promptcmd

Appearance

Caching ​

promptcmd includes built-in response caching to reduce API costs and improve response times for identical requests. When caching is enabled, identical prompts with the same parameters return cached responses instead of making new API calls.

How Caching Works ​

When you run a prompt:

  1. promptcmd generates a cache key based on:

    • Prompt content
    • Model configuration
    • Input parameters
    • STDIN content

  2. If a cached response exists and hasn't expired, it's returned immediately

  3. Otherwise, the request is sent to the AI provider and the response is cached

Enabling Caching ​

Global Caching ​

Enable caching for all providers:

toml
[providers]
cache_ttl = 60  # Cache for 60 seconds

Set to 0 to disable caching globally:

toml
[providers]
cache_ttl = 0  # No caching

Provider-Specific Caching ​

Override global settings per provider:

toml
[providers]
cache_ttl = 30  # Global: 30 seconds

[providers.anthropic]
cache_ttl = 60  # Anthropic: 60 seconds

[providers.openai]
cache_ttl = 15  # OpenAI: 15 seconds

[providers.ollama]
cache_ttl = 0  # Ollama: No caching

Variant-Specific Caching ​

Fine-tune caching per variant:

toml
[providers]
cache_ttl = 20

[providers.anthropic]
cache_ttl = 40

[providers.anthropic.frequently-used]
cache_ttl = 120  # Cache for 2 minutes

[providers.anthropic.always-fresh]
cache_ttl = 0  # Never cache