kernex-pipelines
Multi-agent topology execution for the Kernex AI agent runtime.
Provides declarative pipelines that chain multiple agents, route outputs between them, and enforce per-step gates. Topologies are described as TOML and executed against any Provider implementation.
Used to build patterns like:
- 2-phase analyst → synthesizer (a planner reviews a problem, a builder turns the plan into output)
- N-way reality-check fan-out (one agent produces, several review, a reducer aggregates)
- Sequential evaluators with a final ship/reject gate
You usually consume pipelines through kernex-runtime, which loads topologies from disk and binds them to a runtime instance.
Documentation
- API reference: https://docs.rs/kernex-pipelines
- Project overview: https://github.com/kernex-dev/kernex
License
Apache-2.0 OR MIT.