Developers

Start with lightweight instrumentation. Then add eval gates and runtime controls as coverage expands.

Request a demo Product

1) Instrument a route

Capture traces for prompts, retrieval, tool calls, and outputs with consistent metadata (env, route, user, version).

2) Establish a baseline eval suite

Create a small, versioned dataset and a scorecard for quality and safety. Run it in CI.

3) Add runtime controls

Enforce budgets, policies, and fallbacks so production behavior is predictable under load and drift.

Instrumentation (illustrative)

TypeScript

// Pseudocode for instrumentation
import { Infrand } from '@infrand/sdk';

const infrand = new Infrand({
  apiKey: process.env.INFRAND_API_KEY,
  environment: process.env.NODE_ENV,
});

export async function handler(req) {
  return await infrand.trace('support.reply', async (span) => {
    span.setTag('user.id', req.userId);
    span.setTag('release', process.env.RELEASE_SHA);

    const context = await span.step('retrieval', () => retrieve(req.query));
    const result = await span.step('model', () => callModel({ context }));

    return result;
  });
}

This is illustrative pseudocode to explain the workflow. Request a demo for an implementation aligned to your stack.

Eval gate (illustrative)

YAML

# Example CI gate (pseudocode)
name: AI Evals
on: [pull_request]
jobs:
  evals:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - run: npm ci
      - run: npx infrand eval run --suite baseline --threshold 0.90

This is illustrative pseudocode to explain the workflow. Request a demo for an implementation aligned to your stack.

Policy example (illustrative)

JSON

{
  "route": "support.reply",
  "budgets": { "maxTokens": 1200, "maxCostUsd": 0.02 },
  "policies": [
    { "type": "pii_redaction", "mode": "strict" },
    { "type": "citation_required", "minCitations": 1 }
  ],
  "fallback": { "onTimeoutMs": 8000, "model": "smaller-fast-model" }
}

This is illustrative pseudocode to explain the workflow. Request a demo for an implementation aligned to your stack.

Operational guidance

Design for observability and control from day one

The fastest teams treat AI behavior like distributed systems: instrumented, gated, and governed.

•Tag every interaction with environment, route, user, and release version
•Run eval scorecards in CI before routing traffic to a change
•Use budgets, timeouts, and fallbacks for predictable degradation
•Keep provenance so incidents link back to the exact deployed configuration

Request a demo