crymeariva/igemapplication

Purpose and Top-Level Shape

iGEM OSAGE is a laboratory hardware-control and experiment-cycle design application. It combines a React dashboard, a Node.js service with an LLM-backed agent, and separate Raspberry Pi/Arduino hardware code. The central product concept appears to be creating, saving, loading, and executing/representing experimental “cycles” composed of hardware nodes.

Runtime Entry Points

• React browser client: my-app/src/index.js mounts my-app/src/App.js; static hosting begins at my-app/public/index.html.

• Node backend: my-app/server/index.js is the server entry point, with persistence initialized through my-app/server/database.js.

• Arduino motor firmware: ArduinoScript/MotorScript/MotorScript.ino.

• Raspberry Pi I²C integration/legacy test script: PiScript/i2cTest.py.

Interactive Experiment Canvas

The frontend’s primary orchestration layer is my-app/src/App.js, which assembles the dashboard and likely owns canvas-level state. Hardware is modeled as distinct visual/configurable node types:

• my-app/src/Components/HardwareNodes/PeristalticPumpNode.js

• my-app/src/Components/HardwareNodes/SyringePumpNode.js

• my-app/src/Components/HardwareNodes/SpectrometerNode.js

• my-app/src/Components/HardwareNodes/ElectroporatorNode.js

• my-app/src/Components/HardwareNodes/ThermometerNode.js

This creates a boundary between experiment composition in the browser and device-specific execution/configuration.

Cycle Lifecycle and Persistence

Cycles move between canvas state, serialized storage, and restored canvas state:

• Canvas reconstruction: my-app/src/canvas/addCycleToCanvas.js

• Canvas serialization: my-app/src/utils/serializeCanvasContext.js

• Save/load/delete lifecycle hooks: my-app/src/hooks/useCycleSave.js, my-app/src/hooks/useCycleLoader.js, my-app/src/hooks/useCycleDelete.js

• Backend client boundary: my-app/src/api/cyclesApi.js

• Server persistence boundary: my-app/server/database.js

The load dialog and side menu are the user-facing access points for stored workflows: my-app/src/Components/LoadCycleDialog.js and my-app/src/Components/SideMenu/Sidemenu.js.

Agent-Assisted Cycle Generation

A backend agent pipeline interprets user requests and creates valid experimental cycles:

• Request/message coordination: my-app/server/agent/handleMessage.js

• Intent classification: my-app/server/agent/intent.js

• Cycle generation: my-app/server/agent/generateCycle.js

• Generated-cycle validation: my-app/server/agent/validateCycle.js

The frontend invokes this service through my-app/src/api/agentApi.js, with user controls in my-app/src/Components/AgentMenu/AgentMenu.js and settings in my-app/src/Components/AgentSettingsDialog/AgentSettingsDialog.js.

LLM and Knowledge Boundary

The LLM integration is isolated in my-app/server/llm.js; agent identity/configuration is separated in my-app/server/agentKeys.js. Prompt/domain knowledge is file-based under:

• my-app/server/context/INSTRUCTIONS.md

• my-app/server/context/GENERATE_CYCLE.md

• my-app/server/context/MOSMAGE.md

This keeps language-model prompting and agent policy on the server rather than exposing it to the browser.

Hardware Integration Boundary

The web system is architecturally separate from low-level device runtimes. Arduino C++ firmware controls motor behavior in ArduinoScript/MotorScript/MotorScript.ino; Raspberry Pi Python handles I²C-side communication in PiScript/i2cTest.py. The README labels Pi code as legacy, so the Node service and frontend should not be assumed to directly depend on it without confirming protocol wiring. Material runtimes are browser/React, Node.js, Python on Raspberry Pi, and Arduino firmware.