Agent Cache Python v0.7.0
Fixed
- LlamaIndex adapter: tool definitions now included in cache key when
toolsis passed toprepare_params(). Only serializable metadata is used; callable closures are never serialized.
Changed
- Cache keys changed for tool-using requests on the LlamaIndex adapter. Existing cached entries for those requests will be a one-time miss after upgrade — intended, as the prior entries were not safe to reuse across differing tool sets.
prepare_params()now accepts atoolskeyword argument (andLlamaIndexPrepareOptions.tools). Omitting it falls back to messages-only keying (prior behavior).
Known limitations
- LangChain adapter: tool-schema drift is not reflected in the cache key — the framework's
BaseCacheinterface only exposes(prompt, llm_string). Documented limitation, unchanged.
Full Changelog: agent-cache-py-v0.6.1...agent-cache-py-v0.7.0