dsh-llm-verifier
Best-of-3/5 orchestration and LLM-as-a-Verifier selection for DeepSeek Harness
853 results
Best-of-3/5 orchestration and LLM-as-a-Verifier selection for DeepSeek Harness
Bridge the vercel-labs/skills ecosystem into dsh: LLM-driven skill search, install, and lifecycle for temp/project/global scopes
DeepSeek vision bridge for the dsh web GUI: route image attachments to a vision model (pi-ai / llama.cpp Qwen3-VL) and continue the conversation with the text description on a text-only LLM (DeepSeek)
DSH host plugin — smart outbound HTTP(S) routing: named proxies (http/https/socks4/4a/5/5h), per-host/provider/plugin rules, direct-first fallback with health memory (global fetch + node http/https).
Multi-route high availability, fault isolation and observability for DeepSeek Harness. Route groups fail over between authorized provider routes on rate limits (429 / Retry-After), transport errors and auth/quota failures, with per-route quarantine (exponential backoff) and disablement.
DeepSeek Harness plugin: a General-settings selector that chooses which language the model reasons in, via a system-prompt section.
Task planning with experience muscle-memory for DeepSeek Harness: condition-reflex recall of past solutions + LLM-driven capability matching + auto-persisted lessons. Zero keys, zero hard-coded paths.
Safe model-request retry for DeepSeek Harness with exponential backoff, model-availability recovery, and configurable unknown-error limits.
Learning assistant extension for DeepSeek Harness (dsh): DeepTutor tutoring for your agent — deep explanations, self-test questions, learning paths, personal knowledge-base search (RAG), and note archiving (HTTP/WS first, CLI fallback; auto-adapts local/remote deployment).
OpenCode Go subscription key rotator for DeepSeek Harness (dsh): per-request active-key routing, 429/quota cooldown rotation, cost accounting, dashboard quota scraping, and a web GUI settings section. Self-contained data dir under the plugin install folder — no machine-specific paths.
DSH web plugin: 为自定义模型(OpenAI-compatible pi-ai provider)增加思考强度(reasoning effort)配置 —— 设置页新增「思考强度」分区,按模型配置思考级别与线格式、按 provider 配置默认思考强度,直接写入 llm-pi-ai 设置,composer 模型选择器随即生效。
Ollama Cloud models, OpenAI-compatible chat, and Web Search/Fetch for DeepSeek Harness
Free Kilo Gateway and OpenCode Zen models for DeepSeek Harness (DSH) via native dsh-llm adapters.
Captain 控制台:面向求职公司筛查的可恢复、多智能体 DSH 工作流。
让 DeepSeek Harness 自动重试商汤 429/QUOTA 等瞬时网关故障直到任务完成——启动时给所有 llm-pi-ai provider 热注入加固版 retryPolicy,运行中新增的 provider 也会自动补上,再也不用手动发「继续」。
Claude Code CLI as an LLM provider for dsh (DeepSeek Harness): live model list, per-session resume, approval relay, images, a memory/rewind/changes panel, and workspaces on remote SSH boxes
Per-provider & per-model rate limiting (token bucket, reject/queue modes) plus gateway identity rules for DeepSeek Harness
DSH plugin: query provider quota/balance (Kimi / Zhipu GLM / DeepSeek / Volcengine / custom API) and show a live usage badge bar under the composer.
Witness — the quality gate for agent-built web apps. Independent browser acceptance testing: CLI, MCP, AI-drafted checklists, visual regression, multi-browser. Agents self-test and pass; real browsers tell the truth.
DSH tool-call compatibility: strips sandbox_permissions/justification from tool-call arguments for third-party models (GPT etc.), and lets the user skip possibly-stuck tool calls with an LLM-facing notice.
Token usage dashboard, provider balances, and subscription quotas for the dsh web GUI
深海事务所 (Abyss) — an operations console for DeepSeek Harness agent fleets: every agent is a character at a desk, every line comes from the durable session log, and any past case replays and reports from disk
Two-tier model router for DeepSeek Harness — automatic execution/judgment routing with LLM Judge, multi-model fallback chains, and exponential-backoff runtime failover. A DSH adaptation of pi-shift-router.
macOS Vision OCR tool plugin for DeepSeek Harness: read text from screen capture, clipboard images, or image files (zh-Hans + en-US). Unofficial community project.