API Reference#

The InferencePort AI Lightning backend exposes these hosted API services on https://api.inferenceport.ai.

Generation API: The Generation API (/gen) is the subscription-backed API used for cloud chat, image, video, and audio generation. It resolves your plan’s daily usage quotas and is best for regular app-like workloads.
Pay-2-Go (P2G) API: The Pay-2-Go (P2G) API (/v1) is the credit-backed API for enterprise and production integrations. It meters each request against your wallet balance rather than plan quotas.
Embeddings API: The Embeddings API (/v1/embeddings) is an OpenAI-compatible text embedding endpoint, billed per input token from your Pay-2-Go wallet.
Decision models / System One API: The Decision API (/v1/systemone) evaluates typed choice, noul and score questions with probabilities using configurable providers and P2G pricing.
Realtime API: The Realtime API (/v1/realtime) is a WebSocket for live voice conversations and streaming transcription over the OpenAI Realtime protocol, billed while the session runs.
AI Shield: The AI Shield (/ai-shield) is a detection API for abuse, fraud and attacks: prompts and agent tool output, signups and payments, emails and links, HTTP requests, file uploads, passwords, listings, domains, tokens and more. Send signals (or use the single-purpose /check endpoint) and get back a risk score, decision and detailed evidence.
Rewards System: The Rewards System (/v1/rewards) lets users earn badges and unlock discounts by using the platform. Rewards are earned automatically based on spending, usage diversity, request volume, and streak milestones.

Which API should I use?#

Use the Generation API when you want the same behavior as the app — chat, generate images, video, and audio with subscription-based daily limits and automatic router model selection.
Use the P2G API when you need predictable credit-based billing, enterprise isolation, a stable production API surface, or want to select specific override models, embeddings, decisions, or realtime voice.
Use the AI Shield when you need to screen signups, messages, requests, files, or indicators such as emails, domains and IPs for abuse, fraud, attacks or policy violations before they reach your application logic.

All of them share a common authentication scheme and base URL.

Authentication#

Every request requires a Bearer token in the Authorization header:

curl https://api.inferenceport.ai/gen/chat/completions \
  -H "Authorization: Bearer YOUR_TOKEN"

Tokens can be either:

A Supabase access token (from the signed-in user’s session).
A Lightning API key (created from the web console at https://console.inferenceport.ai or from the app).

How they work together#

A single account can use all three APIs simultaneously:

Subscription access controls the generation/chat experience and the daily plan limits shown in the app.
P2G credits are separate and are spent from the wallet when you call the /v1 routes.
Shield usage counts against plan-based daily analysis limits.
The hosted console at https://console.inferenceport.ai is where you can review balances, inspect usage, buy credit packs, and create API keys.

Create and manage API keys#

Sign in at https://console.inferenceport.ai.
Open the API key section in the console dashboard.
Enter a label for the key.
Optionally add an ISO-8601 expiration timestamp.
Copy the key immediately after creation. The raw secret is only shown once.

API keys are long-lived bearer tokens prefixed with ipa_live_. They are stored hashed in the database and can be revoked from the console at any time.

Common status codes#

Code	Meaning
`400`	Invalid request payload or missing required fields
`401`	Invalid, expired, or revoked bearer credential
`402`	Insufficient credits (P2G API only; realtime sessions close with `4402`)
`404`	Resource not found
`413`	Prompt exceeds configured size limits
`429`	Plan limit or spam-limit exhaustion
`500`	Upstream provider or server configuration failure
`502`	Upstream generation service failure

API Docs Contents