FAVA Trails exposes lifecycle hooks at key moments — before_save, on_recall, before_propose, and more. Protocols are pre-built hook implementations that plug into these points. Each is independent; enable any combination.
Three pieces of recent academic research turned into composable protocols:
- SECOM — extractive compression based on the SECOM paper (Tsinghua University and Microsoft, ICLR 2025).
- ACE — playbook-driven reranking and anti-pattern detection based on Stanford, UC Berkeley, and SambaNova ACE.
- RLM — MapReduce orchestration for fan-out / collect workflows based on MIT RLM.
The full reference — hook points, install commands, configuration examples, and a combining pattern for production stacks — lives on the Machine Wisdom AI site, where the protocol design is discussed in depth.
Continue reading on Machine Wisdom AI
SECOM, ACE, and RLM with config examples, hook points, known limitations, and a composable production stack pattern.
Read the full article on Machine Wisdom AI →If you’re here to install — pip install fava-trails[secom] and fava-trails ace setup --write are the entry points. The FAVA Trails README has the full setup, and the article above has the per-protocol config tables.