Run LiteLLM on Mingles Router
LiteLLM is the layer half the ecosystem already sits on. Adding us is a model prefix and a base URL, in code or in the proxy config.
Pick your model and paste your key — every command and value on this page updates live.
Endpoint: https://router.mingles.ai/v1 · key stays in your browser ·
no key entered? commands show <your-key>.
Wire it up
1. As a library
The openai/ prefix is what tells LiteLLM to use the generic OpenAI route rather than guessing the provider from the model name:
from litellm import completion
resp = completion(
model="openai/deepseek-ai/DeepSeek-V4-Flash-0731",
api_base="https://router.mingles.ai/v1",
api_key="<your-key>",
messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)2. As a proxy
Put us in config.yaml and every app pointed at the proxy gets it:
model_list:
- model_name: cheap
litellm_params:
model: openai/deepseek-ai/DeepSeek-V4-Flash-0731
api_base: https://router.mingles.ai/v1
api_key: <your-key>3. Use it as a fallback, or as the default
A common shape: cheap model first, frontier model only when the first fails.
router_settings:
fallbacks: [{"cheap": ["expensive"]}]Exact values
| model | openai/deepseek-ai/DeepSeek-V4-Flash-0731 |
| api_base | https://router.mingles.ai/v1 |
| api_key | <your-key> |
Switch model
Anything built on LiteLLM inherits this — Aider, CrewAI and a long tail of tools all resolve models through it.
Before you file a bug: read the limits
Reasoning models spend the output budget on internal thinking, output is capped at 8192 tokens, there is no KV cache, and there are no built-in web tools. Most “it broke on Mingles Router” reports are one of these. See Model limits & behavior →
Frequently asked
LiteLLM says it does not know the context window for this model. +
Expected — it has no metadata for third-party ids. Either ignore the warning or declare model_info with max_input_tokens in the proxy config.
Do I need the openai/ prefix? +
Yes, when the model id is not one LiteLLM recognises. Without a prefix it tries to infer the provider from the name and picks the wrong route.