Frequently Asked Questions (FAQ)

Q: What is One API?

A: One API is an open-source project that provides a unified interface for accessing various Large Language Models (LLMs) through a standardized OpenAI-compatible API format. It simplifies the integration of LLMs into your applications.

Q: What LLMs does One API support?

A: One API supports a wide range of LLMs, including OpenAI (ChatGPT series), Anthropic Claude, Google PaLM2/Gemini, Mistral, 字节跳动豆包, 百度文心一言, 阿里通义千问, 讯飞星火认知, 智谱 ChatGLM, 360 智脑, 腾讯混元, and many more.

Q: How do I deploy One API?

A: You can deploy One API using Docker, Docker Compose, or manually. Detailed deployment instructions are available in the project’s README.

Q: What are channels in One API?

A: Channels represent connections to different LLM providers. You add your API keys to channels, and One API uses these channels to route API requests to the appropriate provider.

Q: How do I manage API usage and costs with One API?

A: One API provides token management features that allow you to set expiration times, usage quotas, IP restrictions, and allowed model access for each token. You can also use user and channel grouping to apply different usage multipliers.

Q: What is the ‘额度’ (quota) in One API and how is it calculated?

A: ‘额度’ (quota) refers to the usage credits within One API. It’s calculated as: Group Multiplier * Model Multiplier * (Prompt Tokens + Completion Tokens * Completion Multiplier). The completion multiplier varies based on the model (e.g., 1.33 for GPT-3.5, 2 for GPT-4).

Q: Why am I getting an ‘insufficient quota’ error even though my account has enough quota?

A: Ensure that the token you’re using has sufficient quota assigned to it. Token quota is separate from the overall account quota.

Q: Why am I seeing the error “No available channels”?

A: Check the user group and channel group settings, as well as the model configurations for the channels.

Q: What does the error message “invalid character ‘<’ looking for beginning of value” mean during channel testing?

A: This indicates that the returned value is not a valid JSON format but rather an HTML page, likely because the IP address of your deployment or the proxy node is blocked by Cloudflare.

Q: What should I do if ChatGPT Next Web shows a “Failed to fetch” error?

A: When deploying, avoid setting the BASE_URL. Ensure the API address and API key are correctly entered. Check if HTTPS is enabled, as browsers may block HTTP requests from HTTPS domains.

Q: What does the error “The current group load is saturated, please try again later” mean?

A: This indicates that the upstream channel (LLM provider) is returning a 429 (Too Many Requests) error.

Q: Will my data be lost after upgrading One API?

A: If you are using MySQL, your data will not be lost. If you are using SQLite, ensure you have mounted a volume to persist the one-api.db database file.

Q: Do I need to make any database changes before upgrading?

A: Generally, no. The system automatically adjusts the database schema during initialization. Any necessary changes will be documented in the changelog with accompanying scripts.

Q: What does it mean if I see the error: “Database consistency has been compromised, please contact the administrator?”

A: This indicates that the ability table contains records with channel IDs that do not exist in the channel table. This often happens if you delete records from the channel table without cleaning up invalid channel references in the ability table. Every channel and supported model needs an ability table record.

Q: What is the relationship between UBOS and One API?

A: One API seamlessly integrates with the UBOS platform, a full-stack AI Agent development environment, UBOS focused on empowering businesses by bringing AI Agents to every department. Our platform allows you to orchestrate AI Agents, connect them to your enterprise data, and build custom AI Agents using your preferred LLM models and Multi-Agent Systems. With the combined power of One API and UBOS, you can unlock unprecedented levels of efficiency and innovation in your AI initiatives.