Calling OmniRoute API Manually

Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!

from openai import OpenAI

Example of Calling OmniRoute Using the Python Library openai¶

client = OpenAI(
    base_url="http://localhost:20128/v1",
    api_key="YOUR_OMNIROUTE_API_KEY" # Generated in your local OmniRoute dashboard
)

response = client.chat.completions.create(
    model="auto/best-coding", # OmniRoute handles the routing behind the scenes
    messages=[
        {
            "role": "user",
            "content": "Write a Python function to reverse a string. Output the code ONLY."
        }
    ]
)

response

ChatCompletion(id='router-27eac306f9725fd458c970d837a083b6', choices=[Choice(finish_reason='stop', index=0, logprobs=None, message=ChatCompletionMessage(content='def reverse_string(s):\n    return s[::-1]', refusal=None, role='assistant', annotations=None, audio=None, function_call=None, tool_calls=None, reasoning_content='We need to output Python code for reversing a string. The instruction says "Write a Python function to reverse a string. Output the code ONLY." So we just give the function definition. Probably simple: def reverse_string(s): return s[::-1]. Or using reversed. We\'ll output code only, no explanation. Ensure it\'s a function. Let\'s do:\n\ndef reverse_string(s):\n    return s[::-1]\n\nThat\'s it.'))], created=1786923527, model='deepseek-v4-flash-free', object='chat.completion', moderation=None, service_tier=None, system_fingerprint=None, usage=CompletionUsage(completion_tokens=101, prompt_tokens=97, total_tokens=198, completion_tokens_details=None, prompt_tokens_details=PromptTokensDetails(audio_tokens=None, cache_write_tokens=None, cached_tokens=None)))

dir(response)

['__abstractmethods__',
 '__annotate_func__',
 '__annotations__',
 '__annotations_cache__',
 '__class__',
 '__class_getitem__',
 '__class_vars__',
 '__copy__',
 '__deepcopy__',
 '__delattr__',
 '__dict__',
 '__dir__',
 '__doc__',
 '__eq__',
 '__fields__',
 '__fields_set__',
 '__firstlineno__',
 '__format__',
 '__ge__',
 '__get_pydantic_core_schema__',
 '__get_pydantic_json_schema__',
 '__getattr__',
 '__getattribute__',
 '__getstate__',
 '__gt__',
 '__hash__',
 '__init__',
 '__init_subclass__',
 '__iter__',
 '__le__',
 '__lt__',
 '__module__',
 '__ne__',
 '__new__',
 '__pretty__',
 '__private_attributes__',
 '__pydantic_complete__',
 '__pydantic_computed_fields__',
 '__pydantic_core_schema__',
 '__pydantic_custom_init__',
 '__pydantic_decorators__',
 '__pydantic_extra__',
 '__pydantic_fields__',
 '__pydantic_fields_set__',
 '__pydantic_generic_metadata__',
 '__pydantic_init_subclass__',
 '__pydantic_on_complete__',
 '__pydantic_parent_namespace__',
 '__pydantic_post_init__',
 '__pydantic_private__',
 '__pydantic_root_model__',
 '__pydantic_serializer__',
 '__pydantic_setattr_handlers__',
 '__pydantic_validator__',
 '__reduce__',
 '__reduce_ex__',
 '__replace__',
 '__repr__',
 '__repr_args__',
 '__repr_name__',
 '__repr_recursion__',
 '__repr_str__',
 '__rich_repr__',
 '__setattr__',
 '__setstate__',
 '__signature__',
 '__sizeof__',
 '__slots__',
 '__static_attributes__',
 '__str__',
 '__subclasshook__',
 '__weakref__',
 '_abc_impl',
 '_calculate_keys',
 '_copy_and_set_values',
 '_get_value',
 '_iter',
 '_request_id',
 '_setattr_handler',
 'choices',
 'construct',
 'copy',
 'created',
 'dict',
 'from_orm',
 'id',
 'json',
 'model',
 'model_computed_fields',
 'model_config',
 'model_construct',
 'model_copy',
 'model_dump',
 'model_dump_json',
 'model_extra',
 'model_fields',
 'model_fields_set',
 'model_json_schema',
 'model_parametrized_name',
 'model_post_init',
 'model_rebuild',
 'model_validate',
 'model_validate_json',
 'model_validate_strings',
 'moderation',
 'object',
 'parse_file',
 'parse_obj',
 'parse_raw',
 'schema',
 'schema_json',
 'service_tier',
 'system_fingerprint',
 'to_dict',
 'to_json',
 'update_forward_refs',
 'usage',
 'validate']

response.choices

[Choice(finish_reason='stop', index=0, logprobs=None, message=ChatCompletionMessage(content='def reverse_string(s):\n    return s[::-1]', refusal=None, role='assistant', annotations=None, audio=None, function_call=None, tool_calls=None, reasoning_content='We need to output Python code for reversing a string. The instruction says "Write a Python function to reverse a string. Output the code ONLY." So we just give the function definition. Probably simple: def reverse_string(s): return s[::-1]. Or using reversed. We\'ll output code only, no explanation. Ensure it\'s a function. Let\'s do:\n\ndef reverse_string(s):\n    return s[::-1]\n\nThat\'s it.'))]

response.choices[0]

Choice(finish_reason='stop', index=0, logprobs=None, message=ChatCompletionMessage(content='def reverse_string(s):\n    return s[::-1]', refusal=None, role='assistant', annotations=None, audio=None, function_call=None, tool_calls=None, reasoning_content='We need to output Python code for reversing a string. The instruction says "Write a Python function to reverse a string. Output the code ONLY." So we just give the function definition. Probably simple: def reverse_string(s): return s[::-1]. Or using reversed. We\'ll output code only, no explanation. Ensure it\'s a function. Let\'s do:\n\ndef reverse_string(s):\n    return s[::-1]\n\nThat\'s it.'))

print(response.choices[0].message.content)

def reverse_string(s):
    return s[::-1]

help(response.construct)

Help on method construct in module openai._models:

construct(_fields_set: set[str] | None = None, **values: object) -> ModelT class method of openai.types.chat.chat_completion.ChatCompletion
    # Override the 'construct' method in a way that supports recursive parsing without validation.
    # Based on https://github.com/samuelcolvin/pydantic/issues/1168#issuecomment-817742836.

The Model That the Requested Routed to¶

response.model

'deepseek-v4-flash-free'