Aider Configuration for docs2prompt

A production-grade Aider configuration for the docs2prompt project, an agentic tool that converts documentation into formats suitable for complex structured prompts.

What This Configuration Does

This skill configures Aider with:

Instructions

1. Create Environment File

Create `.env.aider` in your project root with your API keys:

```bash

ANTHROPIC_API_KEY=your_anthropic_key_here

DEEPSEEK_API_KEY=your_deepseek_key_here # optional

OPENROUTER_API_KEY=your_openrouter_key_here # optional

```

2. Create Project Documentation Structure

Set up the required specs directory:

```bash

mkdir -p specs

touch specs/CONVENTIONS.md specs/project.md specs/mvp.md

```

Populate these files with:

3. Save Configuration

Save the following as `.aider.conf.yml` in your project root:

```yaml

Primary model (best for production)

model: claude-3-5-sonnet-20240620

Alternative models (uncomment to switch)

model: deepseek/deepseek-chat

model: openrouter/openai/o1-preview

Environment and caching

env-file: .env.aider

cache-prompts: true

cache-keepalive-pings: 5

UI and workflow

light-mode: true

auto-commits: true

stream: false

Testing configuration

test-cmd: pytest

auto-test: true

Architect mode for high-level design

architect: true

editor-model: claude-3-5-sonnet-20240620

editor-model: deepseek/deepseek-chat

editor-edit-format: diff

Context files (always loaded)

read:

- specs/CONVENTIONS.md

- specs/project.md

- specs/mvp.md

```

4. Launch Aider

Start Aider in your project directory:

```bash

aider

```

Aider will automatically:

5. Example Usage

```bash

Start with a high-level design question

aider> "Design a module to parse Markdown documentation and extract code blocks"

Request implementation

aider> "Implement the markdown parser with error handling"

Tests run automatically after changes

If tests pass, changes are auto-committed

```

Configuration Options Explained

| Setting | Value | Purpose |

|---------|-------|---------|

| `model` | claude-3-5-sonnet-20240620 | Primary model for code generation |

| `env-file` | .env.aider | Separate env file for API keys |

| `cache-prompts` | true | Reduce API costs via prompt caching |

| `cache-keepalive-pings` | 5 | Keep cache alive between requests |

| `auto-commits` | true | Commit successful changes automatically |

| `test-cmd` | pytest | Command to run tests |

| `auto-test` | true | Run tests after every change |

| `architect` | true | Enable high-level design mode |

| `read` | specs/*.md | Context files loaded on startup |

Switching Models

To use DeepSeek (cost-effective for iteration):

```yaml

model: deepseek/deepseek-chat

editor-model: deepseek/deepseek-chat

```

To use OpenRouter's O1 Preview (advanced reasoning):

```yaml

model: openrouter/openai/o1-preview

```

Best Practices

1. **Keep specs updated**: Aider reads these files on every startup—keep them current

2. **Use architect mode**: Ask design questions before implementation requests

3. **Trust auto-test**: Let Aider verify changes before committing

4. **Review auto-commits**: Check commit messages; they're usually descriptive

5. **Disable streaming for stability**: `stream: false` prevents output issues with some models

Constraints

Related Files