CloudRouter 文档

English

简体中文

English

简体中文

Appearance

Call Guide & Dual Routing ​

Applies to: Developers, Organizations (top-level users), Subusers Updated: 2026-07-02

The new console provides two call names for the same model — one with a fingerprint and one without — and displays them side by side with one-click copy on the "Call Guide" page. This guide explains how each call name is used, which route it takes, which to use in which scenario, and where to get them.

1. The two call names ​

On the console's "Call Guide" page, you'll see every group reachable by your current account; under each one, every model provides two copyable call names side by side:

FormExampleRouting behavior
With fingerprint5MHXZWKA/gpt-4oRoutes to that group by "calling it out" via the fingerprint
Without fingerprint (bare model name)gpt-4oRoutes to the routing group bound to the Key used for the call

5MHXZWKA is the group's fingerprint, gpt-4o is the bare model name, joined by /. The copy button for the fingerprint form only appears when the group has fingerprints enabled.

What is a fingerprint? A fingerprint is a short character identifier (e.g. 5MHXZWKA) that the platform automatically assigns to each routing group; it stays unchanged after creation, and you don't need to create or manage it manually. You can see and copy it on the group card on the console's "Call Guide" page. The fingerprint lets you precisely specify which group to route to during a call, using the <fingerprint>/<model name> format.

2. How the gateway routes ​

When the gateway receives a request, it first checks whether the model field contains a fingerprint prefix (contains /):

  • With fingerprint (e.g. 5MHXZWKA/gpt-4o): parses the fingerprint and routes to the group corresponding to that fingerprint

  • Without fingerprint (e.g. gpt-4o): routes to the routing group bound to this Key

  • With fingerprint: regardless of which group this Key is bound to, it "calls out" the specified group by fingerprint — flexible.

  • Without fingerprint: routes to the routing group bound to this Key, consistent with the legacy console's "one group per Key" behavior — zero changes for legacy clients.

⚠️ When calling without a fingerprint, if this Key is not bound to any routing group, the platform returns 403 directly (bare name not callable). In this case, go to the "Call Guide" and switch to a fingerprint call name, or first bind a routing group to this Key on the "API Keys" page.

When the selected channel is unavailable, the platform returns an error directly; it currently does not auto-switch to another channel and does not do fallback. Handle retries on the client side, see Errors & Troubleshooting.

3. Which to use in which scenario ​

ScenarioRecommendation
Legacy scripts/clients already runningUse the bare model name, keep going with zero changes
One Key wants to temporarily switch to a different groupUse the fingerprint call name to call out the target group
One user with multiple Keys, each bound to a different groupUse the bare model name; each Key goes to its own bound group
Unsure how to write the call nameJust copy directly from the "Call Guide" — safest

4. Where to get the call name ​

Log in to the console and click "Call Guide" in the left navigation bar to enter the page.

Page structure ​

From top to bottom, the Call Guide page consists of:

  • Top info bar: briefly introduces the format rules of the call name (<fingerprint>/<bare model name>)
  • Search box: search by group name, platform, fingerprint, or model name to quickly locate your target
  • Platform tab bar: tabs by model provider (e.g. Anthropic, OpenAI, Google, etc.); click to switch and view only that provider's groups; when searching, the tabs auto-hide and results are shown across all platforms instead
  • Exclusive / Public filter: distinguishes groups exclusive to your organization from groups public to the platform

Group card ​

Each available group is displayed as a card, containing:

  • Group name: at the top of the card, e.g. "Claude Official Direct" or "GPT Value"
  • Billing type label: marks whether the group is "usage-based (deduct balance)" or "subscription-based"
  • Fingerprint prefix: click to copy the entire fingerprint prefix (e.g. 5MHXZWKA/), in the format <fingerprint>/<model name>
  • Available model list: all callable models under the group, each with two copy buttons side by side —
    • "Bare name" button: copies the bare model name (e.g. claude-sonnet-4-6); calls route to the routing group bound to your Key
    • "Fingerprint" button: copies the call name with a fingerprint (e.g. 5MHXZWKA/claude-sonnet-4-6); calls route fixedly to this group. This button appears only when the group has fingerprints enabled

Steps ​

  1. Enter the "Call Guide" page
  2. Find the target model via the platform tabs or the search box
  3. On the corresponding group card, click the "Bare name" or "Fingerprint" copy button to the right of the model
  4. Paste the copied call name into the model field in your code

Which groups you can call depends on the range of models available to your account (configured per organization by the platform). If you can't see the target model, see Admin Manual · Regions & Model Scope.

5. Call examples ​

Both call names use the standard OpenAI-compatible format; only the model field is written differently.

Without fingerprint (bare model name):

bash
curl https://api.cloudrouter.online/v1/chat/completions \
  -H "Authorization: Bearer sk-xxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

With fingerprint (call out a group):

bash
curl https://api.cloudrouter.online/v1/chat/completions \
  -H "Authorization: Bearer sk-xxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "5MHXZWKA/gpt-4o",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

For general rules on endpoints, authentication, streaming, etc., see API Basics. The API address is https://api.cloudrouter.online, with the call prefix /v1.

6. FAQ ​

Q: Do I have to switch to the fingerprint call name? A: No. The bare model name remains valid and routes to the routing group bound to this Key. The fingerprint is just an additional, flexible way to "call out a group".

Q: Is there a billing difference between with and without a fingerprint? A: Billing is priced by the group/channel actually hit (as shown on the "Model Marketplace" or the console's "Purchase" page), independent of how the call name is written — it depends on which group is ultimately reached.

Q: Can I mix bare names and fingerprints on the same Key? A: Yes. The gateway determines routing per request based on whether that request's model carries a fingerprint: bare name goes to the Key's bound group, fingerprint calls out the corresponding group.

Q: What should I do when a fingerprint call reports "group unreachable"? A: It means the group corresponding to that fingerprint is not within your account's available range, or that channel is currently unavailable. Confirm your available range, or contact CloudRouter — see Errors & Troubleshooting.

Q: Does the Call Guide provide both call names at once? A: Yes. On the "Call Guide" page, every model displays both the "Bare name" and "With fingerprint" copyable Model IDs side by side (the fingerprint button appears when the group has fingerprints enabled); copy whichever you need.

Contact us ​