@atlasburn/sdk
v1.5.0
Published
Institutional Forensic SDK for AI Governance and Burn Monitoring. Implements the 4 Laws of SDK Safety with Universal Auto-Detection and Framework normalization.
Maintainers
Readme
@atlasburn/sdk
Institutional Forensic SDK for AI Governance and Burn Monitoring.
Features
- Universal Auto-Detection: Intercepts AI calls from Genkit, LangChain, OpenAI, and Anthropic.
- Streaming Support: Automatically captures telemetry from streaming (SSE) responses.
- 3-Tier Guardrails: Implements Alert, Soft Signal, and Hard Fallback protocols.
- Forensic Ingestion: Deterministic telemetry flush for cost and quality auditing.
Installation
npm install @atlasburn/sdkQuick Start
Add this to the very top of your application entry file:
import { initAtlasBurnAuto } from "@atlasburn/sdk";
initAtlasBurnAuto({
apiKey: process.env.ATLASBURN_KEY
});🚀 Troubleshooting NPM 404
If you see a 404 Not Found when publishing @atlasburn/sdk:
- Scope Requirement: Scoped packages require the organization to exist on NPM.
- Action: Go to npmjs.com/org/create and create the
@atlasburnorganization. - Login: Ensure you are logged in via
npm loginwith an account that has access to that organization.
The 3-Tier Guardrail Model
- Tier 1 (Alert): Observe and log breaches without affecting requests.
- Tier 2 (Signal): Attach budget signals to the LLM response.
- Tier 3 (Fallback): Returns a graceful fallback message instead of hitting the LLM.
