dsh-memoryos
Evidence-first long-term project memory for DeepSeek Harness
AI 分析
核心用途是为智能体引入长期项目记忆。适合需要跨会话保持项目上下文、且本地已运行 MemoryOS 服务的用户。
インストール
$
npx -p @deepseek-ai/dsh dsh plugin --profile web add github:tianhao8687/dsh-memoryosドキュメント
README 全文を読む ↗Configuration
| Environment variable | Default | Meaning |
|---|---|---|
MEMORYOS_ENABLED | 1 | Initial state only when no persisted control state exists; 0 starts in ordinary off mode |
MEMORYOS_CONTROL_ENABLED | 1 | Expose the model-callable on/off/status control tool outside strict no_memory |
MEMORYOS_ONBOARDING_NOTICE | 1 | Ask the Agent to relay the first-success notice once |
MEMORYOS_STATE_FILE | OS config directory | Optional per-profile path for persistent switch/onboarding state |
MEMORYOS_BASE_URL | http://127.0.0.1:8000 | Local MemoryOS endpoint |
MEMORYOS_AUTH_TOKEN | none | Local MemoryOS bearer token |
MEMORYOS_CONDITION | msc_progressive | Context-delivery condition |
MEMORYOS_BUDGET_TOKENS | 6000 | MemoryOS response budget |
MEMORYOS_MAX_CONTEXT_CALLS | unlimited | Per-session context-call ceiling; 0 means unlimited |
MEMORYOS_RESPONSE_FORMAT | json | json, deepseek-compact, or progressive compact mode |
MEMORYOS_TOOL_PROFILE | read-only | read-only or explicit cross-session-write |
MEMORYOS_REPOSITORY | none | Fixed repository scope; required for writes |
MEMORYOS_TASK | none | Controller-owned task description |
MEMORYOS_TIMEOUT_MS | 30000 | Local MemoryOS request timeout |
Usage-ledger and controlled-eviction variables are evaluation infrastructure;
see cordis.patch.yml and the architecture document before
enabling them.