* fix(litellm): fall back to last user message when hindsight_query not provided inject_memories=True no longer requires an explicit hindsight_query. The injection path now falls back to extracting the last user message, matching the documented Quick Start behavior that was broken since #167 (v0.4.18). * test(litellm): add regression tests for inject_memories without hindsight_query |
||
|---|---|---|
| .. | ||
| agno | ||
| ai-sdk | ||
| chat | ||
| crewai | ||
| hermes | ||
| langgraph | ||
| litellm | ||
| nemoclaw | ||
| openclaw | ||
| pydantic-ai | ||