PyFluent: Bridging Open-Source Python Automation with Proprietary Liquid Handling for AI-Driven Laboratories

Engineering notes

Key topics: self-driving, control. See the paper for implementation details and experimental results.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

Liquid-handling automation is increasingly central to high-throughput experimentation, closed-loop optimization, and AI-assisted laboratory workflows. Although commercial liquid-handling platforms provide robust hardware capabilities, integration with modern scripting environments, reproducible software practices, and adaptive experimental pipelines often remain limited by proprietary control ecosystems. We present PyFluent, a Python-native interoperability layer for the Tecan Fluent platform that bridges vendor-mediated hardware control with open, scriptable laboratory automation workflows. Rather than replacing FluentControl, PyFluent provides structured interfaces that expose Fluent execution pathways through reusable Python APIs while preserving compatibility with the vendor software stack. PyFluent is organized as a three-layer architecture consisting of a user-facing scripting layer, an execution engine, and the underlying proprietary Fluent control environment. The framework supports two complementary user surfaces: a Fluent-specific direct API pathway that preserves access to Fluent-native command semantics, and a PyLabRobot-compatible backend that enables higher-level, hardware-agnostic workflow development. Both pathways converge through a shared execution controller before dispatch through the FluentControl .NET interface. The execution engine includes XML command generation, optional state tracking, simulation support, and structured command routing designed to improve protocol development, testing, and integration with external software systems. We characterize the dispatch boundary of the Fluent control interface, showing that flexible-channel-arm (FCA) operations can be issued directly through the command channel while multi-channel-arm (MCA) and gripper operations are routed through a vendor-subroutine bridge, and we implement a hybrid execution model that exposes both behind a uniform Python API. Representative workflows including serial dilution, plate-to-plate transfer, and reagent addition, are composed in Python and validated in FluentControl’s simulation environment before hardware execution. PyFluent also supports simulation-first development, enabling protocol validation without requiring active hardware access during early workflow design. By exposing each Fluent operation as a discrete, callable method — usable from notebooks, optimization loops, and tool-calling AI agents — and pairing it with a simulation mode that runs before any hardware is engaged, PyFluent lets autonomous systems generate, validate, and execute protocols within a single Python session. More broadly, it shows how commercial laboratory hardware can become a programmable component of self-driving laboratories rather than an isolated endpoint, without sacrificing vendor-supported execution pathways.

Links and sources

Comments