🐋 dsh-think-translate

> Mirror notice — This repository mirrors the third-party plugin dsh-think-translate (original author: unclk, npm: dsh-think-translate, original repo: UncleK/dsh-think-translate, MIT). This mirror is unaffiliated with the original author and is archived; the original npm package and repo remain the source of truth.

Languages: English · 中文 · 日本語 · 한국어 · Español · Français · Deutsch · Русский

Translate the reasoning / thinking chain (chain-of-thought), task cards and answers of the DeepSeek Harness Web UI into one of 8 target languages — in real time, on the display layer only. The originals stay untouched in the transcript, and the translated text never enters the model context.

✨ Why dsh-think-translate

DeepSeek-class models often reason in Chinese — or in whatever language they happen to think in. dsh-think-translate renders the Think row, task cards and answer in your language while you watch, like subtitles for the model's thinking.

• 🕵️ Read any thinking chain — reasoning, chain-of-thought, task cards and answers translated in real time, streamed batch by batch
• 🌍 8 languages, one consistent UI — 中文 / English / 日本語 / 한국어 / Español / Français / Deutsch / Русский; the settings panel, thinking rows and task cards all follow your choice, and it persists across reloads
• 🔒 Private & offline-first — local Ollama (qwen2.5:7b / 14b or custom) is the default provider: free, unlimited, nothing leaves your machine. First local-model selection auto-downloads the model with a live progress bar and enables it when done
• 🧠 Zero context cost — pure display layer: the model still sees the original text, and translated text never consumes the context window
• ☁️ Google / Bing fallback — automatic switch when the local model is unavailable (google goes through a Node CONNECT tunnel using the system proxy, bypassing anti-bot blocks)
• 🛡️ Code-safe — file paths, commands, URLs, regexes and pure-code lines are never translated
• 🧩 Paragraph & sentence-aware chunking — long thinking chains are split on blank lines (paragraph structure preserved) and further batched by sentence, so even a small local model keeps quality
• ⏱️ Resilient — 3× backoff retries, browser-direct fallback, failed results never cached
• 🎚️ Adjustable translation timing — pre-translate everything, lazy-load historical chains (default), or translate only the expanded chain

📦 Installation

# Option 1: npm (recommended)
dsh plugin --profile web add dsh-think-translate
# then restart web

# Option 2: GitHub
dsh plugin --profile web add github:mtdx2001/dsh-think-translate

# Option 3: manual (junction + patch)
#  1. link the package into the profile's node_modules
New-Item -ItemType Junction -Path "$HOME\.dsh\profiles\node_modules\dsh-think-translate" `
-Target ""
#  2. add to "$HOME\.dsh\profiles\web\cordis.patch.yml":
# - insert:
#     - id: dsh-think-translate
#       name: dsh-think-translate
#  3. restart web

🚀 Usage

1. Open Settings → Think Translation
2. Pick the target language (e.g. 日本語) — the settings panel, thinking rows and task cards all switch to it
3. Pick the preferred provider:
   - Local model (Ollama) — on first selection a download prompt appears (qwen2.5:7b / 14b or custom); it auto-enables when finished. The "+" button next to the model picker downloads more models anytime
   - google gtx / bing — works out of the box (auto system proxy / VPN)
4. Send a message that makes the model think, then expand the Think row to read the translation and compare with the original

⚙️ How it works

browser → POST /_xlate/translate (same-origin, no CORS)
→ host provider chain (fail-open):
openai-compatible (local Ollama, Node fetch to loopback)
→ google gtx (Node https + CONNECT tunnel through system proxy)
→ bing (curl form)
→ browser-direct fallback

• Host half (lib/index.js): provider adapters, LRU cache (600), /_xlate/models listing, /_xlate/model/pull + pull-status model download management (auto-configures on completion)
• Client half (lib/client.js): 8-language UI, sentence/paragraph-batched translation, streaming Think rows, localStorage persistence (settings + translation cache)
• Pure display layer: originals remain in the transcript and model context

🛠 Development

• No build step: lib/client.js is the browser bundle (source = artifact), lib/index.js is the host ESM
• Client changes apply on page refresh; host changes need a web restart
• The 8-language strings live in the UI_TEXT dictionary in lib/client.js

📄 License

MIT

Advanced: pair with think-zh (75k curated TM)

Chinese users on local models: think-zh ships a
self-hosted observer with a 75,353-sentence curated translation memory (DeepSeek-refined),
a built-in OpenAI-compatible endpoint and a live observer page at 127.0.0.1:18765.
Point this plugin's provider at it, or follow its INSTALL.md for the full pipeline
(1.8B realtime + 7B idle correction, MIT).