GateFlowRouting-Native Evaluation

Why GateFlow? ​

Most evaluation tools are disconnected from production. You run evals in notebooks, see results in dashboards, but nothing changes automatically. GateFlow closes the loop.

GateFlow enables quality-driven AI infrastructure:

• Gateway-Native Eval: Evaluation built into routing, not bolted on after
• Production-Driven: Continuous sampling of live traffic, not just test sets
• Curated Database: 10+ pre-built suites, not a blank canvas
• Automatic Compliance: EU AI Act and ISO 42001 reports generated from eval history

Quick Example ​

Evals run automatically on production traffic:

from openai import OpenAI

client = OpenAI(
    base_url="https://api.gateflow.ai/v1",
    api_key="gf-..."  # Your GateFlow API key
)

# Standard inference - evals sample automatically
response = client.chat.completions.create(
    model="auto",  # Routing informed by eval scores
    messages=[{"role": "user", "content": "Hello!"}]
)

# Or run explicit eval suites
from gateflow import EvalClient
eval_client = EvalClient(api_key="gf-...")

results = eval_client.run_suite(
    suite="safety-core",
    model="gpt-4o"
)
print(f"Safety score: {results.aggregate_score}%")

That's it. Your requests flow through GateFlow with automatic evaluation, quality-driven routing, and compliance reporting.

What's New ​

Explore the Docs ​