# UnifiedAI SDK
OpenAI-compatible Python SDK unifying multiple providers (Cerebras, AWS Bedrock) with Solo and Comparison modes, strict models, and built‑in telemetry.
## Highlights
- OpenAI-like API: `UnifiedAI().chat.completions.create(...)` (sync) and `AsyncUnifiedAI` (async)
- Pluggable adapters: Cerebras, Bedrock (extensible)
- Modes: Solo and side‑by‑side Comparison
- Observability: structured logs, Prometheus metrics (SDK), tracing hooks
- Credentials: pass at client construction or use env
## Install
From PyPI (core):
```bash
pip install unifiedai-sdk
```
Optional extras:
```bash
# Cerebras Cloud SDK integration
pip install "unifiedai-sdk[cerebras]"
# HTTP/2 support for httpx
pip install "unifiedai-sdk[http2]"
```
From GitHub (optional):
```bash
pip install git+https://github.com/<your-org-or-user>/<your-repo>.git#subdirectory=cerebras
```
## Usage
### Sync (scripts/CLI)
```python
from unifiedai import UnifiedAI
client = UnifiedAI(
provider="cerebras",
model="llama3",
credentials={"api_key": "csk-..."}, # or set CEREBRAS_KEY in env
)
resp = client.chat.completions.create(
messages=[{"role": "user", "content": "Hello"}]
)
print(resp.choices[0].message["content"])
```
### Async (web backends)
```python
from unifiedai import AsyncUnifiedAI
async with AsyncUnifiedAI(provider="cerebras", model="llama3") as client:
resp = await client.chat.completions.create(
messages=[{"role": "user", "content": "Hello"}]
)
```
### Streaming (async)
```python
async with AsyncUnifiedAI(provider="cerebras", model="llama3") as client:
async for chunk in client.chat.completions.stream(
messages=[{"role": "user", "content": "Stream this"}]
):
print(chunk.delta.get("content", ""), end="")
```
### Comparison (two providers)
```python
from unifiedai import AsyncUnifiedAI
async with AsyncUnifiedAI() as client:
result = await client.compare(
providers=["cerebras", "bedrock"],
model="llama3",
messages=[{"role": "user", "content": "Compare outputs"}],
)
print(result.winner, result.comparative_metrics.speed_difference_ms)
```
## Credentials
- Precedence: per‑provider credentials > global client credentials > environment (SDKConfig).
- Cerebras: set `CEREBRAS_KEY` or pass `credentials={"api_key": "..."}`.
- Bedrock: planned; wire `credentials_by_provider` similarly.
## FastAPI demo (Swagger UI)
```bash
uvicorn apps.chat.backend:app --reload --port 8000
# Swagger UI: http://localhost:8000/docs
```
## Project Structure
- `src/unifiedai/`: SDK implementation (clients, adapters, models, core)
- `examples/`: usage examples (solo, streaming, comparison)
- `apps/chat/`: demo FastAPI backend
- `tests/`: unit tests
## License
MIT
Raw data
{
"_id": null,
"home_page": null,
"name": "unifiedai-sdk",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.9",
"maintainer_email": null,
"keywords": "bedrock, cerebras, comparison, llm, openai-compatible, sdk",
"author": null,
"author_email": "Your Team <team@example.com>",
"download_url": "https://files.pythonhosted.org/packages/a7/5d/b936b5fb3af5fa47caab0935b9346dd997a49c35a2809b2c85f172415067/unifiedai_sdk-1.0.4.tar.gz",
"platform": null,
"description": "# UnifiedAI SDK\n\nOpenAI-compatible Python SDK unifying multiple providers (Cerebras, AWS Bedrock) with Solo and Comparison modes, strict models, and built\u2011in telemetry.\n\n## Highlights\n- OpenAI-like API: `UnifiedAI().chat.completions.create(...)` (sync) and `AsyncUnifiedAI` (async)\n- Pluggable adapters: Cerebras, Bedrock (extensible)\n- Modes: Solo and side\u2011by\u2011side Comparison\n- Observability: structured logs, Prometheus metrics (SDK), tracing hooks\n- Credentials: pass at client construction or use env\n\n## Install\nFrom PyPI (core):\n```bash\npip install unifiedai-sdk\n```\n\nOptional extras:\n```bash\n# Cerebras Cloud SDK integration\npip install \"unifiedai-sdk[cerebras]\"\n\n# HTTP/2 support for httpx\npip install \"unifiedai-sdk[http2]\"\n```\n\nFrom GitHub (optional):\n```bash\npip install git+https://github.com/<your-org-or-user>/<your-repo>.git#subdirectory=cerebras\n```\n\n## Usage\n\n### Sync (scripts/CLI)\n```python\nfrom unifiedai import UnifiedAI\n\nclient = UnifiedAI(\n provider=\"cerebras\",\n model=\"llama3\",\n credentials={\"api_key\": \"csk-...\"}, # or set CEREBRAS_KEY in env\n)\nresp = client.chat.completions.create(\n messages=[{\"role\": \"user\", \"content\": \"Hello\"}]\n)\nprint(resp.choices[0].message[\"content\"])\n```\n\n### Async (web backends)\n```python\nfrom unifiedai import AsyncUnifiedAI\n\nasync with AsyncUnifiedAI(provider=\"cerebras\", model=\"llama3\") as client:\n resp = await client.chat.completions.create(\n messages=[{\"role\": \"user\", \"content\": \"Hello\"}]\n )\n```\n\n### Streaming (async)\n```python\nasync with AsyncUnifiedAI(provider=\"cerebras\", model=\"llama3\") as client:\n async for chunk in client.chat.completions.stream(\n messages=[{\"role\": \"user\", \"content\": \"Stream this\"}]\n ):\n print(chunk.delta.get(\"content\", \"\"), end=\"\")\n```\n\n### Comparison (two providers)\n```python\nfrom unifiedai import AsyncUnifiedAI\n\nasync with AsyncUnifiedAI() as client:\n result = await client.compare(\n providers=[\"cerebras\", \"bedrock\"],\n model=\"llama3\",\n messages=[{\"role\": \"user\", \"content\": \"Compare outputs\"}],\n )\n print(result.winner, result.comparative_metrics.speed_difference_ms)\n```\n\n## Credentials\n- Precedence: per\u2011provider credentials > global client credentials > environment (SDKConfig).\n- Cerebras: set `CEREBRAS_KEY` or pass `credentials={\"api_key\": \"...\"}`.\n- Bedrock: planned; wire `credentials_by_provider` similarly.\n\n## FastAPI demo (Swagger UI)\n```bash\nuvicorn apps.chat.backend:app --reload --port 8000\n# Swagger UI: http://localhost:8000/docs\n```\n\n## Project Structure\n- `src/unifiedai/`: SDK implementation (clients, adapters, models, core)\n- `examples/`: usage examples (solo, streaming, comparison)\n- `apps/chat/`: demo FastAPI backend\n- `tests/`: unit tests\n\n## License\nMIT\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "Unified SDK for multi-provider LLM comparison (Cerebras, AWS Bedrock) with OpenAI-compatible interface.",
"version": "1.0.4",
"project_urls": null,
"split_keywords": [
"bedrock",
" cerebras",
" comparison",
" llm",
" openai-compatible",
" sdk"
],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "b24f92cce0e0c3b10a3caa971c020d8d164ba6fcf2788dfbb2c13e490df588e2",
"md5": "3b20492fdc558eefe4175e3b3ce2cad2",
"sha256": "c7e12d1667cea55b7fd3616c65f3d5819cca906328ae06505b8b6fdb9fb15639"
},
"downloads": -1,
"filename": "unifiedai_sdk-1.0.4-py3-none-any.whl",
"has_sig": false,
"md5_digest": "3b20492fdc558eefe4175e3b3ce2cad2",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.9",
"size": 23693,
"upload_time": "2025-10-13T08:52:52",
"upload_time_iso_8601": "2025-10-13T08:52:52.449693Z",
"url": "https://files.pythonhosted.org/packages/b2/4f/92cce0e0c3b10a3caa971c020d8d164ba6fcf2788dfbb2c13e490df588e2/unifiedai_sdk-1.0.4-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "a75db936b5fb3af5fa47caab0935b9346dd997a49c35a2809b2c85f172415067",
"md5": "b7fda3991a3fd359d217dd9255485775",
"sha256": "2fe2c7a8054c171b7b4e1a0c3a9687a1cd02e3e2f187f7e5c5b5744129b80b60"
},
"downloads": -1,
"filename": "unifiedai_sdk-1.0.4.tar.gz",
"has_sig": false,
"md5_digest": "b7fda3991a3fd359d217dd9255485775",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.9",
"size": 17218,
"upload_time": "2025-10-13T08:52:54",
"upload_time_iso_8601": "2025-10-13T08:52:54.710591Z",
"url": "https://files.pythonhosted.org/packages/a7/5d/b936b5fb3af5fa47caab0935b9346dd997a49c35a2809b2c85f172415067/unifiedai_sdk-1.0.4.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-10-13 08:52:54",
"github": false,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"lcname": "unifiedai-sdk"
}
JSON