Overview
dsh-warm
DeepSeek Harness host plugin: prompt-cache warmer. While a session's agent loop is blocked inside a tool execution (bash, wait_agent, job_output, ...), re-dispatches the session's last conversation request through the llm waterfall with max_tokens=1 every interval so the provider's prompt-cache prefix stays warm; never warms idle sessions or user-input tools. Journal + budget at $DSH_HOME/storages/dsh-warm/.
README / EN
Package documentation
Registry summary
DeepSeek Harness host plugin: prompt-cache warmer. While a session's agent loop is blocked inside a tool execution (bash, wait_agent, job_output, ...), re-dispatches the session's last conversation request through the llm waterfall with max_tokens=1 every interval so the provider's prompt-cache prefix stays warm; never warms idle sessions or user-input tools. Journal + budget at $DSH_HOME/storages/dsh-warm/.
dsh.pub verifies the pinned bundle contract, runtime facts, and distribution semantics. The complete README remains in the source repository.
Read the full README on GitHubLIMITATIONS
Known limitations
- Provider scope: only providers with an entry in the `providers` table are warmed, and each needs a measured or documented TTL floor. deepseek and opencode-go are absent by default for that reason (see the configuration section). - The re-dispatch goes through the profile's main `llm` service; a provider adapter registered only on an isolated service would serialize differently (none exist in this profile today). - The 302 s TTL floor is a Mistral measurement for `zai-glm-5-3`; other providers need their own probe before being added to the `providers` table.