🐛 Bug: Fix the bug of undefined review content.

💻 Code: Optimize log display: add error provider name display.

✨ Feature: Add feature: Support that as long as the prefix of the API key exists in the configuration file, the API key is valid.

🐛 Bug: The bug where the provider's name, which is purely numeric, was not converted to a string has been fixed.

🐛 Bug: Fix the bug where the response body error returned by the error report incorrectly uses raise.

🐛 Bug: Fix the issue where an error is returned directly when Authorization is empty.

✨ Feature: Add feature: support setting rate limit for each model individually

💻 Code: Add log display for model not found.

💻 Code: Optimize to reduce unnecessary log output.

💻 Code: Modify the default behavior of API key cooldown, change it to default off for API key cooldown. When there is only one API key in the channel, it will not cooldown under any circumstances. When API_KEY_COOLDOWN_PERIOD is 0, it will not cooldown. After enabling cooldown, any error with the API key will trigger a cooldown.

🐛 Bug: Fix the bug of incorrect function parameters entering the cooldown state.

🐛 Bug: 1. Fix the bug where the closed client cannot be found when closing the request client.

📖 Docs: Update documentation

💻 Code: When there is only one optional provider, cooling will not take effect.

🐛 Bug: Fixed the bug with limited concurrency and removed unnecessary asynchronous mutex locks.

🐛 Bug: Fix the bug where Vercel fails to start successfully.

🐛 Bug: Fix the bug where Vercel fails to start successfully.

🐛 Bug: Fix the bug of incorrect weight allocation

✨ Feature: Add feature: Support model-level cooling. When a model under a channel reports an error, it does not affect other models under the same channel, only the model that reported the error is cooled.

🐛 Bug: Fix the bug in the chart display.

🐛 Bug: Fixed the bug where the cooling model could not take effect in this request.

✨ Feature: Add feature: Support channel cooldown. When an API channel response fails, the channel will automatically be excluded and cooled down for a period of time, during which no requests will be made to that channel. After the cooldown period ends, the model will automatically be restored until it fails again, triggering another cooldown.

🪞 Frontend: Fixed the bug where the chart size was displayed incorrectly when there was no data.

🐛 Bug: Fix the bug in the moral check database update error.

🪞 Frontend: Add request model statistics bar chart

💰 Sponsors: Thanks to @PowerHunter for the CNY 200 sponsorship, sponsorship information has been added to the README.

🐛 Bug: Fix the bug where the model cannot be found.

🐛 Bug: Fix the bug where weight polling cannot match the model.

🐛 Bug: Fix the bug where weight polling did not check if the weight channel conforms to the request model.

📖 Docs: Add /docs/markdown endpoint to output markdown API documentation

📖 Docs: Update documentation

🐛 Bug: Fix the bug that did not check if weights were None.

🪞 Frontend: Added sidebar support to the front end, removed the menu bar.

🐛 Bug: Fix the bug of weight load balancing round-robin.

📖 Docs: Update documentation

✨ Feature: Add support for embeddings model

✨ Feature: Add feature: support wildcard matching like gpt* to match models such as gpt-3.5 and gpt-4.

🐛 Bug: Fix the bug where the model is not persisted to the file after being automatically retrieved.

🐛 Bug: Fix the bug where Gemini cannot use non-streaming output.

📖 Docs: Add documentation for serv00 deployment steps.

🐛 Bug: Fix the bug where vercel cannot set app.state.config.

✨ Feature: Add support for completely disabling the database.

✨ Feature: Added lottery scheduling algorithm and support for random scheduling algorithm.

🐛 Bug: Fix the bug where the front-end sheet table can only display one key when there are multiple API keys.

🐛 Bug: Fix the bug where the sheet page cannot fully display when there are too many models.

🐛 Bug: Fix the bug where after filtering through the filter, if there are a total of four rows, when clicking the edit configuration on the fourth row, the sheet still displays the data from the fourth row before the filter was applied.

🐛 Bug: Fix the bug where the sheet cannot be closed when clicking the cancel button without filling in the content on the frontend add model sheet.

✨ Feature: Add support for frontend page operation configuration files.

✨ Feature: 1. Add support for experimental frontend.

🐛 Bug: 1. Fix the bug where the option request attempts to parse the OK stream message.