# R3AL.AI

> Universal vision ONNX quantization: any .onnx model in, quantized .onnx out.

## Docs

- [Introduction](https://docs.r3al.ai/introduction.md): Quantize any vision model on the R3AL platform, driven from a thin Python SDK.
- [Quickstart](https://docs.r3al.ai/quickstart.md): Quantize your first vision model on the R3AL platform in six steps.
- [Installation](https://docs.r3al.ai/installation.md): Install the thin client and connect it to the R3AL platform.
- [Quantization methods](https://docs.r3al.ai/concepts/methods.md): PTQ methods and when to use QAT instead.
- [QAT explained](https://docs.r3al.ai/concepts/qat.md): Quantization-aware training: wbit, abit, epochs, and when to use it.
- [Integer arithmetic](https://docs.r3al.ai/concepts/integer-arithmetic.md): ptq_static and ptq_dynamic: real compression and speed.
- [Calibration](https://docs.r3al.ai/concepts/calibration.md): When and why calibration images are needed.
- [Output & manifest](https://docs.r3al.ai/concepts/manifest.md): Every job produces a quantized ONNX plus a verifiable manifest.
- [Using the platform](https://docs.r3al.ai/guides/platform.md): Accounts, API keys, monitoring jobs, and your plan on platform.r3al.ai.
- [End-to-end workflow](https://docs.r3al.ai/guides/workflow.md): From framework weights to a validated quantized ONNX deliverable.
- [Benchmarking](https://docs.r3al.ai/guides/benchmarking.md): Measure latency and size of the quantized model locally, on your own hardware.
- [QAT pipeline](https://docs.r3al.ai/guides/qat-pipeline.md): Quantization-aware training for any vision model, via the R3AL platform.
- [Python SDK](https://docs.r3al.ai/sdk/quantizer.md): Quantizer, QuantConfig, PlatformClient, Job, and authentication for r3alai 2.0.
- [CLI](https://docs.r3al.ai/cli/overview.md): Quantize vision models from the terminal with the r3al command.
- [Command reference](https://docs.r3al.ai/cli/commands.md): Every r3al command and flag.
- [Coding agents](https://docs.r3al.ai/cli/agents.md): Use R3AL from Claude Code and other coding agents.
- [API overview](https://docs.r3al.ai/api/overview.md): The R3AL platform HTTP API: uploads, quantization jobs, artifacts, keys, and usage.
- [Authentication & roles](https://docs.r3al.ai/api/authentication.md): API keys, session tokens, the master key, and what Admin/Developer/Member can each do.
- [Rate limiting](https://docs.r3al.ai/api/rate-limiting.md): Per-caller sliding-window limits on every /v1/* call, with a tighter tier on login/register.
- [Jobs: upload, quantize, qat, download](https://docs.r3al.ai/api/jobs.md): Upload a model, submit a job, poll it, and download the deliverable.
- [API keys](https://docs.r3al.ai/api/keys.md): Mint, list, and revoke standalone API keys for scripts and integrations.
- [Team: accounts, invites, roles](https://docs.r3al.ai/api/team.md): Register/login, invite teammates with a role, and manage the roster.
- [Monitoring: alerts, deployments, logs](https://docs.r3al.ai/api/monitoring.md): Operator monitoring: alert rules over job history, model-serving deployments, and a log tail.
- [Errors](https://docs.r3al.ai/api/errors.md): One consistent error envelope, whether it's an SDK failure, a bad request, or a rate limit.