dsh-cache-miss
Run the following command in DeepSeek Harness:
dsh plugin install wefio/dsh-cache-miss
Paste the following prompt into your AI chat to install this plugin:
To install the dsh-cache-miss plugin in DeepSeek Harness, run the command dsh plugin install wefio/dsh-cache-miss. Source: https://github.com/wefio/dsh-cache-miss
About this plugin
In multi-turn conversations, the first model call is often the moment a cache miss occurs, triggering a full re-prefill. DSH's native statistics, whether displayed below the input box or in the artifact file, often fail to provide immediate feedback at the exact moment of a miss, making it difficult for developers to monitor cache efficiency in real-time.
This plugin renders a yellow prompt directly under the assistant's reply, precisely displaying the cache status of the current request. It not only counts the missed input tokens (re-billed) and the cached portion, but also provides key metrics like idle duration and Time-to-First-Token (TTFT), making the comparison between re-calculation volume and cache hits instantly clear.
This tool is designed for developers and agent builders who need to fine-tune LLM costs, monitor cache hit ratios, or debug performance in long conversations within DSH.
Screenshots
Use Cases
- Monitor cache hit rates in long conversations
- Identify and reduce costly token re-calculation
- Debug model response Time-to-First-Token latency
Best For
- Developers seeking to optimize LLM costs
- Agent engineers debugging multi-turn performance
- AI researchers focusing on cache efficiency
Related Plugins
Traceable, searchable cross-session memory for AI agents that turns conversation knowledge into a typed knowledge graph and recalls relevant subgraphs instead of replaying full history, natively integrated with DeepSeek Harness.
Gives DSH AI cross-session long-term memory, to-do and skill management, plus multi-session orchestration, external AI delegation, and an infinite canvas that grows with you.
dsh-mnemon is a three-tier, pluggable, Agent-driven memory system for DeepSeek Harness, combining Runtime memory, Project Documents, and replaceable Memory Spaces with nine long-term providers.
An opinionated, zero-infrastructure file-based memory protocol: human-readable markdown files plus a bounded always-loaded index, curation discipline, and no database or embeddings.