Universal OpenAI-compatible Client with structured output guarantees and provider abstraction.
pip install vidaisdkimport os
from vidai import Vidai
from pydantic import BaseModel
# Optional: Alias as OpenAI for drop-in compatibility if desired
# from vidai import Vidai as OpenAI
class User(BaseModel):
name: str
age: int
client = Vidai(
api_key=os.environ.get("OPENAI_API_KEY"),
# It sends requests to OpenAI by default,
# but strictly validates/repairs responses locally
)
# If you want to use a custom base_url or proxy, you can do this:
# client = Vidai(
# api_key="your-key",
# base_url="https://your-proxy.com/v1"
# )
response = client.chat.completions.create(
model="gpt-4",
messages=[{"role": "user", "content": "Create a user named John, age 25"}],
response_format=User
)
user = response.choices[0].message.parsed
print(user.name) # "John"
print(user.age) # 25Vidai acts as a smart facade when connected to a Vidai Server proxy. It automatically discovers available models and applies provider-specific optimizations (like Claude tool-use) without any client-side configuration changes.
- Set your environment:
export VIDAI_PROVIDER=vidai
export VIDAI_BASE_URL=http://localhost:8000/v1
export VIDAI_SERVER_API_KEY=sk-...- Use the client as usual:
client = Vidai() # Automatically detects 'vidai' provider from env
# Automatically routes to Anthropic, DeepSeek, Google based on model name
response = client.chat.completions.create(model="claude-3-haiku", ...)- Drop-in Compatibility: 100% compatible with OpenAI SDK API (Completions & Responses)
- Responses API: Type-safe, object-based access normalized across all providers.
- Structured Output Guarantees: Automatic JSON repair and validation
- Performance Tracking: Monitor SDK overhead and repair times
- Provider Agnostic: Works with any AI provider through your proxy
- Developer Friendly: Clear error messages and debugging support
Full documentation is available at https://Vidai.uk/docs/sdk/python/.
See the GitHub Repository for the latest updates.
Pair with Vidai.Server
The SDKs work standalone with any OpenAI-compatible endpoint, but represent a complete, enterprise-grade stack when paired with Vidai.Server. This Rust-based Enterprise Gateway handles 6,000+ RPS on a single Intel core with <20ms p95 latency, meeting the most demanding production requirements.
| SDK Handles | Server Handles (Enterprise Gateway) |
|---|---|
| ✨ Structured Output | 📈 Rate Limiting & Quotas |
| 🛠️ Auto JSON Repair | 💰 Cost Tracking & Budgets |
| ✅ Type Safety (Zod/Pydantic) | 🔑 API Key Management |
Apache-2.0