Flywheelby OpSpot

Platform

Train a private model

Model-Training-as-a-Service: we train a model on your specific business — warm-started from the niche closest to your trade — and host it as a private, single-tenant model only your API keys can call.

What you get

A private, single-tenant modeltuned to your business — your services, hours, policies, and voice. It is never pooled with anyone else and is reachable only by your organization’s API keys. Because it’s built on the same base as every Flywheel niche model, it runs the same OpenAI-compatible API and the same runners — nothing else in your integration changes.

How it works

We don’t train from scratch. We warm-start from the shipped niche model closest to your trade — so a dental office starts from healthcare-frontdesk, an auto shop from automotive — then continue training that model on your business profile. The result keeps the trade expertise the niche already learned and adds your specifics on top.

  1. Intake. You describe your business (services, hours, policies, FAQs, voice) on the train page.
  2. Warm-start training. We continue-train the closest niche model on your profile as a single-tenant adapter.
  3. Package & publish. The model is packaged and stored in a private repository only our inference servers can read.
  4. Live. It appears on your train page as Live and answers immediately on your existing key.

Get started

Open the train page, pick the closest trade, and fill in your business details. The more real FAQs and policies you give, the sharper the model. You can refine the profile and re-train at any time — a re-train rolls forward automatically.

Tip.Already on a Hosted API plan? Your private model uses the same key and the same billing — there is nothing new to wire up. It just becomes another model id you can call.

Calling your model

Once it’s live, call it like any other model — use the id private (or the exact cust-<slug> id shown on your train page):

python
from openai import OpenAI

client = OpenAI(base_url="https://YOUR-DOMAIN/api/v1", api_key="fw_live_…")

resp = client.chat.completions.create(
    model="private",                 # your trained model — or its exact "cust-<slug>" id
    messages=[{"role": "user", "content": "What are your Saturday hours?"}],
)
print(resp.choices[0].message.content)

Isolation & privacy

Your model is a single-tenant build scoped to your organization. Only your API keys can call it — a request for it from any other organization’s key returns a not-found, with no signal that it exists. Training data you provide trains your model only and is never pooled into a shared model. For liability-bearing or privileged verticals, this single-tenant path under a DPA keeps your data fully isolated.

Heads up.A private model is a commercial hosted product, not an open-weight release. Unlike the public niche weights, it is not published to Hugging Face — it stays private to your organization.

Retrains & deletion

Send us more data whenever your business changes and we re-train; the new version becomes live and older versions age out. Withdraw consent or request erasure at any time — your data is excluded from future training within our committed SLA and the affected model versions are queued for retrain and deprecation. The full terms live on the consent page and in the DPA.

← The flywheel