I would like to use Aider with models hosted on glhf.chat.

I use this configuration:

~/aider.conf.yml

openai-api-base: /api/openai/v1
openai-api-key: glhf_MY_SECRET_API_KEY
model-settings-file: ~/.aider.model.settings.yml
model: deepseek-ai/DeepSeek-R1
weak-model: deepseek-ai/DeepSeek-V3
editor-model: Qwen/Qwen2.5-Coder-32B-Instruct

~/.aider.model.settings.yml

- name: deepseek-ai/DeepSeek-V3
  edit_format: diff
  use_repo_map: true
  reminder: sys
  examples_as_sys_msg: true
  extra_params:
    max_tokens: 8192
  caches_by_default: true

- name: deepseek-ai/DeepSeek-R1
  edit_format: diff
  weak_model_name: deepseek-ai/DeepSeek-V3
  use_repo_map: true
  examples_as_sys_msg: true
  extra_params:
    max_tokens: 8192
    include_reasoning: true
  caches_by_default: true
  editor_model_name: deepseek-ai/DeepSeek-V3
  editor_edit_format: editor-diff

- name: Qwen/Qwen2.5-Coder-32B-Instruct
  edit_format: diff
  weak_model_name: Qwen/Qwen2.5-Coder-32B-Instruct
  use_repo_map: true
  editor_model_name: Qwen/Qwen2.5-Coder-32B-Instruct
  editor_edit_format: editor-diff

But when I start aider it says:

───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Warning for deepseek-ai/DeepSeek-R1: Unknown context window size and costs, using sane defaults.
Did you mean one of these?
- deepseek/deepseek-chat
- deepseek/deepseek-coder
Warning for deepseek-ai/DeepSeek-V3: Unknown context window size and costs, using sane defaults.
Did you mean one of these?
- deepseek/deepseek-chat
Warning for Qwen/Qwen2.5-Coder-32B-Instruct: Unknown context window size and costs, using sane defaults.
Did you mean one of these?
- openrouter/qwen/qwen-2.5-coder-32b-instruct
You can skip this check with --no-show-model-warnings

Why? How can I get rid of those warnings?

I would like to use Aider with models hosted on glhf.chat.

I use this configuration:

~/aider.conf.yml

openai-api-base: https://glhf.chat/api/openai/v1
openai-api-key: glhf_MY_SECRET_API_KEY
model-settings-file: ~/.aider.model.settings.yml
model: deepseek-ai/DeepSeek-R1
weak-model: deepseek-ai/DeepSeek-V3
editor-model: Qwen/Qwen2.5-Coder-32B-Instruct

~/.aider.model.settings.yml

- name: deepseek-ai/DeepSeek-V3
  edit_format: diff
  use_repo_map: true
  reminder: sys
  examples_as_sys_msg: true
  extra_params:
    max_tokens: 8192
  caches_by_default: true

- name: deepseek-ai/DeepSeek-R1
  edit_format: diff
  weak_model_name: deepseek-ai/DeepSeek-V3
  use_repo_map: true
  examples_as_sys_msg: true
  extra_params:
    max_tokens: 8192
    include_reasoning: true
  caches_by_default: true
  editor_model_name: deepseek-ai/DeepSeek-V3
  editor_edit_format: editor-diff

- name: Qwen/Qwen2.5-Coder-32B-Instruct
  edit_format: diff
  weak_model_name: Qwen/Qwen2.5-Coder-32B-Instruct
  use_repo_map: true
  editor_model_name: Qwen/Qwen2.5-Coder-32B-Instruct
  editor_edit_format: editor-diff

But when I start aider it says:

───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Warning for deepseek-ai/DeepSeek-R1: Unknown context window size and costs, using sane defaults.
Did you mean one of these?
- deepseek/deepseek-chat
- deepseek/deepseek-coder
Warning for deepseek-ai/DeepSeek-V3: Unknown context window size and costs, using sane defaults.
Did you mean one of these?
- deepseek/deepseek-chat
Warning for Qwen/Qwen2.5-Coder-32B-Instruct: Unknown context window size and costs, using sane defaults.
Did you mean one of these?
- openrouter/qwen/qwen-2.5-coder-32b-instruct
You can skip this check with --no-show-model-warnings

Why? How can I get rid of those warnings?

• configuration
• large-language-model
• aider

1 Answer 1

Billy (one of the glhf.chat founders) here! Thanks for using our app!

To remove the warnings, you can tell Aider about the models you're using with a .aider.model.metadata.json file. See Context window size and token costs

Here's one that should work for your models:

{
  "openai/hf:deepseek-ai/DeepSeek-R1": {
    "max_tokens": 8192,
    "max_input_tokens": 128000,
    "max_output_tokens": 8192,
    "input_cost_per_token": 0.000003,
    "output_cost_per_token": 0.000007,
    "mode": "chat",
  },
  "openai/hf:deepseek-ai/DeepSeek-V3": {
    "max_tokens": 8192,
    "max_input_tokens": 128000,
    "max_output_tokens": 8192,
    "input_cost_per_token": 0.00000125,
    "output_cost_per_token": 0.00000125,
    "mode": "chat",
  },
  "openai/hf:Qwen/Qwen2.5-Coder-32B-Instruct": {
    "max_tokens": 8192,
    "max_input_tokens": 32000,
    "max_output_tokens": 8192,
    "input_cost_per_token": 0.00000125,
    "output_cost_per_token": 0.00000125,
    "mode": "chat",
  },
}

(Aider also adds openai/ to specify OpenAI compatible providers so we also do that in the configs.)

Note that for glhf.chat's API, we need you need to add an hf: (HuggingFace) prefix for model names:

// .aider.model.settings.yml

- name: openai/hf:deepseek-ai/DeepSeek-V3
  edit_format: diff
  use_repo_map: true
  reminder: sys
  examples_as_sys_msg: true
  caches_by_default: true

- name: openai/hf:deepseek-ai/DeepSeek-R1
  edit_format: diff
  weak_model_name: openai/hf:deepseek-ai/DeepSeek-V3
  use_repo_map: true
  examples_as_sys_msg: true
  extra_params:
    include_reasoning: true
  caches_by_default: true
  editor_model_name: openai/hf:deepseek-ai/DeepSeek-V3
  editor_edit_format: editor-diff

- name: openai/hf:Qwen/Qwen2.5-Coder-32B-Instruct
  edit_format: diff
  weak_model_name: openai/hf:Qwen/Qwen2.5-Coder-32B-Instruct
  use_repo_map: true
  editor_model_name: openai/hf:Qwen/Qwen2.5-Coder-32B-Instruct
  editor_edit_format: editor-diff

// .aider.conf.yml

openai-api-base: https://api.glhf.chat/v1
openai-api-key: glhf_MY_SECRET_API_KEY
model-settings-file: ~/.aider.model.settings.yml
model: openai/hf:deepseek-ai/DeepSeek-R1
weak-model: openai/hf:deepseek-ai/DeepSeek-V3
editor-model: openai/hf:Qwen/Qwen2.5-Coder-32B-Instruct

Hope that helps!