dsh-llm-proxy

Accessing international large language model APIs from within China often results in connection timeouts or unreachability, and free API quotas are frequently disrupted by transient errors like HTTP 429 rate limiting or 5xx server errors. The dsh-llm-proxy plugin addresses these issues with an intelligent proxy solution. Its core capability is routing requests based on the target domain of each model: users select which models should go through the proxy, and the plugin automatically forwards those requests via a specified proxy (e.g., Clash), while others connect directly, avoiding the impact of a global proxy on domestic APIs. It also includes automatic retry on failure for disconnections, rate limiting, and server errors to preserve free quotas, along with multimodal model mirroring and a test connection feature for easy verification. This plugin is particularly useful for developers needing to access international model APIs, especially those relying on free quotas but hindered by network restrictions, as well as users wanting fine-grained control over proxy settings to optimize network performance and costs.