Extend Aider's cache TTL from 5 minutes to 1 hour for long sessions.
npx skills add https://github.com/OnlyTerp/prompt-cache-skills --skill aider-1h-ttl
Please help me install this Agent Skill. Skill: aider-1h-ttl Source: https://github.com/OnlyTerp/prompt-cache-skills/tree/main/skills/aider-1h-ttl Command: npx skills add https://github.com/OnlyTerp/prompt-cache-skills --skill aider-1h-ttl
This Skill extends Aider's cache TTL from 5 minutes to 1 hour, reducing the need for keepalive pings and preserving the cache across long sessions.
Enable the 1-hour cache for Aider with the command: "Apply aider-1h-ttl."
High-intent search queries and answers about installing and using this skill.
To extend Aider's cache TTL, you can apply a configuration that increases the expiration window from 5 minutes to 1 hour. This prevents cache invalidation during prolonged human-in-the-loop pauses in coding sessions.
The Anthropic prompt cache expires because its default TTL is set to 5 minutes. When human-in-the-loop reasoning tasks exceed this window without keepalive pings, the cached data is cleared to optimize resource allocation.
Yes, you can enable Anthropic's extended cache TTL feature within Aider. This modifies the cache management to sustain a 1-hour TTL, which is ideal for repetitive coding tasks that involve extended human delays.
The best way to reduce keepalive pings is to enable a 1-hour cache TTL for Aider. This extends the cache expiration window, cutting down on unnecessary API requests and saving on billing costs during long sessions.
Extending the cache TTL to 1 hour optimizes for Anthropic's extended cache feature, which applies a 2x write cost premium. However, it reduces overall billing by cutting down on unnecessary keepalive pings and cache rebuilds.
You should not use a 1-hour cache TTL for short, rapid-fire coding sessions where prompts are exchanged quickly. The 2x write cost premium for extended caching is only ideal for long, multi-hour sessions with extended human delays.
OnlyTerpEnable Aider's --cache-prompts flag by default for supported models.
ssdeanxExecute Gemini CLI commands for code analysis and generation.
devendrapratapsinghOptimize AI and LLM operations for token usage, cost, and inference.
OpenCovenMeasure and lay out text without triggering DOM reflow.
