Universal AI Provider SDK (uai-sdk-python)
A universal, modular AI infrastructure layer for Python that abstracts multiple Chinese LLM providers behind a single, stable API with an opt-in middleware architecture.
Installation
🚧 Under Active Development — This SDK has NOT been released on PyPI yet. We're targeting our first release for Q3 2026.
Local Install (Development Mode)
# Clone the repository
git clone https://github.com/uai-sdk/uai-sdk-python.git
cd uai-sdk-python
# Install in development mode
pip install -e .
Once published to PyPI, the command will be:
pip install uai-sdk
Quick Start
from uai import UniversalAI
# One client per provider; the SDK reads DEEPSEEK_API_KEY from the
# environment if api_key is omitted.
client = UniversalAI(provider="deepseek", model="deepseek-chat")
# Chat
result = client.chat(
messages=[{"role": "user", "content": "Translate to English: 你好"}],
)
print(result.content)
# Streaming
for chunk in client.chat(
messages=[{"role": "user", "content": "Tell me a story..."}],
stream=True,
):
print(chunk.content, end="", flush=True)
# Structured output (validated by the provider adapter; see docs/structured_output.md)
from pydantic import BaseModel
class KeyPoints(BaseModel):
points: list[str]
result = client.chat(
messages=[{"role": "user", "content": "Extract the key points"}],
output_schema=KeyPoints,
)
print(result.parsed)
Features
- ✅ Unified API across multiple LLM providers
- ✅ Chat, streaming, tool-calling, structured outputs
- ✅ Embeddings (
client.embed) and rerank (client.rerank) via provider adapters - ✅ Vision via chat content blocks (image_url) on vision-capable models
- ✅ Provider adapters with strict capability enforcement
- ✅ Opt-in middleware — retry, cache, circuit breaker, logging, tracing, metrics (
client.use(...)) - ✅ Benchmark CLI covering all chat models (
uai benchmark) - ✅ KPI regression suite — <5 ms SDK overhead, ≥1,000 req/min, memory caps (see docs/performance.md)
- ✅ Security-first design (no secret logging, input validation)
- ✅ Extensible — easily add new providers via the PDK
- ✅ Prometheus-style metrics — in-process
MetricsRegistrywithrender() - ⏳ Audio / TTS / transcription — not yet implemented (deferred)
Supported Providers
| Provider | Chat | Streaming | Tools | Vision | Embeddings | Rerank | Audio |
|---|---|---|---|---|---|---|---|
| DeepSeek | ✅ | ✅ | ✅ | ❌ | ✅ | ❌ | ❌ |
| Qwen | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
| GLM | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
| Kimi | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ |
| StepFun | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
| Doubao | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
| MiniMax | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
| Hunyuan | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
Documentation
- Architecture — middleware pipeline, adapter contract, UnifiedRequest lifecycle
- Chat — conversational completions, history management, system prompts
- Streaming — SSE handling and TTFT
- Tools — function calling with OpenAI-style tool definitions
- Embeddings — text embedding operations via adapters
- Vision — multimodal image interpretation via chat content blocks
- Rerank — document ranking (Qwen, GLM)
- Structured Output — schema validation & parsing
- Middleware — creating and composing interceptors (retry, cache, circuit breaker, logging, tracing, metrics)
- Benchmark — offline benchmarking CLI covering all models
- Performance — KPI targets & systemic testing (Module 1.6), mock provider server
- Telemetry — Prometheus-style metrics and GenAI tracing spans
- Configuration — env vars, YAML config, API key management
- PDK — Provider Development Kit guide
Contributing
Contributions are welcome! Please read our Contributing Guide first.
Issue Label Policy
All new issues are automatically evaluated and tagged:
| Label | Meaning |
|---|---|
roadmap-aligned |
On our planned roadmap — we intend to build it. |
needs-discussion |
Requires design discussion before we decide. |
won't-implement |
Out of scope for this project's direction. |
help-wanted |
We explicitly welcome PRs for this. |
See .github/LABELS.md for the full label set and conventions.
License
This project is licensed under the Apache License 2.0 — see LICENSE for details.
Metadata
Release files for uai-sdk 0.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| uai_sdk-0.1.0.tar.gz | 65.7 kB | With provenance | Details for uai_sdk-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| uai_sdk-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for uai_sdk-0.1.0-py3-none-any.whl |
Total release size: 165.4 kB
Release files / uai_sdk-0.1.0.tar.gz
| Download URL | uai_sdk-0.1.0.tar.gz |
|---|---|
| Size | 65.7 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
c359c881a61e7e0b7e56675750d4bb13c102739b5212bf69fda712c1d4547414
|
|
BLAKE2b-256 checksum How to use checksums |
eeb6b014fc5f589c489ca014ccd8caa9f109b86d3c0a4fabe75ff86f11e6ddad
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Aug 9, 2026.
Transparency logRelease files / uai_sdk-0.1.0-py3-none-any.whl
| Download URL | uai_sdk-0.1.0-py3-none-any.whl |
|---|---|
| Size | 99.7 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
563f1929f388c5460b884458fb89399908364738d73d20ca6c02a219cd1e93c7
|
|
BLAKE2b-256 checksum How to use checksums |
405931e5bc5cf3e2fc23190d0772dc04b5cd180e20f22df130649f2fec4459c0
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Aug 9, 2026.
Transparency logSecurity for uai-sdk 0.1.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
Both files in this release have been uploaded with provenance metadata.
Attestation for uai_sdk-0.1.0.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Aug 9, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.