Developer FAQ
Applicable role: Developer Last updated: 2026-07-14
This document is an index of frequently asked developer questions, with each question pointing to detailed documentation.
1. Onboarding and Authentication
Q: What should I put in the Base URL?
YOUR_API_BASE_URL, see API Basics
Q: Where do I create an API Key / how do I get one? Both organizations and sub-users can create Keys on the "API Keys" page in the console. When a sub-user creates a Key, routing groups are restricted to the organization's authorized scope, see Quickstart
Q: Getting a 401 error?
Check the format of the Authorization: Bearer header, see Error Codes and Troubleshooting
2. Models and Invocation
Q: What exactly should I put in the model field?
Enter the model name (e.g., gpt-4o). You can view the list of available models in the console's "Invocation Guide" or "Model Marketplace". If your organization has enabled group-identifier routing (an advanced feature, disabled by default), the Invocation Guide page also lets you copy an invocation name with a group identifier (e.g., 5MHXZWKA/gpt-4o) to precisely specify a group, see Invocation Guide and Routing
Q: What's the difference between a model name and one with a group identifier? A model name uses the routing group bound to that Key or the default routing configuration; one with a group identifier targets a specified group (only available after the organization enables group-identifier routing), see Invocation Guide and Routing
Q: Why can't I reach a certain model? The invocation name is misspelled, or the group is not within your account's available scope (configured by the platform per organization). Confirm the available scope or copy it from the Invocation Guide.
Q: What's the difference between the OpenAI format and the Anthropic format? See API Basics, Section 5
Q: How do I switch models?
Change the model field, see API Basics
3. Billing and Rate Limiting
Q: How is billing calculated? Who gets charged? Billed according to the pricing of the group/channel actually hit (as displayed in the "Model Marketplace"); actual charges are always deducted from the owning organization's balance.
Q: Is there a billing difference between using a group identifier and a model name? It's unrelated to how the invocation name is written; billing is based on the group/channel ultimately used, see Invocation Guide and Routing
Q: What do I do about 429 rate limiting?
Reduce concurrency and use exponential backoff; distinguish rate limits (API_KEY_RATE_*, wait for the window to reset) from quotas/concurrency (the exact code depends on what the API actually returns), see Error Codes and Troubleshooting
Q: What's the concurrency limit? How do I raise it? The organization-level concurrency limit is configured uniformly by the platform. Coordinate with the owning organization, and contact the platform if adjustments are needed.
4. Stability
Q: How do I handle a 502? When a model provider returns a 5xx, the platform uniformly maps it to a 502. Please retry with exponential backoff, see Error Codes and Troubleshooting
Q: What do I do when an invocation with a group identifier reports "group unreachable"? Group-identifier routing is an advanced feature (disabled by default) and is only available after the organization enables it. If it's enabled but you still get this error, it means the group is not within the available scope or is currently unavailable; the platform returns an error directly without falling back. Switch to an available group, see Invocation Guide and Routing
Q: How do I distinguish between a 500 and a 502? A 500 is a failure on the platform itself, while a 502 is a failure on the model provider's side, see Error Codes and Troubleshooting
Q: The model occasionally outputs Korean/Japanese? Model language drift, see Claude Language Drift Notes