Installation¶
Install the Flockadoo Provider Client on the machine that will serve models. Confirm it meets hardware requirements: 32 GB of system memory and 24 GB NVIDIA VRAM on Linux; on Windows 32 GB of RAM plus a fixed 64 GB page file (or 64 GB of RAM) and 24 GB NVIDIA VRAM; or 48 GB unified memory on a Mac. If you do not have a provider API key yet, complete Sign up and create a flock first.
The install script places the binary, then runs flockadoo setup. See the setup wizard for what it asks (engine, model, API key) and for non-interactive flags. Re-run anytime instead of editing YAML by hand:
Linux / macOS one-liner¶
Works on Linux (amd64/arm64) and macOS (Apple Silicon). Intel Macs are not supported.
The wizard asks Provider API key or join code. PROVIDER_API_KEY fills the default. It skips the question only when you also pass --yes or --api-key.
The wizard installs:
| OS | Engine |
|---|---|
| Linux | vLLM (needs an NVIDIA GPU) |
| macOS | vLLM Metal or llama.cpp (Apple Silicon) |
Windows (PowerShell)¶
Requires 64-bit Windows. The wizard installs llama.cpp. Native vLLM is not supported. Windows only lists catalog rows tagged llama_cpp. Encoder and embedding models such as baai/bge-m3 stay off Windows — see Approved model builds.
The script does not need Administrator. It installs flockadoo to %LOCALAPPDATA%\flockadoo\bin, runs setup, and adds a Start Menu → Flockadoo → Flockadoo Client shortcut. On Windows the client runs in the notification area by default — closing the terminal does not stop it. See Setup: Windows.
You can also download scripts/windows-install.bat from the flockadoo repo and double-click it (no administrator prompt).
Skip the wizard (scripts / CI)¶
For scripts, CI and Docker, where nobody is there to answer prompts. You run the inference engine yourself and tell the installer which model it serves. Set FLOCKADOO_SKIP_SETUP=1:
export FLOCKADOO_SKIP_SETUP=1
export PROVIDER_API_KEY="your-provider-api-key"
export PROVIDER_TYPE="vllm" # vllm, vllm-metal or llama_cpp
export PROVIDER_MODEL="Qwen/Qwen2.5-7B-Instruct" # the model your engine serves
export LLM_URL="http://127.0.0.1:8000"
export SERVER_PORT="8080"
curl -fsSL https://raw.githubusercontent.com/flockadoo/flockadoo/main/scripts/install.sh | bash
PROVIDER_MODEL is required. It must be an approved model for your engine, and your engine must serve it under that exact name. The install stops with an error if it is not. To list the models this machine can serve, run flockadoo compatibility.
With a Team Flock key or a join code, leave PROVIDER_MODEL unset. The flock's model is used.
The client does not install or start the engine in this mode. The script runs the non-interactive setup for you.
To override defaults (provider type, LLM URL, port), see Configuration.
Launch the client¶
Default config:
Custom config path:
On Windows, start from Start Menu → Flockadoo → Flockadoo Client, or run flockadoo. The client stays in the notification area; right-click the icon and choose Open dashboard for live status. Use flockadoo --console if you want the terminal UI.
Docker¶
The official image is flockadoo/flockadoo on Docker Hub.
If the client runs in Docker but the LLM engine runs on the host, set LLM_URL so the container can reach the host (for example http://host.docker.internal:8000). The engine on the host must listen on 0.0.0.0, not only loopback.
The client listens on 127.0.0.1 inside the container. Flockadoo reaches you through the outbound connection the client opens — you do not need to publish port 8080. To open the local status dashboard from the host, set server.host to 0.0.0.0 and publish the port to loopback (for example -p 127.0.0.1:8080:8080). See Configuration.
Docker quick start¶
docker run -d \
--name flockadoo \
-e PROVIDER_API_KEY="your-key" \
-e PROVIDER_TYPE="vllm" \
-e LLM_URL="http://host.docker.internal:8000" \
flockadoo/flockadoo:latest
Docker Compose¶
version: '3.8'
services:
flockadoo:
image: flockadoo/flockadoo:latest
environment:
- PROVIDER_API_KEY=your-key
- PROVIDER_TYPE=vllm
- LLM_URL=http://host.docker.internal:8000
restart: unless-stopped
Build from source¶
docker build -t flockadoo .
docker run -d \
--name flockadoo \
-e PROVIDER_API_KEY="your-key" \
flockadoo
After first run¶
When the client starts, it publishes your available models with default costs. Open www.flockadoo.ai, go to My Flocks → your flock → Models, and set prices you actually want — see Model pricing.