The outcome

Keep production inference observable, bounded, recoverable, and isolated across environments.

FriendliAI is a AI inference platform with serverless model APIs, dedicated GPU endpoints, private containers, observability, and OpenAI-compatible interfaces. Serving open models through instant APIs, deploying pinned Hugging Face models on dedicated GPUs, running private inference containers, and operating production inference with explicit scaling and observability. This guide narrows that broad capability into one repeatable outcome, with checkpoints that keep the source material and your judgment in the loop.

Before you begin

Set the boundary before the tool starts.

Choose one real task, identify who will use the result, and decide what evidence or test will make the result acceptable. Gather only the source material needed for that task. If the work contains confidential, personal, regulated, or client-owned information, confirm that the platform and account are approved before sharing it.

Troiana principle

AI should make the work easier to inspect. If the workflow removes the source, the owner, or the review step, redesign the workflow.

Step by step

A workflow you can repeat.

  1. 01

    Separate projects, teams, and keys by environment and tenant, define roles, model rights, data handling, retention, logging, SLOs, budgets, and incident ownership.

  2. 02

    Store keys in a secret manager, restrict deployment operations, pin model revisions, cap input, output, batch, and replicas, and require reviewed version comments for updates.

  3. 03

    Monitor endpoint phase, replicas, GPU memory, queue, tokens, latency, errors, content-logging state, availability, and spend using non-sensitive metadata.

  4. 04

    Test sleeping and wake-up, failed download, invalid token, memory exhaustion, scale limits, provider outage, bad version, key rotation, rollback, and termination.

  5. 05

    Canary every update, retain a prior version and alternate endpoint, audit access and costs, delete retired resources, and reconcile that termination stopped billing.

Working standard

What good use looks like.

  • Separate deployment and inference credentials.
  • Canary every endpoint version.
  • Verify termination stops billing.

OpenAI compatibility does not guarantee identical tokenization, tools, errors, moderation, or output. Dedicated endpoint management APIs are marked beta, content logging is configurable, and a model repository can change unless its revision is pinned. Verify model rights, keep keys server-side, minimize logged content, separate environments and teams, cap autoscaling spend, test sleeping and failure states, and never update a production endpoint without an evaluated version and rollback.

Official references

Check the current product documentation.

Features, plan limits, availability, and data controls change. These official pages are the starting points used for this collection.