ForgeOS — Persistent Memory for Businesses

Inspiration

As AI agents become more capable, they're increasingly being asked to complete multi-step work that normally requires an entire team. The problem is that most agents are stateless—they forget previous work, cannot coordinate effectively, and have no persistent organizational memory.

We wanted to explore a different idea:

What if AI agents could operate like a real company instead of isolated chatbots?

ForgeOS was built as an operating system for AI organizations. Instead of a single assistant answering prompts, ForgeOS lets specialized agents collaborate, remember previous work, execute missions, and continuously improve over time.

This aligns closely with the vision of persistent AI systems powered by Qwen.

What it does

ForgeOS transforms a natural language request into an organized execution workflow.

Users can:

create missions
compile organizational knowledge
maintain long-term memory
coordinate specialized AI workers
persist previous work between sessions

Rather than generating isolated responses, ForgeOS builds a persistent organizational knowledge base that agents can reuse across future tasks.

How we built it

The project is deployed entirely on Alibaba Cloud ECS.

Our stack includes:

Alibaba Cloud ECS for deployment
Qwen via DashScope API for reasoning
Docker & Docker Compose
FastAPI backend
SQLite persistent storage
REST API
Browser frontend

The deployment automatically:

provisions the application
configures Docker
connects to DashScope
launches the production server
verifies deployment through an automated proof script

The backend communicates directly with Qwen models through the DashScope compatible API while storing persistent organizational memory locally on ECS.

Challenges we ran into

Deployment turned out to be one of the biggest challenges.

Some of the issues we solved included:

ECS account risk-control restrictions during provisioning
Docker installation differences on Ubuntu images
missing Docker Compose packages
security group configuration preventing external access on port 8000
ensuring DashScope authentication worked correctly inside the deployed environment
creating an automated deployment proof that verifies both ECS and Qwen connectivity

These challenges ultimately helped us produce a more reliable deployment workflow.

What we learned

Building ForgeOS reinforced an important idea:

LLMs become significantly more useful when combined with persistent memory, structured execution, and cloud infrastructure.

We also gained valuable experience deploying AI-native applications on Alibaba Cloud ECS and integrating Qwen through DashScope in a production environment.