dsh-our-free-model: A No-Login, No-Key Model Provider for DeepSeek Harness

What Does It Do?

It is a DSH plugin for DeepSeek Harness that gives you frontier models such as DeepSeek V4.1 Flash and Kimi K3 with no login, no sign-up and no API key. It removes the usual account, credential and quota friction by pointing at a single public upstream gateway and refreshing the model list from that upstream. Maintained by Ebony-Vinyl under the MIT license, it mounts purely as a plugin with no kernel changes, no build step and zero dependencies, and was last updated in 2026-10.

Core Features

• Zero-configuration access: no account, key or quota request is needed; models appear in the model picker as soon as the plugin is enabled.
• Live model catalog: model set, context length and capabilities are re-fetched from upstream on every refresh with no static snapshot, and models the gateway explicitly refuses to route are dropped from the picker while their rejection reason stays visible in settings.
• Real thinking budgets: Light / Balanced / Deep map to 2,048 / 8,192 / model-maximum output token budgets, logged per call, with each model card in settings showing the actual cap for that tier.
• Local usage dashboard: token heatmap, total curves, output speed and time-to-first-token sampling, all stored on your own machine with nothing uploaded.
• OpenAI-compatible forwarding port: other local tools can call these models through a base URL and key, with configurable listen address and key rotation.
• In-app upgrade and hot reload: download, checksum, backup, atomic replace and read-back run in one click with automatic rollback on failure, and code changes take effect without restarting the app.

How to Use This Plugin?

After enabling it in DSH, open the model picker in the input box and choose from the Our Free Model group; models blocked by regional policy appear in a separate region-limited group that stays visible but isolated. The settings page lives under Settings → Our Free Model and is split into model catalog, announcement center, usage dashboard, local forwarding, plugin settings and plugin upgrade, where you can inspect availability, context windows and measured latency, change the probe interval and default output cap, or turn on local forwarding and copy the base URL. A five-page onboarding announcement appears on first launch and will not show again once confirmed. On desktop hosts, avoid symlinks or directory junctions when installing, since they trip the profile gate; use the in-app plugin manager or a real directory instead.

How to Troubleshoot This Plugin?

If the desktop app refuses to start with PROFILE_UPGRADE_REQUIRED: offline dependency migration is not yet available, the profile tree contains a symlink or directory junction outside what dsh itself generates; follow the official guidance and use the in-app plugin manager or a real directory. A duplicate loader entry id: our-free-model error means the same patch layer was registered twice, so stop adding entries to cordis.patch.yml and let the bundle reference apply it. When a model reports Model is unavailable or returns 404, upstream is not routing it; the plugin removes it from the picker and records the reason in settings, and it returns automatically once a later probe succeeds.