01
Inference edge
OpenAI-compatible /v1/chat/completions and /v1/models. Streaming SSE supported. Clients never see provider base URLs or upstream model ids.
Product
Fabrix owns authentication, provider abstraction, usage tracking, and prepaid billing so product teams stop wiring each vendor separately.
01
OpenAI-compatible /v1/chat/completions and /v1/models. Streaming SSE supported. Clients never see provider base URLs or upstream model ids.
02
Organisation Login manages Projects, API Keys, Razorpay top-ups, low-balance alerts, and appeals. One shared dashboard identity per Organisation.
03
Platform Admins define Fabric Models and Token Prices, map Provider Models, grant Model Access, credit wallets, and suspend Organisations.
Customer-facing slugs (for example apparatus-core) with USD token prices. Each maps to exactly one Provider Model in v1.
Allow-lists which Fabric Models an Organisation may call. Remap upstream without changing client code.
Every settled call records tokens, Provider Cost, customer charge, latency, and billing mode for ops review.
Apparatus internal companies can run without wallet debit while still metering tokens and upstream cost.
Optional monthly hard caps and soft thresholds. Spend resets each calendar month in Asia/Kolkata.
Responses echo Fabric Model slugs only. Upstream names and base URLs stay with Platform Admins.
01
Admin creates the Organisation and grants Model Access.
02
Top up the wallet, create a Project, rotate a key.
03
Point SDKs at Fabrix and request Fabric Model slugs.
https://gateway.fabrix1.comReach out anytime
Docs cover authentication, Chat Completions, models, rate limits, and billing behaviour.