Models & Providers

Runs a four-task, 150-point model intelligence benchmark inside DeepSeek Harness (Qin 60, Pompeii 45, Pelican 22.5, Candy 22.5): each selected model × reasoning-effort pair executes every task in its own isolated subagent session, Qin and Pelican are graded by the selected judge model against rubrics, Pompeii and Candy programmatically, and batches keep scores, artifacts and cost. Adds a 0–1000 real-world performance score per model × effort computed from local session logs across seven dimensions (output speed, first-token latency, reasoning speed, tool cadence, tool latency, tool success rate, cache hit), with baseline deltas and controlled speed runs, plus a sidebar page with ranking, performance, history and settings views. Each batch spends real tokens of the candidate and judge models, and candidate subagents get file and bash tools inside an isolated run directory.

Agent gateway manager: picks a working gateway among proxies and direct, verifies the model channel, and switches online; DSH plugin plus standalone CLI.

DSH rotation pool: a sidebar panel to pick configured provider/model routes and fail over across them on quota errors, plus optional multi-key rotation.

Fusion for DeepSeek Harness: one model entry backed by a frontier Lead that reads, briefs and reviews, and a much cheaper Sidekick that writes the code.

Disables the built-in llm-deepseek-account route so the DeepSeek account model group no longer appears in the picker.

Account-based access to the ToTonAI gateway inside DeepSeek Harness: sign in or register in the settings panel, use 65 mainstream models (GPT, Claude, Gemini, DeepSeek, Doubao, Grok, GLM), and pick a plan then pay by QR code in the same panel. Upstream API keys stay on the server.

Asks which provider and model to use before each in-process subagent starts, listing the models from your configured LLM providers; if nothing is chosen within 60 seconds the subagent keeps the DSH default route.

Provider pool for DeepSeek Harness. It runs a local OpenAI-compatible proxy that tries configured LLM providers in priority order and falls back on authentication, quota, rate-limit or server errors, with per-provider usage statistics.

Configuration panel for an existing `oneapi` route of the pi-ai provider adapter (declare that route in the llm-pi-ai settings first, with api: openai-completions): edit display name, Base URL, API key and default model, test connectivity per model and sync the model list from the gateway, with six agent tools and an optional left-sidebar panel. Written for a Baidu-internal gateway: the Base URL placeholder and token link point there, so set your own OneAPI-compatible gateway.

Adds a Settings section to adjust the advertised context window of each connected WorkBuddy and Trae model, writing the override into the connector's context budgets.

Provider and account hub for DeepSeek Harness: several API-key accounts per provider with one-click switching and failover when a call dies on quota or credential errors, a per-account usage and quota panel in the conversation header, and built-in Grok Build device-code OAuth.

Replaces the composer model picker with a model + thinking-effort picker: arrow keys left/right cycle the effort level and arrow keys up/down switch model; an Auto level scores each turn by how far the conversation has drifted from the main objective and by repeated failures (one extra small judge call per turn), picks the effort from that score, and labels every finished turn with the effort it sent.

Unified provider pools for DeepSeek Harness with session-sticky balancing, capability-aware model mapping, and automatic failover.

A zero-config composer model picker with one provider per tab, independent scrolling per provider, cross-provider search with provider badges and reasoning-effort chips; selections stay in sync with the built-in /model dialog.

Native DeepSeek Harness panel for local CLIProxyAPIPlus with Qoder and Codex authentication, quota views, automatic provider and model catalog sync, YAML editing, service control, request trends, and runtime logs.

Per-model wire protocol for DeepSeek Harness (desktop profile only): compiles model-level protocol intent into legal llm-pi-ai provider routes, probes Chat Completions, Responses and Anthropic Messages with a differential control probe, and lets you hide models no protocol can serve. Mutually exclusive with dsh-gearbox.

Lists RouterAI prompt, completion and cache-read prices per million tokens for the models configured in this profile, highlighting the active one. Requires a model provider whose config key is `routerai`; without it the table shows no data.

Manages local llama.cpp servers and GGUF launch profiles on Windows and exposes the profiles in the DSH model picker.

Per-provider API key pool for DeepSeek Harness with manual switching, background reachability probing and plan-usage adapters for DeepSeek, Kimi, MiniMax, Zhipu, OpenRouter and other providers.

Stores provider API keys in one local pool and hands out the least-used enabled key over a loopback-only endpoint, tracking usage counters without proxying requests.

一个将 WorkBuddy 订阅反代到 DeepSeek Harness 使用的 DSH 插件

DSH plugin: bring your ChatGPT subscription into DeepSeek Harness — Codex models, GPT Image, quota and recovery checks, plus a per-plugin scoped HTTP(S) proxy. Community alpha fork of dsh-codex-connect, published on the alpha dist-tag; it registers the same openai-codex provider, so it cannot be installed alongside the upstream plugin; the image, quota and proxy features are opt-in and off by default.

Adds ChatGPT sign-in, Codex subscription quota (5-hour and weekly windows) and Codex-backed web search to DeepSeek Harness; installing it switches the profile's web search provider to openai-codex.

Runs the local llama.cpp Qwen3.5-4B from a Settings page — Windows only, NVIDIA GPU (8 GB VRAM handles the 4B Q4) — with start, stop and restart, launch parameters, deployment paths with a presence check, provider-route and compaction-policy sync in the profile patch, a loopback request-capture proxy, and a local_worker tool that hands translation, OCR and video-frame chores to the local model by file path. Needs llama-server.exe (verified on b12069) and Qwen3.5-4B GGUF + mmproj files on disk; video frames also need ffmpeg.