Gemini API relay detection Compatible protocol

Verify Gemini API relay protocol and capabilities

pangolin probes Gemini relays via OpenAI-compatible APIs for protocol, capabilities, and usage pass-through.

OpenAI-compatible path Function Calling usage Gemini 3 aware

Focuses on compatible protocol and thinking-by-default behavior — not Claude signature-grade identity.

Gemini highlight: compatible path & usage

Measures real Gemini behavior on OpenAI-compatible endpoints used by third-party relays.

Compatible pathProbes /chat/completions covering Google /v1beta/openai and common third-party /v1.
Gemini 3Accounts for default thinking token and empty-output edges on Gemini 3.
Usage fieldsChecks token field completeness, consistency, and response model stability.

Closest to how third-party Gemini relays are actually used.

Parent of /chat/completions. Google uses /v1beta/openai; third parties often use /v1.

Enter the Gemini Native /v1beta API root.

Relay base URL

Parent of /chat/completions. Google uses /v1beta/openai; third parties often use /v1.

API protocol

API key

Used only for this job; sent via Authorization: Bearer.

Target model

Prefer Gemini models your relay actually enables.

Scan depth

Checks that will run

Updates with Scan depth above. Full mode covers all applicable checks. Long-context probes need an extra opt-in.

Standard · all applicable (recommended) · 14

Basic request

Confirm that Gemini OpenAI compatible paths can be returned stably.

Model response shape

Check response.model and the choices[].message Chat Completions shape.

function call

Check the function call name, parameter object, and call structure.

Structured output

Verify whether response schema / JSON output constraints are in effect.

Protocol normativeness

Verify id, object, choices, finish_reason, usage, and cross-protocol usage fingerprints.

streaming consistency

Compare the output and end reasons of stream and non-stream for consistency.

Token usage

Check prompt / completion / total token field integrity and metering relationships in usage.

Official comparison of the same title

The same question compares input consumption and response behavior between channels and official (or baseline).

Accurate command reproduction (anti-casing)

The model is asked to reproduce a fixed mark verbatim, checking how deterministic the instructions are followed.

System conflict testing

Compare user only to explicit system+user, checking whether fixed words are overridden by hidden directives.

Minimalist Input Token Audit

Use minimalist requests to observe whether input_tokens is abnormally high and identify implicit System injection.

Extreme reasoning ability (number theory counting)

Use canned reasoning questions to check whether verifiable answers are correct.

Hide injection detection

Randomized variants cross-check for abnormal signals such as nonce leaks, fixed replies, and suspicious cache reuse.

Security rejection and policy stability

Check if the security policy is stable using denial-of-answer scenarios and rewriting/encoding/progressive variants.