Install
$ | 0.1.2 及以后——npm 的 @deepseek-ai/dsh（现在的 latest 是 0.2.0-rc.2）、内核 0.1.2 以后的各家桌面版、或自建 master | dsh-maze 2.x，latest 标签 | dsh plugin --profile web add dsh-maze |Plugin Overview & Capabilities
AI-assisted organization based on the public repository snapshot. The content must be grounded in source evidence and does not replace compatibility or security verification.
Key Capabilities
- Interactive maze timeline visualizing main execution path, parallel tool branches, retries, and subagent runs
- Data tracks monitoring tool invocation density, token usage pulses, and context pressure curves with compression markers
- Deterministic execution analysis including failure recovery chains, scatter plots for tool latencies, and validation metrics
- Multi-session timeline comparison across 2 to 5 runs with turn alignment and branch inventories
- Execution replay up to 300x speed and vector/raster export (SVG, 2x PNG)
- Real-time session tab and dedicated sidebar session-log inspector supporting .jsonl and .jsonl.zstd
Useful For
- Analyzing where and why a DeepSeek Harness agent branched into failed tool calls or blind retries
- Comparing performance, token consumption, and execution differences across various models or prompts on identical tasks
- Diagnosing context window usage spikes and compression effects during prolonged multi-turn sessions
- Auditing and sharing agent execution logs with exportable timeline visuals
Who It Fits
- DeepSeek Harness developers and agent engineers debugging complex execution trajectories
- Prompt engineers and AI researchers benchmarking multi-turn tool calling performance across models
- Power users troubleshooting long-running DSH autonomous tasks
Documented Limitations
- Requires DeepSeek Harness host versions compatible with Cordis/DSH client splits (v2.x requires DSH 0.1.2+; specific host version alignment needed for subagent features)
- Real-time tab only visualizes loaded session history windows; full historical analysis requires uploading raw session log files
- Code mode internal script commands (within run_code) are not individually parsed at the granular sub-command level
DSH Compatibility
Version-specific runtime evidence collected by DSH Plugin. A missing result means we have not tested that combination yet.
Security Signals
Objective signals discovered from package metadata and source inspection. These are not a guarantee that a plugin is safe.
package.json declares 1 runtime, 25 development, 16 peer, and 0 optional dependencies.
package.json declares DSH bundle metadata.
GitHub reports the repository license as MIT.
A root package.json was captured and can be inspected by the registry.
Public GitHub source metadata is available for this registry snapshot.
Source & Registry Notes
Public provenance, Registry classification, and the latest source check for this entry, kept separate from runtime verification.
- Source repository
- lamost423/dsh-maze
- Registry source
- GitHub · dsh-plugin topic
- Registry classification
- Plugin
- Source checked
- fb83f30 · 2026-10-07
This project is independently indexed from public source information. DSH Plugin is not affiliated with DeepSeek or the plugin author. Always check the author repository before installation.
Repository Activity
- GitHub stars
- 85GitHub stars
- Forks
- 4
- Open issues
- 1
- Last commit
- 2026-10-05
- Last release
- 2026-10-05
Related DSH Plugins
Ranked by overlapping capabilities, use cases, plugin type, categories, and DSH profile.
Context visualization, inspection, and lifecycle management dashboard and /context command for DeepSeek Harness.
View plugin →DeepSeek Harness balance and token monitor with whale-girl companion animations and visual billing rules.
View plugin →A comprehensive Web GUI plugin suite and AI agent workbench for DeepSeek Harness, featuring task boards, mobile remote access, SSH operations, and visual tools.
View plugin →Enhanced sidebar workbench and extensible foundation for DeepSeek Harness with an interactive editor, file tree, Git diffs, tasks, and side chat.
View plugin →