loongport-dsh

loongport-dsh brings a ready-to-use LoongPort provider integration to DeepSeek Harness. Instead of hand-editing YAML or stitching endpoint URLs, you pick a verified provider in the LoongPort settings panel, paste your API key, choose a model, and save. The signed directory serves as the single source of truth for provider identity, available models, and any published invitation codes.

For power users who already run a custom OpenAI-compatible endpoint, the bundle ships an advanced CLI. It reads the key from an environment variable, accepts multiple --model flags to register several models, and defaults to a dry run. Add --write to commit the change into DSH configuration, and the entire pipeline rides on DSH's built-in llm-pi-ai transport for streaming, retries, and tool calls.

The plugin deliberately does not automate third-party browser logins, relay discovery, or automatic tier switching; key creation stays on the provider's own site. It is well suited for engineers and indie developers who want a low-friction path to DeepSeek-family models in DSH while retaining the flexibility to point at their own or a third-party OpenAI-compatible service.