It reads and files pages
Your container takes a share of the reading list, works through it, and hands back finished, searchable knowledge. Every page it files is a page nobody has to wait for later.
Indexer · Docker
ONE
COMMAND
Any Docker host.
Windows, Linux or macOS.
If it runs Docker, it can run a node.
A single container that does a small share of the reading that keeps the index fresh. It uses only what the host can spare, touches nothing else, and comes off with one command.
What the machines have done so far
Every one of these is a container on someone's machine.
No datacentre, no rack, no procurement — just spare capacity, wrapped in a container. The figures below come straight from the fleet as it runs, and they move while you read them. Adding a host moves them faster.
Connecting…
—
Pages searchable
everything the fleet has filed
—
Machines online
dedicated to reading
—
Processor cores
spare capacity, pooled
—
Pages read by the fleet
these machines' own tally
—
Web pulled down
read, cleaned, filed
—
Shared card throughput
across — shared graphics cards
What the container actually does
Four jobs, none of them yours.
Once it's running there is nothing to operate. It takes work when the fleet has work, pauses when the host is busy, and never asks you for anything.
It rescues the hard ones
Some pages refuse to be read by anything simple. Those get passed around the fleet until a machine can handle them — and your node takes its turn.
It shares a graphics card
With the NVIDIA container toolkit installed, the container joins the pool that turns text into something searchable — the single most valuable thing a node can offer.
It stays in its lane
The container runs isolated, with a hard ceiling on memory and processor time. It reads the public web — never the host's files, and nothing outside the container it's given.
Step one · run
One command, then walk away.
With Docker running, paste the line below into a terminal. On Windows use a WSL, Git Bash or Docker Desktop shell. It pulls the image, starts a capped container named askfinz-indexer, and registers it with the fleet.
Any Docker host · Linux, macOS, Windows (WSL)
curl -fsSL https://get.askfinz.ai/docker | shAny askFinz account can add a node. askFinz is in private access today, so the way in is to request an account — your node key comes with it.
- Nothing on the host changes — it's just one more container.
- The container is capped before it starts, not throttled after the fact.
- It picks its own name and number — you never configure anything.
askfinz · docker run
Illustration of a first run. Your node number is assigned by the fleet.
What you need
Docker, and not much else.
If a machine already runs Docker, it can run a node. Each thing you add unlocks a little more of what the container can take on — but none of it is required to be useful.
| Tier | What it needs | What it unlocks |
|---|---|---|
The host | Docker Engine or Docker Desktop, 64-bit | Anywhere Docker runs works — a Linux server, a Mac, or Windows via Docker Desktop / WSL2. Nothing else on the machine changes. |
Memory | 8 GB or more recommended | The container claims a slice, never the machine. More memory simply means it takes on more reading at once. |
Connection | Always-on broadband | Steady reading. If the host sleeps or restarts, the container resumes and rejoins the fleet on its own. |
Graphics card | Optional — NVIDIA + container toolkit | The shared embedding pool. Pass a GPU through with the NVIDIA container toolkit and the node joins it; without one it's still a full member of the fleet. |
Step two · verify
Find yourself on the live board.
Every node in the fleet shows up on a public status page. Your container appears within a few minutes, under the name it was given. That page is the only confirmation you need — or run docker logs askfinz-indexer if you like watching.
A healthy node climbs. If uptime is going up and the page count is going up, it is doing its job — the rest of the row is detail.
A healthy row
- Name
- askfinz-indexer-021 assigned to you at start
- Uptime
- 6h 12m climbing, not resetting
- Pages
- 8,417 rising every refresh
- Reading
- 4 at a time sized to the host
- Shared card
- Active only with GPU passthrough
Illustration. Live values are on the status page.
If something looks wrong
Five things people ask about.
Most of these aren't faults at all. Nodes are built to recover on their own, so the usual answer is to give it a few minutes.
The command can't reach Docker
Docker isn't running, or your user can't talk to the daemon.
Start Docker Desktop (Windows/macOS) or the Docker service (Linux), then run the line again.
Nothing appears on the status page after 10 minutes
The image was still pulling, or the container hadn't registered yet.
Give it a few minutes on the first run. Check `docker logs askfinz-indexer` if you're curious.
Your node was listed, then disappeared
The host slept, lost its connection, or the container was stopped.
Bring the host back and give it a couple of minutes. The container rejoins on its own and keeps its number.
No shared-card badge on your row
The NVIDIA container toolkit isn't installed, or no GPU is passed through.
Install the toolkit and re-run the command to enable GPU sharing. The node is still a full contributor without it.
Pages sat at zero for the first few minutes
Nothing is wrong. A new node waits to be given its share of the reading list.
No action. Pages start climbing once work is handed to it.
Whenever you like
Leaving is one command too.
No notice period, nothing to negotiate. Remove the container and the node hands back whatever it was working on and releases its number to the pool. The host keeps only the image, which you can drop too with docker rmi askfinz/indexer.
Any Docker host
docker rm -f askfinz-indexerNothing is left behind — no leftover service, no stray folder, no account to close.
askfinz · container removal
Questions
Before you
run it.
- Which operating systems does this work on?
- Any host that runs Docker: Linux (Docker Engine), macOS and Windows (Docker Desktop, which uses WSL2). The same one-line command works on all of them — on Windows run it from a WSL, Git Bash or Docker Desktop terminal.
- Will it slow the machine down?
- It shouldn't. The container runs with a hard ceiling on memory and CPU and at low priority, so anything else on the host takes precedence. You can stop it at any moment.
- Can it see my files?
- No. It runs isolated in its own container and reads public pages from the web. It has no access to the host's files, browser or accounts beyond what the container is given.
- How do I share a GPU?
- Install the NVIDIA container toolkit on the host and re-run the command — the installer detects it and passes the card through so the node can join the shared embedding pool.
- How do I remove it?
- One command — `docker rm -f askfinz-indexer` — stops and deletes the container. It hands back its work first, and its node number returns to the pool. Nothing is left on the host.
Other ways to run one