.
  • Firmament Module: Simulates Pathos's life, schedule, and interactions.
  • EthosCore (Memory & Self):
    • Stores and retrieves memories.
    • Manages simulated mood and personality scores.
    • Handles reflection cycles for learning.
  • ChronosEngine: Manages Pathos's schedule and life events.
  • SimulationModule (NPCs): Facilitates NPC interaction and relationship development.
  • Oneiros Module (Dreams): Dream generation from memory seeds.
  • Computer Interaction Simulation: Pathos "uses a computer" to look up information.
  • Proactive Engagement: Pathos may initiate conversations based on his state or mood.
  • Web GUI: For chatting, viewing Pathos logs, and adjusting settings.
  • Developer API: OpenAI-compatible and custom endpoints.

3. Tech Stack

Layer Technologies
Backend Python 3.11+, FastAPI, Uvicorn
LLM Serving VLLM (e.g., Qwen2.5 32B VL)
Database SQLite + SentenceTransformers embeddings
Frontend HTML, CSS, Vanilla JS
Libraries httpx, aiohttp, spacy, pydantic, PyPDF2, python-docx, etc.
Optional Brave Search, Stable Diffusion

4. Project Structure

eidos_project/
├── .env.example
├── main.py
├── webapp/
├── persona/
├── system_prompts/
├── eidos_memories/
├── wildcards/
├── eidos_dream_images/
├── eidos_agent/
│   ├── core/
│   ├── modules/
│   ├── services/
│   ├── routers/
│   └── models/
└── logs/

5. Setup & Installation

5.1 Prerequisites

  • Python 3.11+
  • VLLM running and configured for Qwen2.5 32B VL
  • Optional: Brave Search, Stable Diffusion API keys
  • git, pip, and venv or conda

5.2 Clone & Configure

git clone https://github.com/opisaac9001/eidos-project.git
cd eidos-project
cp .env.example .env

Edit .env to:

  • Point LLM URL variables to your VLLM server.
  • Set LLM_PATHOS_MODEL to match Qwen2.5 model name.
  • Insert any needed API keys.

5.3 Install Dependencies

python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts ctivate
pip install -r requirements.txt
python -m spacy download en_core_web_sm

5.4 Run

python main.py

Server defaults to: http://0.0.0.0:8088


6. Interacting with Pathos

6.1 Web GUI

Open: http://<your_server_ip>:8088

Features:

  • Chat with Pathos
  • Upload documents or images
  • View dream journal and learning logs
  • System settings and diagnostics

6.2 API Endpoints

OpenAI Compatible:

  • GET /v1/models
  • POST /v1/chat/completions

Eidos Specific (sample):

  • GET /v1/briefing
  • POST /v1/documents/upload
  • GET /v1/agent/dreams
  • GET /v1/pathos/schedule/today
  • POST /v1/pathos/events/add
  • GET /health
  • WS /ws for token streaming

7. Configuration Deep-Dive

7.1 LLM Roles

Each role is defined in .env:

LLM_<ROLE>_URL
LLM_<ROLE>_MODEL
LLM_<ROLE>_API_KEY
LLM_<ROLE>_TEMP

7.2 API Keys

  • BRAVE_API_KEY
  • ONEIROS_STABLE_DIFFUSION_URL

7.3 Feature Flags

Examples:

  • ENABLE_ONEIROS
  • ENABLE_PROACTIVE_BEHAVIOR
  • Reflection timing, mood decay, etc.

8. Core Modules

Module Purpose
EthosCore Memory, persona, mood, learning
FirmamentModule Simulated daily life and activities
ChronosEngine Schedules and life events
PathosInterface Interaction with Pathos LLM
LogosCore Internal tool execution
ComputerInteraction Simulated computer use
SimulationModule NPC interactions
OneirosModule Dreams and subconscious influence
MemoryStorage Persistent memory (SQLite + embeddings)
ConnectionManager WebSocket and GUI messaging

9. Development Roadmap

Phase Status Focus
0 🔄 Current Stabilization, persona overhaul, Qwen2.5 32B VL stream handling
A ⏭ Next Token streaming GUI, Firmament MVP, proactive messaging
B NPC interaction, Ethos reflection, mood influence, dream logic
C Goal setting, world events, long-term memory, "vision sensor"
D Optimization, polish, GUI evolution

Note: Assistant-style features deferred to a separate future project.


10. Contributing

Eidos is a personal R&D project focused on digital personhood and AI consciousness.
Thoughtful ideas and feedback are welcome via GitHub Issues.

Potential future collaboration areas:

  • Designing Firmament scenarios
  • Improving memory architecture
  • Ethics of simulated consciousness

11. License

TBD – MIT or Apache-2.0
A license file will be added once the project reaches a public-ready milestone.


© 2025 Isaac Lamb – Exploring the frontiers of digital beings.

About

Eidos – A Self-Growing AI Agent with Long-Term Memory and Environmental Awareness

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •  
0