Self-learning LLM runtime — TurboQuant KV-cache (6-8x compression), SONA adaptive learning, FlashAttention, speculative decoding, GGUF inference

A persistent key/value cache

Improve Pi prompt/KV cache hit rates with stable prompts, OpenAI-compatible cache keys, proxy compat warnings, and footer cache stats.

Persistent KV-cache/session-state manager for DeepSeek Harness (llama.cpp slot snapshots)

- Supports [Cloudflare KV](https://developers.cloudflare.com/kv/) cache for `@envelop/response-cache` plugin - Suitable for graphql servers running on [Cloudflare Workers](https://workers.cloudflare.com/)

Typed KV cache for nifra - TTL, stale-while-revalidate, tag invalidation, and single-flight stampede protection on a pluggable store (in-memory default; bring CF KV / Redis). Dependency-free, runs on Bun/Node/Deno/Workers.

Codex-style context visibility for DeepSeek Harness: live context-window occupancy, composition breakdown, KV-cache hit rate, and a GitHub-style 53-week usage heatmap

Routes requests to KV assets

OpenCode plugin for LLM prompt caching, token savings, and KV cache reuse across AI agent sessions

Official Mingxin MCP server (stdio): signed AI-inference-storage benchmarks (KV cache tiering, throughput +29-40%, TTFT -26-32% on AMD MI308X), technical article search, and KV-cache tiering ROI estimation. Every figure carries a test-report citation.

Self-learning LLM runtime — TurboQuant KV-cache (6-8x compression), SONA adaptive learning, FlashAttention, speculative decoding, GGUF inference

AI agent memory & session orchestrator for MCP — persistent KV-Cache, Soul Board, immutable Ledger

Bundled dsh badge skill provider for DeepSeek Harness

Durable immutable attachment storage seam for the DeepSeek Harness

Storage hub (ctx.storage): named backend registry plus mounted data-form facilities for the DeepSeek Harness

Improve OMP prompt/KV cache hit rates with safer prompt churn cleanup, proxy compat warnings, and footer cache stats.

Fold the DSH skill catalog prompt surface: static KV-cache-stable catalog + BM25/bge-m3 hybrid skill_search + autoRoute hints. Rendering-only fold: source.entries stay untouched, the skill tool and /name injection stay host-owned. v0.3.0 adds a semantic r

DSH 上下文透镜 + 成本治理插件：token 用量可视化（每次请求的上下文构成、KV 缓存命中率）+ 每轮/每会话预算控制（超限提醒/拦截）。DSH web plugin: context lens (token usage visualization, KV cache hit rate) + cost governance (per-turn / per-session budget with warn/deny).

Official DeepSeek Harness brand occupants for the Web client's sidebar and conversation Hero slots

Read-only Remote projection of current Cordis Loader plugin state