yachay 0.1.1

Python SDK for Yachay — Condor Models inference API (OpenAI-compatible)

Yachay Python SDK

Python client for Yachay — Condor Models inference API. Designed to feel like the OpenAI SDK so migrating existing code takes minutes.

Install

pip install yachay

Requires Python 3.9+.

Quickstart

import yachay

client = yachay.Client()  # reads YACHAY_API_KEY from environment

response = client.chat.completions.create(
    model="llama-3.1-8b",
    messages=[{"role": "user", "content": "Explain gradient descent in one paragraph."}],
)
print(response.choices[0].message.content)

Authentication

Create a customer API key in the Yachay dashboard. Keys are prefixed yck_.

Recommended — environment variable:

export YACHAY_API_KEY="yck_your_key_here"

Or pass directly:

client = yachay.Client(api_key="yck_your_key_here")

Available models

Two models are currently available for inference:

Model ID	Description
llama-3.1-8b	Meta Llama 3.1 8B — fast general-purpose chat
deepseek-r1-distill-llama-8b	DeepSeek R1 distilled into Llama 8B — strong reasoning

To list available models programmatically:

models = client.models.list()
for model in models.data:
    print(model.id)

Chat completions

response = client.chat.completions.create(
    model="deepseek-r1-distill-llama-8b",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "What is the capital of Peru?"},
    ],
    max_tokens=256,
    temperature=0.7,
)

# Access the generated text
print(response.choices[0].message.content)

# Access token usage
print(f"Tokens used: {response.usage.total_tokens}")

Context manager

with yachay.Client() as client:
    response = client.chat.completions.create(
        model="llama-3.1-8b",
        messages=[{"role": "user", "content": "Hello!"}],
    )
    print(response.choices[0].message.content)
# HTTP connection pool is closed automatically

Migrating from OpenAI

The client mirrors the OpenAI SDK's nested-namespace pattern. Replace:

# Before
import openai
client = openai.OpenAI(api_key="sk-...")
response = client.chat.completions.create(model="gpt-4o", messages=[...])

with:

# After
import yachay
client = yachay.Client(api_key="yck_...")
response = client.chat.completions.create(model="llama-3.1-8b", messages=[...])

Response objects share the same shape: .choices[0].message.content, .usage.total_tokens, etc.

Error handling

from yachay import AuthError, RateLimitError, APIError

try:
    response = client.chat.completions.create(...)
except AuthError:
    print("Invalid API key")
except RateLimitError:
    print("Slow down — you are being rate-limited")
except APIError as e:
    print(f"API error {e.status}: {e.body}")

All errors are subclasses of yachay.YachayError.

Custom base URL and timeout

client = yachay.Client(
    api_key="yck_...",
    base_url="https://api.yachaymodels.com/v1",  # default; override for testing
    timeout=30.0,  # seconds
)

What v1 does NOT include

• Streaming — stream=True is not yet supported. Submit a feature request at yachaymodels.com/support.
• Fine-tuning job management — submitting training jobs, downloading adapters, and managing the fine-tuning pipeline are not available through this SDK in v1. Use the Yachay web dashboard for job management.
• Embeddings — not available in v1.

Documentation

Full API reference and guides at yachaymodels.com/docs.

License

MIT — see LICENSE for details.