NIYAM AI Developer Documentation
Welcome to the official developer documentation for NIYAM AI. NIYAM provides deterministic, low-latency language model APIs designed for enterprise applications, compliance verification, and qualitative research synthesis.
Unlike probabilistic chatbots, every NIYAM model runs through a formally verified inference pipeline: identical inputs produce identical outputs, every response carries a proof trace, and no customer data is ever retained. This makes NIYAM suitable for regulated industries — legal, medical, financial — where auditability is not optional.
What you can build
Deterministic Chat & Reasoning
Zero-hallucination completions with proof traces via 70b-sovereign. Ideal for contracts, filings, and compliance Q&A.
Multimodal Understanding
Vision, video, and audio analysis with omni-vision — from 4K focus-group footage to scanned documents.
Offline & Edge Inference
Run edge-3b fully offline in the browser via WebAssembly. No data ever leaves the device.
Research Synthesis
Synthesize interviews, diaries, and panels into codebooks and sentiment maps with reproducible pipelines.
API environments
| Environment | Base URL | Use for |
|---|---|---|
| production | https://api.niyam.ai/v3 | Live workloads, air-gapped Swiss clusters |
| sandbox | https://sandbox.niyam.ai/v3 | Free testing, rate-limited, non-retained |
Start here
- Quickstart — make your first API call in under five minutes.
- Authentication — keys, vaults, rotation, and headers.
- Model Overview — pick the right engine for your workload.
- Chat Completions — full endpoint reference with examples.
- Errors & Rate Limits — handle failures and back off correctly.
- SDKs — Python and Node.js client libraries.
- Glossary — platform and model terms.
- Changelog — what shipped recently.