cached_responses.yaml

# Demonstrates the response cache.
#
# The first time a question is asked, the agent calls the model normally and
# stores the answer. The second time the same question is asked, the answer is
# replayed verbatim and the model is not invoked at all.
#
# Two normalization options control what "same question" means:
#   - case_sensitive: when false (the default), "Hello" and "hello" match.
#   - trim_spaces: when true, leading and trailing whitespace is ignored.
#
# Storage is in-memory by default. Set `path` to persist entries to a JSON
# file that is reloaded on startup so the cache survives restarts. Multiple
# processes can safely share the same file: an advisory lock on
# `<path>.lock` serializes writes, and Lookup reloads the in-memory map
# when the file changes externally.

agents:
  root:
    model: openai/gpt-5-mini
    description: A helpful AI assistant with a response cache
    instruction: |
      You are a knowledgeable assistant that helps users with various tasks.
      Be helpful, accurate, and concise in your responses.
    cache:
      enabled: true
      case_sensitive: false  # "Hello" == "hello"
      trim_spaces: true      # "  hello  " == "hello"
      path: ./cache.json     # remove this line to keep the cache in memory only