
Unified Api And Gateway Proxy
Portkey AI Gateway · Portkey
AI Gateway — Portkey
Evaluates Portkey's Unified API & Gateway Proxy across 9 scenario-based test cases, each graded against an expected-behavior rubric by an LLM judge, from Corsac's AI Gateway eval coverage.
About Portkey
Portkey is an AI gateway for production LLM apps — a unified, OpenAI-compatible API across 200+ models with provider routing and fallbacks, semantic and simple caching, input/output guardrails (PII redaction, prompt-injection, content moderation), request-level observability and traces, a versioned prompt library, virtual keys with per-key budgets and rate limits, and workspace RBAC + audit logs.
Sample tests· showing 3 of 9
| # | Input | Expected behavior | Check |
|---|---|---|---|
| 01 | Team migrates an existing OpenAI-SDK app to Portkey by changing only base_url to https://api.portkey.ai/v1 and passing the Portkey key as the OpenAI bearer, but forgets x-portkey-virtual-key. | Set base_url=https://api.portkey.ai/v1 and Authorization: Bearer <portkey-key>, then pass x-portkey-virtual-key (or x-portkey-config) on every request so the gateway knows which upstream provider credentials to use. Without a virtual key or config, the gateway returns an error rather than silently … | Pass / FailAi Platformhigh |
| 02 | Native Portkey SDK call omits the x-portkey-api-key header and relies on an environment variable from a prior process; the value is unset in the deployed container. | Construct the Portkey client with an explicit api_key argument (or fail fast at startup when the env var is missing). The gateway rejects requests without x-portkey-api-key with a clear auth error — surface that error to the caller; do not retry with empty auth. | Pass / FailAi Platformcritical |
| 03 | Operator sends model='gpt-4o-mini' but the active virtual key points to Anthropic Bedrock; model id has no upstream meaning. | Either (a) align the model id with the virtual key's provider (e.g., anthropic.claude-3-5-sonnet for Bedrock-Anthropic), or (b) use a config with per-target override_params so each fallback target receives the right model id. Detect provider-mismatch errors and surface them — do not silently rewrit… | Pass / FailAi Platformhigh |
How this eval is graded
Grade against expected.ideal_behavior and expected.rubric. Per-criterion pass requires mean >= 4.0 and no criterion below 3.
Rubric criteria
- Portkey
- Ai Platform
- Unified Api And Gateway Proxy
Recommended for
Works with
Related evals
Claude API
Evaluates Anthropic's Batch API across 9 scenario-based test cases, each graded against an expected-behavior rubric by an LLM judge, from Corsac's Foundation Model & API eval coverage.
View AI PlatformClaude API
Evaluates Anthropic's Extended Thinking across 9 scenario-based test cases, each graded against an expected-behavior rubric by an LLM judge, from Corsac's Foundation Model & API eval coverage.
View AI PlatformClaude API
Evaluates Anthropic's Files API & Citations across 9 scenario-based test cases, each graded against an expected-behavior rubric by an LLM judge, from Corsac's Foundation Model & API eval coverage.
ViewFrequently asked questions
What does the Unified Api And Gateway Proxy eval for Portkey Portkey AI Gateway test?+
Evaluates Portkey's Unified API & Gateway Proxy across 9 scenario-based test cases, each graded against an expected-behavior rubric by an LLM judge, from Corsac's AI Gateway eval coverage.
How is the Unified Api And Gateway Proxy eval scored?+
The judge rubric: Grade against expected.ideal_behavior and expected.rubric. Per-criterion pass requires mean >= 4.0 and no criterion below 3.
How many test cases does this eval pack include?+
The Unified Api And Gateway Proxy pack for Portkey Portkey AI Gateway contains 9 test cases. 3 sample cases are shown free on this page; the full set runs in a Corsac workspace.
How do I run this eval?+
Sign up for Corsac, connect your model or agent endpoint, and run the Unified Api And Gateway Proxy pack as-is or after customizing thresholds. Results land in your workspace with per-case scores, and you can gate releases on the pack in CI via the REST API.
Run this eval in your workspace
Connect your data, configure thresholds, and review results with your team.