DeepSeek Harness Plugins · Memory
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
Integrates the tokensave semantic graph engine into DSH, giving agents 80+ semantic tools to replace brute-force grep and saving 900-2000 tokens per query with zero configuration.
Cited memory over DSH's lossless session log, distilling durable facts into auditable markdown files with lossless escalation to exact original context.
A zero-dependency DSh plugin that eliminates JSONL cold reads via tiered caching, session preheating, and write coalescing to speed up session switching and reduce SSD wear.
A long-term memory plugin for DeepSeek Harness powered by a local Markdown vault, with fused retrieval injection, per-turn auto-capture, and model-driven recall, with zero runtime dependencies.
Cross-session persistent cost ledger for DeepSeek Harness that auto-logs every LLM token-usage event to SQLite with built-in DeepSeek pricing and record_cost / query_cost / set_budget agent tools.
Pin assistant replies or tool evidence in a session, then recall them into the model context via Web card or commands.
Export DSH conversations to an Obsidian vault as Markdown notes, with deduplication, incremental updates, and tools for reading, searching, tagging, and backlinking.
Connects DeepSeek Harness to a Noc Memory MCP server for session boot, daily briefings, and full memory CRUD operations.
A local memory and knowledge hub for DeepSeek Harness that unifies update, retrieve, and subscribe services with pluggable memory implementations and a transactional Rust storage backend.
A provider-neutral memory service that gives AI coding agents durable cross-session memory through a recall-reflect loop, automatically accumulating project knowledge without manual notes.
A thin on-demand context-retrieval layer for DeepSeek Harness that exposes three model-facing tools to read persisted conversation history.
Integrates MemoryHub into DeepSeek Harness with auto memory loading, checkpoint saving, six CLI tools, and a web UI memory tab.