dsh-py-codeact
CodeAct agent loop for the DeepSeek Harness: a persistent IPython session as the model's action space, with harness tools bridged in as a virtual `__dsh__.tools` module.
236 results
CodeAct agent loop for the DeepSeek Harness: a persistent IPython session as the model's action space, with harness tools bridged in as a virtual `__dsh__.tools` module.
Edit a sent user message and resend it: fork a new branch from that message's turn and regenerate the reply with the edited text.
Persistent long-horizon autonomous mission runtime for DeepSeek Harness: dependency DAG, independent verification, replan-on-failure, lease ownership.
Replace the 'Deep diving...' running status in the DeepSeek Harness Web GUI with alternating witty phrases and practical tips.
Pin tool results or assistant replies, then recall them into the next model turn via /pin /pins /recall /unpin and Web card actions.
Turn DeepSeek Harness sessions into redacted, README-ready animated demos — as a local CLI or DSH plugin.
Shadow-replaces the built-in glob tool: auto-excludes bottomless directories (node_modules, dist, build, .venv, ...) and requires an explicit include whitelist to search inside them.
Controlled A/B comparisons and evidence-backed reports for DeepSeek Harness plugins and presets.
Replace DSH web tool icons with emoji: think 🤔, read/search 🧐, pwsh 👨💻, edit ☝️🤓, error 😫
Browser patrol and website inspection plugin for DeepSeek Harness: teach once, replay deterministic runbooks with managed Chromium, credentials, screenshots, checkpoints, and selector recovery.
Windowed, reversible context management for DeepSeek Harness; replaces Basic compaction.
Turn successful agent tool calls into reviewed, parameterized routines you can actually replay. DSH, MCP, and WebMCP.
OpenViking memory & context bundle for DeepSeek Harness: auto-recall, session capture, threshold commit, offline outbox replay, self-contained MCP tool bridge, viking:// URI guard.
Evidence-governed experience memory for DeepSeek Harness
Register each agent preset's own skills/ directory as discoverable, preset-isolated skills in DeepSeek Harness, including blank-session preset switching and a startup roster prewarm that shrinks the web catalog warm race. Replacement for the old preset-skills copy-in mechanism.
Keeps Charm Hyper requests under the gateway's 10 MiB body cap: replaces the oldest images with placeholders (newest survive) before sending, and retries with a tighter projection when the gateway still rejects the body.
Voice conversation mode for the DeepSeek Harness Web GUI: mic input with auto-submit, streaming spoken readout of replies, and a hands-free multi-turn loop with barge-in.
DSH web plugin: manageable quick-reply text chips in the current session input area; one tap sends the stored text as an ordinary user message (queue while idle, steer when the top-level session runs). Independent source, user data and DSH install stay separate for easy upgrade recovery.
消息气泡修改工具:在每条消息上提供「修改」按钮,编辑后通过 DSH 核心 surface replace 机制改写模型可见上下文(append-only 日志保留原事件),气泡文字经 localStorage 覆盖层同步显示
A library for patching, replacing and decorating DeepSeek Harness plugins during runtime
Voice for DeepSeek Harness: an AI voice-reply tool, a persistent voice bar, a Voice settings page, and a boot sound module — all backed by Alibaba Cloud Bailian TTS.
Unified DSH checkpoints with linked-worktree rollback, session replay, and a graphical conversation branch tree
Skill usage statistics for DSH: counts every skill invocation across sessions (replay of existing logs + live session/event subscription), detects zombie skills, and helps decide which skills to keep.
Custom sidebar plugin with a replacement brand logo (session tree, search, grouping).