BuildOS Full Platform
Full agentic AI platform tying together a Next.js dashboard, FastAPI backend, vector-search knowledge layer, and a Node.js runtime for orchestrating agent-driven infrastructure tasks.
Overview
BuildOS Full Platform is the production build-out of the BuildOS ecosystem: a Next.js 15 dashboard talking to a FastAPI backend, with Postgres/pgvector as the knowledge and semantic-search layer, Redis for queuing, and a separate Node.js runtime service that executes agent-issued operations and reports results back over WebSockets.
The Problem
Earlier BuildOS prototypes were single-process tools — useful for one task at a time, but with no durable state, no way to search accumulated knowledge semantically, and no clean separation between "the agent decides" and "the agent executes."
The Solution
BuildOS Full splits these concerns into dedicated services:
- Dashboard (Next.js 15): the operator-facing UI.
- API (FastAPI + SQLAlchemy2): async request handling, backed by Postgres.
- pgvector: semantic search over stored knowledge and context, not just keyword lookup.
- Redis: queuing and caching between services.
- Node.js runtime: a dedicated execution service, connected over WebSockets, that carries out agent-issued operations and streams results back — keeping "decide" and "do" cleanly separated.
Architecture
+-------------------+ +-------------------+
| Dashboard (Next) | <----> | API (FastAPI) |
+-------------------+ +-------------------+
| |
v v
+---------+ +-----------+
| Postgres| | Redis |
| pgvector| | (queue/ |
| | | cache) |
+---------+ +-----------+
^
| WebSocket
v
+------------------+
| Node.js Runtime |
+------------------+
Technical Decisions
- pgvector over a separate vector DB: keeps knowledge and relational data in one Postgres instance instead of syncing two systems.
- A dedicated Node.js runtime, separate from the FastAPI API: agent-issued operations run in their own service over a WebSocket channel, so a long-running or risky operation can't block the API layer.
Roadmap & Future Work
- Expand the knowledge layer with more automated ingestion sources.
- Tighter integration between the Node.js runtime and the homelab's real infrastructure (BuildOS Node Commander).