Conceptual Overview
ReasonChip is a modular framework for automating complex intellectual labor through declarative pipelines and dynamic execution units called Chips. Its architecture is designed to be scalable, distributed, and developer-friendly, making it ideal for building advanced reasoning systems, automation tools, and smart agents.
Workflows
A workflow is the execution of one or more pipelines - moving step by step through a reasoning or automation process. Workflows are top-down and sequential by default, but can branch, pause, restart, or cancel based on logic defined in the pipeline. They can span multiple pipelines and respond dynamically to real-time results.
A pipeline is a YAML-defined sequence of tasks that express how work is carried out. Pipelines are clean, configurable, and declarative, removing the need for orchestration code. They support conditional logic, branching, sidequests, async tasks, and more.
A task is a single step in a pipeline. It is used to call chips to perform the function of the task.
A Chip is an async Python function that performs a focused task—such as sending a message, making a decision, or calling an API.
Chips are registered using @Registry.register and are discoverable by pipelines.
Each Chip uses Pydantic for structured input/output validation and can be composed into larger workflows.
System Architecture
ReasonChip operates as a distributed system composed of three roles:
- Clients: Submit workflows and receive results
- Brokers: Manage execution flow and dispatch tasks to workers
- Workers: Run chips and return results, logs, or intermediate states
This architecture supports secure, scalable execution locally or across cloud environments.
Developer Experience
Developers using ReasonChip focus on logic - not orchestration. They:
- Write async Chips with validated input/output
- Register Chips with the system for dynamic execution
- Compose pipelines in YAML to model reasoning and automation
- Debug workflows via streaming logs and result tracking
Why ReasonChip?
- Modular: Reuse small Chips across many workflows
- Declarative: Define what should happen, not how
- Async-native: Designed for concurrency and responsiveness
- Composable: Easily build intelligent systems from simple parts
Summary
| Concept | Description |
|---|---|
| Workflow | A sequence of pipelines executed in sequence to achieve an outcome |
| Pipeline | YAML-defined sequence of tasks |
| Task | A single chip execution in a pipeline |
| Chip | A single function that performs an operation |
| Client | Sends pipeline jobs and receives output |
| Broker | Manages job routing, state, and lifecycle |
| Worker | Executes pipelines and returns results |
© 2025 South Patron LLC. All rights reserved.
AI everywhere.