Agentic Science Worker Configuration

An Aider configuration that transforms your AI assistant into an autonomous agent for computational materials science research. This setup pre-loads comprehensive context about materials science workflows, simulation tools, and research methodologies.

What This Does

This configuration automatically loads specialized skills and agent instructions when Aider starts, enabling:

Instructions for AI Agent

When this configuration is active:

1. **Load Context Files**: You have access to `AGENTS.md` and six specialized skill files covering materials science research workflows. Reference these to understand research protocols, tool syntax, and best practices.

2. **Auto-Commit Behavior**: Auto-commits are disabled by default (`auto-commits: false`) to give precise control during research benchmarking and experimentation. Enable if you want automatic version control.

3. **Model Selection**:

- Uncomment your preferred model in the configuration

- Claude 3.5 Sonnet recommended for complex materials science reasoning

- GPT-4o for broad general knowledge

- DeepSeek Coder for code-heavy simulation scripting

4. **Research Workflow Support**:

- Assist with setting up LAMMPS input files (force fields, ensembles, thermostats)

- Generate Quantum Espresso input decks (k-points, pseudopotentials, convergence)

- Write SLURM/PBS job submission scripts optimized for HPC resources

- Search literature and extract relevant findings

- Query materials databases (Materials Project, OQMD, AFLOW)

- Analyze simulation outputs (RDF, MSD, band structures, DOS)

5. **File Organization**: Maintain clear project structure:

- `simulations/` - LAMMPS/QE input files

- `jobs/` - HPC submission scripts

- `analysis/` - Post-processing notebooks and scripts

- `data/` - Raw simulation outputs

- `reports/` - Findings and visualizations

6. **Safety Checks**: Before submitting HPC jobs:

- Validate input file syntax

- Check resource requests (cores, memory, walltime)

- Verify data paths and dependencies

- Confirm convergence criteria are reasonable

7. **Documentation**: Generate clear README files explaining:

- Research objectives

- Simulation parameters and rationale

- Analysis methodology

- Key findings and next steps

Setup

1. Copy configuration to project root:

```bash

cp configs/aider/.aider.conf.yml .aider.conf.yml

```

2. Ensure required context files exist:

- `AGENTS.md` - Agent instructions and research protocols

- `skills/lammps-simulation/SKILL.md`

- `skills/quantum-espresso/SKILL.md`

- `skills/hpc-cluster/SKILL.md`

- `skills/literature-search/SKILL.md`

- `skills/materials-database/SKILL.md`

- `skills/data-analysis/SKILL.md`

3. Start Aider - context will auto-load:

```bash

aider

```

Configuration Options

**Model Selection** (uncomment one):

**Auto-Commits**:

**Other Options**:

Use Cases

**Starting a New Simulation Study**:

```

"Set up a LAMMPS simulation to study the glass transition temperature of

polymer melt using NPT ensemble and Nose-Hoover thermostat"

```

**DFT Convergence Testing**:

```

"Generate a series of Quantum Espresso input files to test k-point

convergence for silicon band structure calculation"

```

**HPC Job Optimization**:

```

"Write a SLURM script to run 20 parallel LAMMPS jobs with optimal

resource allocation for our cluster's node configuration"

```

**Literature Review**:

```

"Search for recent papers on machine learning potentials for

metallic glasses and summarize the key methodologies"

```

**Data Analysis**:

```

"Analyze the LAMMPS trajectory file to compute radial distribution

function and mean squared displacement over time"

```

Important Notes