Agents And Emulator Workflows
This repository keeps the agentic AI connection, but the center of the website is now the quantum-emulator workflow.
The practical division of labor is:
AI agent support -> data preparation and workflow scaffolding
quantum-inspired emulator -> decision-space exploration
human researchers -> interpretation and scientific judgment
Why Agents Still Matter
Environmental decision problems are rarely ready for optimization when a working group first defines them. The data may live in different formats, use different coordinate systems, require documentation, or need to be transformed into a candidate-site table.
An AI agent can help with that surrounding workflow:
- finding and checking data sources
- writing harmonization scripts
- generating decision-table preparation code
- documenting assumptions
- running lightweight validation
- keeping workflow outputs in predictable folders
The agent does not replace scientific review. It accelerates the construction of an inspectable workflow.
Why The Repository Is The System
The repository is the durable scientific object. It contains the code, documentation, instructions, outputs, and tests that make the workflow reproducible.
For this project, the repository defines:
- how ecological monitoring sites are represented
- how binary decision variables are constructed
- how a QUBO rewards biological value and coverage
- how cost and redundancy become penalties
- how emulator output is compared with a classical baseline
- how selected sites are mapped back to geography
That structure matters more than any single chat response. The model helps build the system; the system is what can be inspected, shared, and rerun.
Guardrails For This Project
The agent instructions in AGENTS.md are part of the methods record. They tell
future agents to:
- keep the emulator workflow central
- preserve the harmonizer as a data-preparation layer
- avoid claims of quantum advantage or speedup
- keep D-Wave cloud access out of the default path
- compare emulator results with a clear classical baseline
- document new workflows and append to the prompt action log
These constraints keep the project practical and honest.
From Data To Decision
The scientific value of this repository is not that an AI model or emulator produces an answer by itself. The value is that a working group can trace the path from environmental synthesis to a candidate-site table, from that table to a binary optimization model, and from solver output back to ecological interpretation.
That is the point of the agentic structure here: not less rigor, but a clearer and faster way to build workflows that researchers can inspect.