← DeepSeek Harness Plugins
DeepSeek Harness PluginManifest Valid

Lume

Task execution discipline, methodology artifacts, behavioral triggers, observability metrics, and persona memory system for DSH.

VisionDeveloper Tools
4GitHub Stars0ForksUpdated2026-09-26

Install

No clear install command was found in the current repository evidence.

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.

source-grounded
Lume (微光) is an enhancement plugin for DSH Desktop that enforces task execution discipline and structured workflows. It operates across four core areas: a discipline layer that classifies intent (Q&A, search, discuss, diagnose, execute) and checks evidence before modifying files or drawing conclusions; a methodology layer offering structured ledgers for contracts, changes, hypotheses, design decisions, and workspace-scoped project knowledge; an observability metrics dashboard (`lume_metrics`) logging trigger hits and routing decisions locally; and a persona system providing named characters distilled from real source materials with long-term and temporary memories.

Key Capabilities

  • Intent routing and task boundaries (Q&A, search, discuss, diagnose, execute) to prevent out-of-scope modifications
  • Evidence checks for unverified file line citations, non-existent symbol assertions, requirement coverage, and context window threshold warnings
  • Behavioral triggers (converge, verify-as-you-go, dead-path, criteria-drift) designed to correct trajectory drift with specific numerical metrics
  • Methodology tools: contracts (`lume_contract`), change ledgers (`lume_change`), hypothesis ledgers (`lume_hypothesis`), and design records (`lume_design`)
  • Persistent workspace-scoped project knowledge with sensitive content filtering and idempotent background distillation from past sessions
  • Local runtime metrics logging (`lume-metrics.jsonl`) and efficiency evaluation via `lume_metrics`
  • Persona distillation and long-term memory system with style drift correction and memory star-map visualization
  • Layered injection architecture to preserve prompt prefix caching on supported DSH hosts

Useful For

  • Guiding complex coding sessions with strict verification and automatic change tracking without relying purely on LLM self-discipline
  • Preventing hallucinations and unwarranted negative assertions during technical diagnosis and refactoring
  • Carrying over structured session memories (goals, decisions, open items) across long conversations and workspace reboots
  • Distilling custom character personas with tone adherence, long-term memory, and lexical leak correction

Who It Fits

  • DSH Desktop developers handling multi-step coding, debugging, and refactoring tasks
  • Users needing strict verification, structured reasoning ledgers, and cross-session knowledge retention
  • Developers looking for verifiable behavioral metrics on agent decision-making
  • Users who want nuanced, roleplay-capable personas with anchored speaking styles

Documented Limitations

  • Mechanical criteria check whether factual evidence exists, but cannot judge semantic correctness of LLM reasoning
  • Does not block destructive commands or perform irreversible operation blocking (removed in v0.8.1 to respect explicit user instructions)
  • Requires a full restart of DSH Desktop (including tray processes) after installation or upgrades due to host loading constraints
  • Metrics evaluate observational co-occurrence within a 3-turn window, not strict causality
  • Hosts without `systemPrompt.context` fall back to single-system-prompt injection, which sacrifices prefix caching

DSH Compatibility

Version-specific runtime evidence collected by DSH Plugin. A missing result means we have not tested that combination yet.

Not tested yetNo runtime compatibility tests have been published yet.

Security Signals

Objective signals discovered from package metadata and source inspection. These are not a guarantee that a plugin is safe.

Dependency Counts

package.json declares 0 runtime, 18 development, 5 peer, and 0 optional dependencies.

info
Dsh Bundle Declared

package.json declares DSH bundle metadata.

info
License Declared

GitHub reports the repository license as MIT.

info
Package Manifest Available

A root package.json was captured and can be inspected by the registry.

info
Source Available

Public GitHub source metadata is available for this registry snapshot.

info

Source & Registry Notes

Public provenance, Registry classification, and the latest source check for this entry, kept separate from runtime verification.

Source repository
cayan0x/Lume
Registry source
GitHub · dsh-plugin topic
Registry classification
Plugin
Source checked
88dc68e · 2026-09-28

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
4GitHub stars
Forks
0
Open issues
0
Last commit
2026-09-26
Last release
2026-09-26
For maintainers

Maintaining this plugin?

This listing is generated from public repository data. If you maintain this project, you can review the information and share this listing with your users if you find it useful.

Add to README
Listed on DSHPlugin.app