llm-taxi 0.7.0

Call LLM as easily as calling a taxi.

llm-taxi

Installation

pip install llm-taxi

Usage

Use as a library

import asyncio

from llm_taxi.conversation import Message, Role
from llm_taxi.factory import embedding, llm


async def main():
    # Non-streaming response
    client = llm(model="openai:gpt-3.5-turbo")
    messages = [
        Message(role=Role.User, content="What is the capital of France?"),
    ]

    response = await client.response(messages)
    print(response)

    # Streaming response
    client = llm(model="mistral:mistral-small")
    messages = [
        Message(role=Role.User, content="Tell me a joke."),
    ]
    response = await client.streaming_response(messages)
    async for chunk in response:
        print(chunk, end="", flush=True)
    print()

    # Embed text
    embedder = embedding("openai:text-embedding-ada-002")
    embeddings = await embedder.embed_text("Hello, world!")
    print(embeddings[:10])

    # Embed texts
    embedder = embedding("mistral:mistral-embed")
    embeddings = await embedder.embed_texts(["Hello, world!"])
    print(embeddings[0][:10])



if __name__ == "__main__":
    asyncio.run(main())

Common parameters

• temperature
• max_tokens
• top_k
• top_p
• stop
• seed
• presence_penalty
• frequency_penalty
• response_format
• tools
• tool_choice

Command line interface

llm-taxi --model openai:gpt-3.5-turbo-16k

See all supported arguments

llm-taxi --help

Supported Providers

Provider	LLM	Embedding
Anthropic	✅
DashScope	✅
DeepInfra	✅
DeepSeek	✅
Google	✅	✅
Groq	✅
Mistral	✅	✅
OpenAI	✅	✅
OpenRouter	✅
Perplexity	✅
Together	✅
BigModel	✅