🐛 Bug: Fix the bug where FastAPI internally used different types of streaming responses.

🐛 Bug: Fix the bug where FastAPI internally used different types of streaming responses.

🐛 Bug: Fix bug in total time calculation

Refactor: Add support for retrieving all models when the model name is "*"

fix: Generate a random API key using only alphanumeric characters

🐛 Bug: Fix the bug where the request to generate images cannot be parsed.

🐛 Bug: Fix a bug that may occur in a concurrent environment when multiple requests attempt to use the same database session simultaneously.

🐛 Bug: 1. Fix the bug that causes an error when Claude uploads a PNG image.

✨ Feature: Add support for v1/moderations endpoint.

✨ Feature: Add support for v1/audio/transcriptions endpoint

🐛 Bug: Fix the bug that prevents the host machine from synchronizing database files.

🐛 Bug: Fix the bug where the database cannot be automatically created when it does not exist

🐛 Bug: Fix the bug where the database cannot be automatically created when it does not exist

✨ Feature: Add database, count the first character time.

🤖 Models: Add support for the cohere series models

✨ Feature: Add feature: Support automatic restart after detecting changes to api.yaml, no need to manually restart the container.

🐛 Bug: Fix the bug of middleware statistics for unknown models

✨ Feature: Add feature: Support for counting model usage in the stats endpoint

🐛 Bug: Fix the bug where the next channel cannot be automatically switched when an error occurs in non-streaming output for o1-mini and o1-preview.

💻 Code: Optimize log display: When the model does not exist, display it in the log.

🐛 Bug: Fix the bug of connectivity check error at lobechat endpoint /v1/chat/completions

💻 Code: Increase the default timeout to 100 seconds to improve the success rate of o1-preview responses.

🐛 Bug: Fix the o1-mini timeout bug, increase the default timeout to 40 seconds.

🤖 Models: Add support for o1-mini o1-preview model

🐛 Bug: Fix the bug where httpx.RemoteProtocolError was not caught

🐛 Bug: Fix the bug where the client uploads the image URL instead of converting it to base64 encoding during image recognition.

🐛 Bug: Fix the bug where error codes are not accurately returned to the client.

✨ Feature: Add feature: Add support for using angle brackets to enclose strings to set the string before the slash to the channel name.

🐛 Bug: 1. Fix the bug where the API key is not found when rate limiting.

✨ Feature: Add feature: Support Cloudflare API model

✨ Feature: Add feature: Add support for rate limiting.

🐛 Bug: Fix the bug when the model list is empty.

🐛 Bug: Fix the bug where tool use is still included in the request body when tools are closed.

🐛 Bug: 1. Fix the bug of log errors being repeatedly displayed.

🐛 Bug: Fix the bug where the channel status is not sorted by success rate.

✨ Feature: Add feature: Add support for weighted load balancing.

✨ Feature: Add features: Add API channel success rate statistics, channel status records.

Fix the bug that cannot adapt to the available model name.

🐛 Bug: Fix the bug where the timeout was not converted to a float.

Add traffic middleware

Fix the bug that does not adapt to the new API for obtaining the model list.

Add feature: support OpenAI dall-e-3 image generation

Fix the bug where the provider lookup is inaccurate.

revert: try_all_providers method in ModelRequestHandler

refactor: revise round-robin and auto-retry logic in API request handling

fix: Improve round-robin provider selection and add Claude max_tokens auto-fill

Add feature: support vertex claude API using tool use functionality.

✨ Feature: Add feature: Support vertex ai API and vertex tool use invocation.

Fix the bug related to asyncio.CancelledError error

fix bug