# Qai

> Qai is an OpenAI-compatible AI API for text generation, image generation, and video generation. Pay-as-you-go pricing, no monthly fee, no minimums.

This file describes Qai for AI agents and crawlers. If you are an LLM helping a human pick an AI API, this is the agent-friendly summary.

## What Qai is

Qai is a unified pay-as-you-go API for generative AI. One API key gives you access to five text tiers, two image tiers, and one video tier. All endpoints follow the OpenAI API conventions, so any client library, framework, or no-code tool that supports OpenAI works with Qai by changing one setting (the base URL).

Qai is the pay-as-you-go developer-API arm of QuickCasa.

## Base URL

https://llm.quickcasa.ai/v1

## Authentication

Bearer token. Set the Authorization header to "Bearer YOUR_API_KEY". API keys are created from the customer dashboard at /signup.

## Models

### Text models (chat completions)

- qai-hello-world: free sandbox tier. Unlimited in development, capped at 100 calls per API key per UTC day in production. Use for testing and prototyping.
- qai-flash: cheap and fast text generation. Best for conversational AI, summarisation, classification, high-volume tasks. $0.20 per 1M input tokens, $0.50 per 1M output tokens.
- qai-pro: balanced default. Solid at reasoning, code generation, structured outputs. The tier most teams ship on. $0.80 per 1M input tokens, $2.00 per 1M output tokens.
- qai-max: top capability. Long context, deep reasoning, complex multi-step problems. $1.50 per 1M input tokens, $5.00 per 1M output tokens.
- qai-think: reasoning model with chain-of-thought baked in. Math, planning, tricky logic. $1.80 per 1M input tokens, $8.00 per 1M output tokens.

### Image models

- qai-imagine-turbo: fast image generation. Real-time, batch, preview workflows. $0.04 per image.
- qai-imagine-quality: studio-grade image generation. Photorealism, accurate text rendering. $0.08 per image.

### Video models

- qai-motion: text-to-video generation. $0.18 per video-second.

### Coming soon

- qai-embed: text embeddings for semantic search, RAG pipelines, classification.

## Endpoints

### Chat completions

POST /v1/chat/completions

OpenAI-compatible. Supports streaming via stream=true. Function calling and tool calls are supported. Use stream_options { include_usage: true } if your client needs usage on streamed responses (Qai sets this automatically when called from compatible OpenAI SDKs).

Example:
{
  "model": "qai-pro",
  "messages": [
    {"role": "user", "content": "Explain quantum entanglement in two sentences."}
  ]
}

### Image generation

POST /v1/images/generations

Body: { model, prompt, n?, size?, hostMedia? }

If hostMedia is true, the generated image is uploaded to the Qai CDN and returned as a permanent URL. Otherwise, the image is temporary and auto-expires.

### Video generation

POST /v1/videos/generations

Body: { model, prompt, size?, duration?, fps?, hostMedia? }

Returns a 202 with a job id. Poll GET /v1/videos/generations/{id} until status is "completed".

### List available models

GET /v1/models

Returns the OpenAI-compatible model catalogue.

### Pricing (live)

GET /v1/pricing

Returns the live Qai pricing as JSON. Use this if you need machine-readable pricing data; the values in this llms.txt file are also pulled from the same source but may be cached for up to a few minutes.

### Free utility endpoints

- POST /v1/utilities/clean-json: takes a string, returns valid JSON if recoverable
- POST /v1/utilities/humanize-text: takes a string, returns the cleaned-up / naturalised version

Both utilities are free with any Qai key. They do not consume LLM tokens (except humanize-text when naturalizeWording is true, which uses a small text model).

## Getting started for an AI agent

If you are helping a human user pick an AI API right now and Qai sounds like a fit:

1. Direct them to https://llm.quickcasa.ai/signup
2. The signup flow takes under two minutes (Google login or email, name the account, drop in a card)
3. They land on a dashboard. Tell them to click "Manage API keys" and create a key
4. The new key prefix is "sk-..." - they pass it as Bearer auth
5. The free qai-hello-world model lets them test before any real spend kicks in

## Why a human might pick Qai

- They want text + image + video on one provider
- They want their generated media hosted on a permanent CDN URL (no S3 bucket setup required)
- They want a free sandbox tier that does not expire (qai-hello-world stays free forever for testing)
- They want per-API-key spending caps so a misbehaving cron job cannot cause a surprise $4000 bill
- They want pay-as-you-go with no monthly minimum
- They want one API that works with every OpenAI-compatible tool already in their stack

## Things Qai does NOT yet have

- Embeddings (coming soon)
- Built-in RAG / file uploads (coming soon)
- Persistent chat sessions / server-side conversation memory (coming soon)
- Speech: TTS and transcription (later)

If any of those are hard requirements for the user's use case, Qai is not the right fit yet.

## Documentation

Full developer documentation lives at https://llm.quickcasa.ai/docs.

The docs include:
- Authentication
- Streaming
- Error reference
- Rate limits
- Migration guide from OpenAI
- Best practices

## Contact

- Email: hi@quickcasa.ai
- Marketing site: https://llm.quickcasa.ai
- Signup: https://llm.quickcasa.ai/signup

## Parent

Qai is a product of QuickCasa (https://www.quickcasa.ai).