Aider

Aider is an open-source AI pair programmer for the terminal, with tight git integration. It connects to any OpenAI-compatible endpoint via environment variables — GreenPT works out of the box using the openai/ model prefix.

Prerequisites

• Aider installed (python -m pip install aider-install && aider-install)
• A GreenPT API key from your GreenPT dashboard

Quick start

export OPENAI_API_BASE=https://api.greenpt.ai/v1
export OPENAI_API_KEY=your-greenpt-key
aider --model openai/glm-5.2

The openai/ prefix tells Aider (via LiteLLM) to treat GreenPT as an OpenAI-compatible endpoint. The part after the prefix must exactly match a model ID from GreenPT's /v1/models endpoint.

Available models

Persistent configuration

Instead of flags and exports, put your defaults in ~/.aider.conf.yml (global) or .aider.conf.yml in the repo root (project — takes precedence):

openai-api-base: https://api.greenpt.ai/v1
model: openai/glm-5.2

Keep the API key out of the YAML file — set it in your shell profile or a .env file in the repo root:

# .env
OPENAI_API_BASE=https://api.greenpt.ai/v1
OPENAI_API_KEY=your-greenpt-key

Model metadata (context window & pricing)

Aider warns about unknown models because GreenPT's model IDs aren't in LiteLLM's built-in database. Register them in ~/.aider.model.metadata.json so Aider knows each model's context window and shows accurate costs in-session:

{
  "openai/glm-5.2": {
    "max_input_tokens": 1000000,
    "max_output_tokens": 131072,
    "input_cost_per_token": 0.0000011,
    "output_cost_per_token": 0.0000044,
    "litellm_provider": "openai",
    "mode": "chat"
  },
  "openai/kimi-k3": {
    "max_input_tokens": 1000000,
    "input_cost_per_token": 0.0000033,
    "output_cost_per_token": 0.0000165,
    "litellm_provider": "openai",
    "mode": "chat"
  },
  "openai/deepseek-v4-flash-0731": {
    "max_input_tokens": 1000000,
    "input_cost_per_token": 0.00000014,
    "output_cost_per_token": 0.00000035,
    "litellm_provider": "openai",
    "mode": "chat"
  }
}

Costs are per single token (price per million ÷ 1,000,000). max_output_tokens is only set for glm-5.2 (131,072 confirmed); add it for the other models once known.

Recommended edit format

For strong coding models like these, Aider's diff edit format is usually the most token-efficient. Set it per model in ~/.aider.model.settings.yml:

- name: openai/glm-5.2
  edit_format: diff
  use_repo_map: true
- name: openai/kimi-k3
  edit_format: diff
  use_repo_map: true
- name: openai/deepseek-v4-flash-0731
  edit_format: diff
  use_repo_map: true

If a model produces malformed diffs, fall back to edit_format: whole.

Verify the connection

curl https://api.greenpt.ai/v1/models \
  -H "Authorization: Bearer $OPENAI_API_KEY"

You should get back a list including the model IDs above.

Troubleshooting