Solo-dev optimized Aider setup with low token usage, fast feedback, safe Git history, and multi-model support. Includes color customization, linting, and architect mode.
A comprehensive Aider configuration optimized for solo developers, featuring low token usage, fast feedback loops, safe Git history management, and multi-model flexibility. Works with any codebase.
This `.aider.conf.yaml` file configures Aider (AI pair programming tool) with:
Create a file named `.aider.conf.yaml` in your project root (or `~/.aider.conf.yaml` for global config):
```yaml
model: openrouter/deepseek/deepseek-r1 # primary model for code generation
editor-model: anthropic/claude-3-5-sonnet-20241022 # model used for actual edit operations
weak-model: openrouter/anthropic/claude-3.5-haiku-20241022 # summariser / repo-map helper / commits / chat history
alias:
- "fast:openrouter/openai/gpt-4o" # low-latency, high-quality edits
- "smart:openrouter/openai/o1" # deep reasoning for complex tasks
- "hacker:openrouter/anthropic/claude-3.5-haiku-20241022" # cost-effective wide-context model
- "edit:o4-mini"
#edit-format: diff # Specify what edit format the main LLM should use(API model replies)
#editor-edit-format: editor-diff # Specify the edit format for the editor model(copy/paste or architect mode[editor-diff and editor-whole])
lint-cmd:
python: flake8 # Python linter
yaml: yamllint # CloudFormation templates, SAM templates
shell: shellcheck # Bash/POSIX shell scripts
json: jq . > /dev/null # Check if JSON is valid
markdown: markdownlint # Optional: if you use README.md or docs
cache-prompts: true # skip resending static context
cache-keepalive-pings: 2 # keep cache warm (pingsX5min) = uptime to keep cache alive
max-chat-history-tokens: 4096 # summarise long sessions
map-tokens: 1024 # small but useful repo overview
map-refresh: auto # refresh when tracked files change
auto-commits: true # commit after each successful change
dirty-commits: true # commit even with unstaged changes
system-message: >-
You are Aider, a pair-programming AI: think step-by-step, ask questions, propose designs,
output unified diffs per style guides, include tests and CLI commands.
Never expose secrets, and maximize correctness and maintainability.
commit-prompt: 'Aider-AI🧙: {{summary}} ({{files}})'
read: [CONVENTIONS.md] # optional project style guide
pretty: true # colorized terminal output
stream: false # stream tokens for snappy feedback
suggest-shell-commands: true # propose handy CLI snippets
dark-mode: false
watch-files: true # watch all files and look for any AI coding instructions text
user-input-color: "#9bd255" # your typed commands appear in green
assistant-output-color: "#37e1e6" # AI responses in cyan for clarity
tool-output-color: "#ff3ee2" # normal tool messages in magenta
tool-error-color: "#f52828" # errors in bright red to stand out
tool-warning-color: "#f58528" # warnings in bright orange
completion-menu-color: "white" # selection text in white for contrast
completion-menu-bg-color: "#02009e" # background in blue to keep the cool tone
code-theme: "one-dark" # popular dark theme with purples & blues for code
analytics-disable: true
```
**Linux:**
```bash
pip install flake8
npm install -g markdownlint-cli
sudo apt install yamllint shellcheck jq
```
**Windows (via Chocolatey):**
```bash
pip install flake8
pip install yamllint
npm install -g markdownlint-cli
choco install shellcheck jq
```
```bash
pip install pytest-playwright
playwright install --with-deps chromium
```
```bash
aider
```
**In-Session Slash Commands:**
**CLI Flags (on launch):**
| Mode | Description | Best Use Case | Token Cost |
|------|-------------|---------------|------------|
| `diff` | Classic minimal diffs | Small edits, bug fixes | 🔹 Lowest |
| `udiff` | Unified diffs with context | Most day-to-day changes | 🔸 Low |
| `diff-fence` | Unified diffs in markdown fences | Better readability | 🔸 Low-Medium |
| `whole` | Outputs entire files | Full rewrites, file generation | 🔺 Higher |
| `architect` | Full-project planning and design | Large features, refactoring | 🔺🔺 Highest |
A token is roughly 4 characters or 1 word. Adjust `map-tokens` based on project size:
| Project Size | Recommended `map-tokens` |
|--------------|--------------------------|
| Tiny (1–3 files) | 512–1024 |
| Small (5–10 files) | 2048–4096 |
| Medium (10–25 files) | 4096–8192 |
| Large (25+ files) | 8192–16384+ |
Uncomment these lines in your config for automatic architect mode:
```yaml
architect: true # off by default; toggle via /architect cmd
auto-accept-architect: false # require review when architect mode is used
```
1. Start Aider in your project: `aider`
2. Add files to context: `/add main.py utils.py`
3. Ask Aider to make changes: "Refactor the error handling in main.py"
4. Aider auto-commits with message: `Aider-AI🧙: Refactor error handling (main.py)`
5. Review changes: `/diff main.py`
6. Undo if needed: `/undo`
From: trinib/Awesome-Mint-Setup (Awesome setup for Linux Mint Cinnamon/Ubuntu)
Leave a review
No reviews yet. Be the first to review this skill!
# Download SKILL.md from killerskills.ai/api/skills/universal-aider-configuration/raw