Releases·Tool Release
cv-superding releases unofficial DSH plugin using chat.deepseek.com web login as an LLM provider
Two community plugins for DeepSeek's DSH harness appeared on GitHub: dsh-deepseek-web-login turns a logged-in chat.deepseek.com browser session into an LLM provider, and dsh-prompt-optimizer rewrites user prompts into structured commands for the working AI before sending. Both are unofficial, fully open-source, and target DSH agent workflows without paid API keys for the web-login path.
78 engagement2 sourcesThu, Sep 10, 2026, 01:10 PM
AI Overview
- Unofficial DSH plugin drives DSH agents via chat.deepseek.com's logged-in web session — not the API — covering PoW solving, SSE streaming, and file upload
- Plugin uses browser login capture, SHA3 WASM-based PoW challenge solving, and per-call temporary chat sessions
- Tool calls run through prompting-based approaches; supports agent tool calls, thinking streams, and image understanding
- Requires Node >=22, Apache-2.0 licensed, tested with 111 assertions, flagged unofficial and use-at-your-own-risk
- Second DSH plugin, dsh-prompt-optimizer (v0.1.9beta1) by WestFox, rewrites prompts into commands via a relay AI before sending to DeepSeek-V4.1-Flash
Entities
Related
Releases·Tool Release
Unofficial DeepSeek Harness plugin brings live model catalog to Command Code
A developer released dsh-commandcode-provider, an unofficial DeepSeek Harness LLM provider plugin for Command Code, ported from pi-commandcode-provider under MIT. It adds a live model catalog, reasoning-effort support, and a Models-page card to the editor.
A developer released dsh-commandcode-provider, an unofficial DeepSeek Harness LLM provider plugin for Command Code, ported from pi-commandcode-provider under MIT. It adds a live model catalog, reasoning-effort support, and a Models-page card to the editor.
49 engagement1 sourcegithubFri, Aug 14, 2026, 11:38 AM
Releases·Tool Release
FreeDeepseekAPI-EN: Atharvotech maintains OpenAI-compatible local proxy for DeepSeek Web Chat
Atharvotech maintains FreeDeepseekAPI-EN, an English-adapted version of a local OpenAI-compatible API proxy for DeepSeek Web Chat. The proxy lets users connect DeepSeek to Open WebUI, Claude Code, and OpenAI SDK tools without an official API. Core logic belongs to original author ForgetMeAI.
Atharvotech maintains FreeDeepseekAPI-EN, an English-adapted version of a local OpenAI-compatible API proxy for DeepSeek Web Chat. The proxy lets users connect DeepSeek to Open WebUI, Claude Code, and OpenAI SDK tools without an official API. Core logic belongs to original author ForgetMeAI.
25 engagement1 sourcegithubSun, Sep 6, 2026, 05:07 AM
Releases·Tool Release
dsh plugin offers free, no-key access to frontier models Muse Spark 1.3 and MiMo V2.6 with no usage cap
A GitHub plugin called dsh-our-free-model lets dsh users access frontier models including Muse Spark 1.3 and MiMo V2.6 for free with no usage cap, login, sign-up, or API key. The plugin mounts on dsh without modifying the core, tracks upstream model availability in real time, and implements real thinking-budget controls and a built-in update system.
A GitHub plugin called dsh-our-free-model lets dsh users access frontier models including Muse Spark 1.3 and MiMo V2.6 for free with no usage cap, login, sign-up, or API key. The plugin mounts on dsh without modifying the core, tracks upstream model availability in real time, and implements real thinking-budget controls and a built-in update system.
146 engagement1 sourcegithubThu, Sep 24, 2026, 04:17 PM
Releases·Tool Release
DeepSeek Harness plugin lets ChatGPT, Claude, Grok subscriptions serve as LLM providers
A new GitHub plugin for DeepSeek Harness enables using ChatGPT (Codex), Claude, and Grok (X Premium) subscriptions as LLM providers without API keys. Users log in via OAuth in the web UI, with tokens stored locally and auto-refreshed.
A new GitHub plugin for DeepSeek Harness enables using ChatGPT (Codex), Claude, and Grok (X Premium) subscriptions as LLM providers without API keys. Users log in via OAuth in the web UI, with tokens stored locally and auto-refreshed.
75 engagement1 sourcegithubFri, Aug 14, 2026, 08:27 AM
Releases·Model Release
DeepSeek-V4.1-Flash tech report details post-training recipe: synthesized data, agent environments over algorithmic novelty
DeepSeek published a tech report for DeepSeek-V4.1-Flash, disclosing its post-training methodology: supervised fine-tuning, reinforcement learning, and on-policy distillation. The report emphasizes investment in large-scale automated data synthesis and environment construction over novel optimization methods, reinforcing DeepSeek's reputation as the most open lab on methods disclosure.
DeepSeek published a tech report for DeepSeek-V4.1-Flash, disclosing its post-training methodology: supervised fine-tuning, reinforcement learning, and on-policy distillation. The report emphasizes investment in large-scale automated data synthesis and environment construction over novel optimization methods, reinforcing DeepSeek's reputation as the most open lab on methods disclosure.
2 engagement1 sourceredditThu, Sep 10, 2026, 05:13 PM