System Requirements

This page lists everything you need to run a self-hosted Snippbot instance, from operating system and runtime requirements to ports and optional dependencies.

---

Operating System

Platform	Support Level	Notes
Linux (Debian/Ubuntu/RHEL)	Fully supported	Recommended for production
macOS (12+)	Fully supported	Good for development
Windows (10/11)	Community support	WSL 2 strongly recommended

---

Runtime Requirements

Requirement	Minimum	Recommended
Python	3.11	3.12
Node.js	20.x LTS	22.x LTS
pnpm	9.x	Latest 9.x
RAM	2 GB	4 GB or more
Disk	5 GB	20 GB or more
CPU	Any modern x86_64 or ARM64	2+ cores

Note

Disk usage grows over time with agent memory, workspace files, and conversation history. For production deployments with multiple active agents, allocate at least 20 GB.

---

Optional Requirements

Docker or Podman

Required only if you want to use the sandbox feature, which runs agent-generated code in isolated containers.

Tool	Minimum Version
Docker	24.x
Podman	4.x

The Docker socket must be accessible to the Snippbot process. On Linux, add the snippbot user to the docker group:

sudo usermod -aG docker $USER

GPU (for local Whisper STT)

Snippbot supports optional local speech-to-text via OpenAI Whisper. A GPU is not required, but significantly improves transcription speed.

Hardware	Notes
NVIDIA GPU (CUDA)	Best performance; install CUDA 12+ and cuDNN
Apple Silicon (MPS)	Supported on macOS 12+ via PyTorch MPS
CPU-only	Works but is slow for long audio; use whisper-tiny or whisper-base model

Browser Automation

The browser tool (used for web scraping, form automation, and the browser panel in the UI) requires Chrome or Chromium with the Chrome DevTools Protocol (CDP) enabled.

Option	Notes
Google Chrome	Recommended; install the stable channel
Chromium	Works; available in most Linux package managers

Snippbot launches Chrome automatically when the browser tool is first used. No manual configuration is needed if Chrome or Chromium is installed in a standard location.

---

Ports

Port	Service	Required
18781	Snippbot API + UI	Yes
18790	Channel adapter (webhooks from Slack, Telegram, etc.)	Only if using channels
5173	Vite UI dev server	Dev mode only

Tip

In production, you typically run Snippbot behind a reverse proxy (nginx or Caddy) and only expose port 443 publicly. The reverse proxy forwards requests to the internal ports. See the Reverse Proxy guide.

---

Network Requirements

Outbound

Snippbot requires outbound HTTPS access to the AI provider APIs you configure:

Destination	Purpose
api.anthropic.com	Anthropic Claude models
api.openai.com	OpenAI models (optional)
api.groq.com	Groq inference (optional)
Your Ollama endpoint	Local models via Ollama (optional)

Inbound

Inbound access is only required if you use messaging channels (Slack, Telegram, etc.) that deliver webhook events to your server.

Scenario	Required inbound
UI + API only	None (outbound only)
Channel adapters	Port 18790 reachable from the internet (or via reverse proxy on 443)

---

Database

Snippbot uses SQLite for all persistent storage. No separate database server is required.

Property	Detail
Engine	SQLite 3 (bundled with Python)
Location	~/.snippbot/snippbot.db
Backups	Copy the .db file; it is a single portable file

Note

For high-concurrency production deployments with many simultaneous users or agents, PostgreSQL support is planned for a future release. SQLite is well-suited for single-user and small-team deployments.

---

Recommended Production Configuration

For a reliable self-hosted production installation:

Resource	Recommendation
Provider	Any VPS (DigitalOcean, Linode, Hetzner, Vultr, AWS Lightsail)
CPU	2+ vCPUs
RAM	4 GB
Disk	20 GB SSD
OS	Ubuntu 24.04 LTS
TLS	Caddy (automatic Let’s Encrypt) or nginx + certbot

A $12–$20/month VPS from any major cloud provider comfortably handles Snippbot for a team of 5–10 users with several active agents.