IntelGraph AI Symphony Aider Config

Agent-centric Aider configuration that enables role-based model switching through a LiteLLM proxy. Routes between local (Qwen) and cloud models (Gemini, xAI) based on task requirements.

What This Does

This configuration sets up Aider with:

Instructions

1. Prerequisites

Ensure you have:

2. Configuration Setup

Create a `.aider.conf.yml` file in your project root with these settings:

```yaml

Core routing - default to fast local Qwen coder

model: openai/local/llama

openai-api-base: http://127.0.0.1:4000/v1

openai-api-key: sk-anything

Quality & performance

pretty: true

stream: true

auto-commits: false

git: true

edit-format: diff

show-diffs: true

max-chat-history-tokens: 8192

M2 16GB optimization

cache-prompts: true

restore-chat-history: false

File handling

encoding: utf-8

ignore-globs: |

*.log

*.tmp

**/node_modules/**

**/dist/**

**/build/**

**/.vite/**

**/.next/**

**/coverage/**

rag/index/*.duckdb*

**/*.pyc

**/__pycache__/**

```

3. Agent Role Quick-Switch

Override the model for specific agent roles:

**Architect/Coder (guy)**

```bash

aider --model openai/local/llama

```

**Security Review (aegis)**

```bash

aider --model openai/local/llama-cpu

```

**CI/PR Tasks (hermes)**

```bash

aider --model openai/local/llama-small

```

**Research (elara)**

```bash

export GOOGLE_API_KEY=your_key

aider --model openai/gemini/1.5-pro

```

**Power Burst (high-capacity cloud)**

```bash

export XAI_API_KEY=your_key

aider --model openai/xai/grok-code-fast-1

```

4. Usage Patterns

**Standard coding session:**

```bash

aider src/main.py src/utils.py

Uses default Qwen local model

```

**Security review of changes:**

```bash

aider --model openai/local/llama-cpu src/auth.py

Lower resource usage, focused on security patterns

```

**Quick PR fixes:**

```bash

aider --model openai/local/llama-small --message "Fix lint errors"

Fast, lightweight model for simple tasks

```

**Research and architecture:**

```bash

aider --model openai/gemini/1.5-pro --architect-mode

Cloud model with larger context for design decisions

```

5. Key Configuration Decisions

6. File Ignoring Strategy

Excluded patterns:

7. LiteLLM Proxy Setup

Your LiteLLM config should map the model routes:

```yaml

model_list:

- model_name: local/llama

litellm_params:

model: ollama/qwen2.5-coder

- model_name: local/llama-cpu

litellm_params:

model: ollama/qwen2.5-coder:7b

- model_name: local/llama-small

litellm_params:

model: ollama/qwen2.5-coder:3b

```

Important Notes

Examples

**Multi-file refactoring with architect mode:**

```bash

aider --model openai/local/llama src/**/*.py --architect-mode

```

**Security-focused code review:**

```bash

aider --model openai/local/llama-cpu --read-only src/auth.py src/middleware.py

```

**Bulk PR fixes with small model:**

```bash

aider --model openai/local/llama-small --yes --message "Apply formatting across codebase"

```