dsh-h-v1

Running DeepSeek Harness on a single model provider day in, day out means two things: a monthly inference bill that keeps climbing, and a workflow that breaks the moment that provider hits rate limits, ships an API change, or pulls a free tier. FreeDSH, the community distribution layer on top of DeepSeek Harness, places FreeLLMAPI, OpenRouter :free, OpenCode free/zen and other configured free or low-cost providers ahead of any paid fallback, so routine calls stay at zero or near-zero cost.

The core capabilities orbit one goal: keep every workday moving. A task-aware router reads auto, eco, and ultra profiles to pick the right model tier for each workload; when a provider times out or goes dark, the router silently shifts to the next configured option without anyone watching terminal logs. Core updates are no longer overwrite-and-pray: a new kernel is validated in a parallel instance first, and local snapshots backed by git history make rollback a single command. An integrated status panel surface provider health, routing rules, recent files, and model controls directly inside the Harness UI, replacing the old habit of hunting through config files. The interface ships in Brazilian Portuguese, English, and Chinese, and interactive one-line installers cover both Windows and Linux.

FreeDSH fits two audiences. The first is developers already running DeepSeek Harness who want to cut their monthly inference costs without being chained to any single API. The second is users with a stable setup who only want to layer on smart routing, a status panel, and safe updates without touching core configuration; the DSH bundle install registers plugins only and never touches keys, providers, or existing settings. Licensed under MIT, the project explicitly respects every provider's terms of service and falls back to paid options when free tiers run out, putting reliability first.