Aider Multi-Provider Configuration

A comprehensive Aider configuration that sets up multiple LLM providers with flexible model routing for different task types.

Features

**Multi-Provider API Keys**: Pre-configured slots for 12+ LLM providers (OpenAI, Anthropic, Gemini, DeepSeek, Codestral, Mistral, GitHub, Cohere, xAI, Groq, Together, OpenRouter)

**Three-Tier Model Routing**:

- Main model: `o3-mini` (reasoning-focused)

- Editor model: `claude-3-7-sonnet` (code editing)

- Weak model: `codestral-latest` (fast operations)

**Architect Mode**: Enabled for high-level design tasks

**Git Integration**: Auto-commits with custom attribution settings

**Performance Optimizations**: Prompt caching, subtree-only mode, timeout tuning

**Developer Experience**: Vim bindings, pretty output, multiline support

Configuration

The configuration uses the `.aider.conf.yaml` format and should be placed in your project root or home directory.

Model Selection Strategy

The file includes commented alternatives for each tier:

**Main Model** (complex reasoning):

Active: `o3-mini`

Alternatives: `o1`, `deepseek-reasoner`, `claude-3-7-sonnet`, `gemini-2.0-pro-exp`

**Editor Model** (code editing):

Active: `claude-3-7-sonnet`

Alternatives: `deepseek-chat`, `gpt-4o`, `gemini-2.0-flash`, `codestral-latest`

**Weak Model** (fast operations):

Active: `codestral-latest`

Alternatives: `gpt-4o-mini`, `mistral-large-latest`

API Key Configuration

Replace placeholder values with your actual API keys:

```yaml

api-key:

- openai=sk-...

- anthropic=sk-ant-...

- gemini=...

- deepseek=...

- codestral=...

- mistral=...

- github=ghp_...

- cohere=...

- xai=...

- groq=...

- together=...

- openrouter=...

```

Alternative API Bases

Uncomment one of these to use alternative OpenAI-compatible endpoints:

```yaml

openai-api-base: http://localhost:8080/v1 # Local inference

openai-api-base: https://glhf.chat/api/openai/v1

openai-api-base: https://api.kluster.ai/v1

openai-api-base: https://api.hyperbolic.xyz/v1

```

Key Settings Explained

| Setting | Value | Purpose |

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

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

| `yes-always` | `true` | Auto-approve suggestions (use with caution) |

| `auto-commits` | `true` | Automatically commit changes |

| `show-diffs` | `false` | Hide diffs in output (cleaner logs) |

| `map-tokens` | `2048` | Context map size |

| `subtree-only` | `true` | Only track relevant files |

| `cache-prompts` | `true` | Enable prompt caching for speed |

| `cache-keepalive-pings` | `300` | Keep cache alive for 5 minutes |

| `timeout` | `301` | Request timeout (5 minutes) |

| `vim` | `true` | Enable Vim keybindings |

| `multiline` | `true` | Support multi-line input |

Optional Features

Automated Testing

Uncomment to enable test-driven development workflow:

```yaml

test-cmd: go test ./... -v -race -cover -timeout 10s

auto-test: true

```

Additional Context

Uncomment to include system prompts or design docs:

```yaml

read:

- SYSTEM_PROMPT_PROGRAMMING.xml

- PRELIMINARY_DESIGN.md

```

Usage

1. **Install Aider**:

```bash

pip install aider-chat

```

2. **Create config file**:

```bash

# Project-specific

touch .aider.conf.yaml

# Or global

touch ~/.aider.conf.yaml

```

3. **Add your API keys** to the `api-key:` section

4. **Run Aider**:

```bash

aider

```

5. **Switch models on-the-fly** using commented alternatives:

- Edit `.aider.conf.yaml`

- Uncomment desired model

- Restart Aider

Best Practices

**Start with weak model** for exploration, upgrade to main model for complex tasks

**Use architect mode** for new features, disable for bug fixes

**Enable auto-test** when working on codebases with good test coverage

**Set reasonable timeouts** based on model speed (reasoning models need longer)

**Use subtree-only** for large repositories to reduce token costs

Security Notes

Never commit this file with real API keys to public repositories

Use environment variables or secret management tools for production

Consider using `.gitignore` entry: `.aider.conf.yaml`

Constraints

Requires Aider v0.50.0 or later for full compatibility

Some providers require specific model name formats (e.g. `provider/model-name`)

Prompt caching requires provider support (OpenAI, Anthropic)

Architect mode works best with capable reasoning models

Aider Multi-Provider Configuration

Aider Multi-Provider Configuration

Features

Configuration

Model Selection Strategy

API Key Configuration

Alternative API Bases

openai-api-base: http://localhost:8080/v1 # Local inference

openai-api-base: https://glhf.chat/api/openai/v1

openai-api-base: https://api.kluster.ai/v1

openai-api-base: https://api.hyperbolic.xyz/v1

Key Settings Explained

Optional Features

Automated Testing

Additional Context

Usage

Best Practices

Security Notes

Constraints

Reviews (0)