hermes - 💡(How to fix) Fix Feature: Show prompt cache hit rate in CLI status bar [1 pull requests]

Problem

Major providers (OpenAI, Anthropic, DeepSeek, xiaomi, etc.) support prompt caching, which significantly reduces API costs. However, users have no visibility into how effectively caching is working in their current CLI session.

Proposed Solution

Add a ◎ XX.X% indicator to the CLI status bar showing the prompt cache hit rate (cache_read_tokens / prompt_tokens).

Display behavior

• Color coded: ≥70% green, 40-70% yellow, <40% red (higher is better)
• Adaptive precision: Integer on narrow terminals, one decimal on wide
• Conditional: Only shown when provider returns cache data
• Ultra-narrow: Hidden to save space (<52 chars)

Example

Wide:   ⚕ mimo-v2.5 │ 12K/200K │ 6% │ ◎ 76.2% │ 3m20s
Narrow: ⚕ mimo · 6% · ◎ 76% · 3m20s

Use Case

Users running long sessions or monitoring API costs can immediately see:

• Whether their provider's caching is working effectively
• When cache is being utilized vs when it's not
• How switching providers affects cache performance

Technical Details

• Uses existing fields: session_cache_read_tokens and session_prompt_tokens
• Compatible with all providers that return prompt_tokens_details.cached_tokens
• No new data collection needed