dsh-continual-evolve
DeepSeek Harness 的持续自我演进插件：根据会话轨迹不断优化版本化、可审计且可安全回滚的 Harness 状态（提示词备注、记忆、技能、子代理规范）。
插件会安装到这里；不确定时保持 web。
npx -y @deepseek-ai/dsh plugin --profile web add dsh-continual-evolve@0.6.0Continual Evolve 以 dsh-continual-evolve 发布，当前版本为 0.6.0。Plugin Hub 会校验它的 manifest，并保存精确安装来源，便于复现安装结果。
中文 | English
Continual self-evolution for DeepSeek Harness: a versioned, auditable, rollback-safe harness state layer — prompt notes, memories, skills, subagent specs — refined from session trajectories.
The model proposes, the code guarantees. Every mechanical safety property — schema validation, atomic writes, snapshots, versioning, audit trail, acceptance decisions — is enforced in code, never by prompt discipline.
Agents accumulate reusable experience (repeated failures, durable facts, reusable procedures) and forget it next session. This plugin turns that experience into first-class state:
/evolve consolidate turns write-time conflict hints and zero-use staleness into one approved, fully reversible batch of archives — with merge, near-duplicate content folds into the surviving originalevolve_add, or the automatic review gate proposes them from the session trajectory (turn-interval + compaction checkpoints).conflictHint for later consolidation.auto_regression benchmark).# from npm (installs and activates — ships its own bundle patch)
dsh plugin add dsh-continual-evolve
# or from source (first GitHub installs require approving the allowBuilds step)
dsh plugin add ZK-Andy/dsh-continual-evolve
Restart dsh web after installing or updating.
Commands (in-session):
| Command | Effect |
|---|---|
/evolve | help + current local store |
/evolve list · history · rollback <id> | inspect and revert (add global for the cross-session store) |
/evolve plan [msg] | run the LLM planner against the store |
/evolve wrapup | assess this session's local entries: promote / archive / keep |
/evolve archive · unarchive · demote <id> | hide from injection (data kept, restorable) — demote targets global noise |
/evolve consolidate [apply] [merge] | report (or apply) one batch archive of conflict-hinted + stale zero-use global entries; merge folds near-duplicate content into the survivors |
/evolve failures | aggregated failure classes (gate + benchmark) |
/evolve log [tail N] [session <id>] | plugin log |
/evolve export · import <path> | backup / restore a store |
/evolve mount · unmount <skillId> | hot-mount an executable skill as a live plugin |
/evolve goal [objective · done · block] | round-driven auto-review goal |
/evolve benchmark … | case lifecycle, runs, acceptance |
Model tools: evolve_list / add / update / delete / rollback.
For third-party consumers: every applied evolution (gate or manual) appends a structured evolve_complete event to reviews.jsonl (src/evolve-event.ts defines the shape) alongside the human-readable audit records.
Injection shape: prompt notes and delegation specs inject with content (≤6/kind × 180 chars, relevance-ranked). Memories and skills appear as a directory index ([kind:id] title, capped at 15 lines with a fold counter) — full text via evolve_list. Empty store = zero injected tokens.
| Key | Default | Meaning |
|---|---|---|
baseDir | resolved DSH home | root for the evolve/ stores |
autoReview | false | enable the automatic review gate |
reviewIntervalTurns | 6 | gate cadence on the turn-interval path |
maxReviewInputChars | 40000 | trajectory slice handed to the gate |
reviewBudgetTokens | 4096 | output budget for the gate call |
notifyOnAutoReview | true | visible follow-up notice after an applied gate run |
requireGlobalApproval | true | global edits ask for explicit approval |
localFate | true | gate audits local entries and proposes promote/archive (consulted, never silent) |
fateIntervalTurns | follows reviewIntervalTurns | minimum turns between fate assessments |
goalBlockedWrapupTurns | 3 | consecutive blocked-goal gate runs trigger one fate assessment (0 disables) |
promotionBlockPatterns | POSIX paths, session ids, ~/.dsh | content matching these is project-scoped and never promoted to global |
promotionMinChars | 100 | whole promotions below this length stay local |
injectionDirectoryLines | 15 | entry-directory lines per build before folding into a counter |
sectionOrder | 118 | system-prompt section order |
skillsDir | <dshHome>/skills | where skill entries materialize as SKILL.md bundles |
rubricKey | auto-generated key file | AES-256-GCM passphrase for benchmark rubrics ( overrides) |
Example profile patch:
- id: continual-evolve
config:
autoReview: true
reviewIntervalTurns: 6
pnpm install && pnpm build # deps + tsc -> lib/
pnpm test # vitest (573 tests)
pnpm test:coverage # v8 coverage, thresholds enforced in CI
pnpm lint # oxlint src test
Project layout:
├── src/ # engine, tools, commands, gate, fate, benchmark, usage…
├── test/ # vitest suites (36 files)
├── lib/ # build output (tsc)
├── docs/
│ ├── design.md # full design doc (hardening matrix)
│ ├── FAQ.md # real failure/fix records
│ ├── gap-analysis.md # vs prime-agent /refine + penguin-harness
│ ├── research/pi-dsh-competitor-gap-analysis.md # pi/dsh ecosystem competitors
│ ├── experiment-bootstrap.md
│ ├── archive/ # closed point-in-time reports
│ └── research/ # penguin report + prime-agent annotated source
├── examples/README.md # seed benchmark cases
└── .agents/ # AI collaboration layer (AGENTS.md, skills, ADR notes)
docs/design.md · Pitfalls: docs/FAQ.md · Gap analysis: docs/gap-analysis.md · D2 experiment: docs/experiment-bootstrap.mddocs/research/penguin-harness-self-evolution.md; prime-agent /refine (engineering shape; MIT) — annotated reference source in docs/research/prime-agent-refinement.ts. This package is an original implementation on the DSH plugin surface.DSH_EVOLVE_RUBRIC_KEYlogToFile / logLevel / logMaxBytes | true / 1 / 5 MiB | plugin-owned JSONL file log with rotation |
autoRollbackOnReject | true | deterministic rollback after a benchmark rejection |
autoCase | true | failed evolution attempts are captured as draft regression cases (auto_regression benchmark) |
reviewModel | agent's own | optional cheaper model for the gate ("provider/model") |