orbitals 0.0.3__tar.gz → 0.1.1__tar.gz

orbitals-0.1.1/Dockerfile

@@ -0,0 +1,23 @@
+FROM vllm/vllm-openai:v0.14.1
+
+ARG DEBIAN_FRONTEND=noninteractive
+
+COPY --from=ghcr.io/astral-sh/uv:0.9.8-debian-slim /usr/local/bin/uv /usr/local/bin/uvx /bin/
+
+WORKDIR /app
+ENV PATH="/app/.venv/bin:$PATH"
+ENV UV_COMPILE_BYTECODE=1
+
+ARG MODEL
+ENV MODEL=${MODEL}
+
+RUN hf download ${MODEL}
+
+COPY pyproject.toml uv.lock README.md src /app/src/
+
+RUN pip install -e .[serving]
+
+EXPOSE 8000
+
+ENTRYPOINT ["/bin/bash", "-c"]
+CMD ["orbitals scope-guard serve ${MODEL} --port 8000 --host 0.0.0.0"]

orbitals-0.0.3/Dockerfile → orbitals-0.1.1/Dockerfile copy

@@ -1,7 +1,6 @@
-FROM vllm/vllm-openai:v0.14.1 as builder
+FROM nvidia/cuda:13.1.1-cudnn-devel-ubuntu24.04 AS builder
 
 ARG DEBIAN_FRONTEND=noninteractive
-ARG MODEL
 
 COPY --from=ghcr.io/astral-sh/uv:0.9.8-debian-slim /usr/local/bin/uv /usr/local/bin/uvx /bin/
 
@@ -11,7 +10,7 @@ ENV UV_COMPILE_BYTECODE=1
 
 RUN \
     apt-get update && \
-    apt-get install -y --no-install-recommends git
+    apt-get install -y --no-install-recommends git python3.12
 
 COPY pyproject.toml uv.lock README.md /app/
 
@@ -19,8 +18,8 @@ RUN --mount=type=cache,target=/root/.cache/uv \
     uv venv -p 3.12 && \
     uv sync --frozen --extra scope-guard-serve --no-install-project --no-dev
 
-RUN --mount=type=cache,target=/root/.cache/uv \
-    hf download ${MODEL}
+ARG MODEL
+RUN hf download ${MODEL}
 
 COPY src /app/src/
 
@@ -30,13 +29,15 @@ RUN --mount=type=cache,target=/root/.cache/uv \
 # TODO remove next line
 ENTRYPOINT ["/bin/bash", "-c"]
 
-FROM vllm/vllm-openai:v0.14.1 as runner
-
-ARG MODEL
+FROM nvidia/cuda:13.1.1-cudnn-runtime-ubuntu24.04 AS runner
 
 WORKDIR /app
 ENV PATH="/app/.venv/bin:$PATH"
 
+RUN \
+    apt-get update && \
+    apt-get install -y --no-install-recommends python3.12
+
 COPY --from=builder /app/.venv/ /app/.venv/
 
 RUN mkdir -p /root/.cache/huggingface/hub
@@ -44,5 +45,8 @@ COPY --from=builder /root/.cache/huggingface/hub /root/.cache/huggingface/hub
 
 EXPOSE 8000
 
+ARG MODEL
+ENV MODEL=${MODEL}
+
 ENTRYPOINT ["/bin/bash", "-c"]
-CMD [ "orbitals", "scope-guard", "serve", "principled-intelligence/${MODEL}", "--port", "8000", "--host", "0.0.0.0" ]
+CMD ["orbitals scope-guard serve ${MODEL} --port 8000 --host 0.0.0.0"]

{orbitals-0.0.3 → orbitals-0.1.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: orbitals
-Version: 0.0.3
+Version: 0.1.1
 Summary: LLM Guardrails tailored to your Principles
 Author-email: Luigi Procopio <luigi@principled-intelligence.com>, Edoardo Barba <edoardo@principled-intelligence.com>
 License: Apache-2.0
@@ -11,7 +11,7 @@ Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
 Classifier: Programming Language :: Python :: 3.13
 Classifier: Programming Language :: Python :: 3.14
-Requires-Python: >=3.10
+Requires-Python: ==3.13
 Requires-Dist: aiohttp
 Requires-Dist: pydantic>=2.0.0
 Requires-Dist: requests
@@ -19,35 +19,38 @@ Requires-Dist: typer>=0.12.3
 Provides-Extra: all
 Requires-Dist: accelerate>=1.11.0; extra == 'all'
 Requires-Dist: fastapi[standard]>=0.119.1; extra == 'all'
-Requires-Dist: pynvml; extra == 'all'
+Requires-Dist: nvidia-ml-py; extra == 'all'
 Requires-Dist: transformers<5.0.0,>=4.47.0; extra == 'all'
 Requires-Dist: uvicorn>=0.29.0; extra == 'all'
-Requires-Dist: vllm>=0.11.0; extra == 'all'
+Requires-Dist: vllm<0.13.0,>=0.11.0; extra == 'all'
 Requires-Dist: xgrammar; extra == 'all'
 Provides-Extra: scope-guard-all
 Requires-Dist: accelerate>=1.11.0; extra == 'scope-guard-all'
 Requires-Dist: fastapi[standard]>=0.119.1; extra == 'scope-guard-all'
-Requires-Dist: pynvml; extra == 'scope-guard-all'
+Requires-Dist: nvidia-ml-py; extra == 'scope-guard-all'
 Requires-Dist: transformers<5.0.0,>=4.47.0; extra == 'scope-guard-all'
 Requires-Dist: uvicorn>=0.29.0; extra == 'scope-guard-all'
-Requires-Dist: vllm>=0.11.0; extra == 'scope-guard-all'
+Requires-Dist: vllm<0.13.0,>=0.11.0; extra == 'scope-guard-all'
 Requires-Dist: xgrammar; extra == 'scope-guard-all'
 Provides-Extra: scope-guard-hf
 Requires-Dist: accelerate>=1.11.0; extra == 'scope-guard-hf'
-Requires-Dist: pynvml; extra == 'scope-guard-hf'
+Requires-Dist: nvidia-ml-py; extra == 'scope-guard-hf'
 Requires-Dist: transformers<5.0.0,>=4.47.0; extra == 'scope-guard-hf'
 Provides-Extra: scope-guard-serve
 Requires-Dist: fastapi[standard]>=0.119.1; extra == 'scope-guard-serve'
-Requires-Dist: pynvml; extra == 'scope-guard-serve'
+Requires-Dist: nvidia-ml-py; extra == 'scope-guard-serve'
 Requires-Dist: transformers<5.0.0,>=4.47.0; extra == 'scope-guard-serve'
 Requires-Dist: uvicorn>=0.29.0; extra == 'scope-guard-serve'
-Requires-Dist: vllm>=0.11.0; extra == 'scope-guard-serve'
+Requires-Dist: vllm<0.13.0,>=0.11.0; extra == 'scope-guard-serve'
 Requires-Dist: xgrammar; extra == 'scope-guard-serve'
 Provides-Extra: scope-guard-vllm
-Requires-Dist: pynvml; extra == 'scope-guard-vllm'
+Requires-Dist: nvidia-ml-py; extra == 'scope-guard-vllm'
 Requires-Dist: transformers<5.0.0,>=4.47.0; extra == 'scope-guard-vllm'
-Requires-Dist: vllm>=0.11.0; extra == 'scope-guard-vllm'
+Requires-Dist: vllm<0.13.0,>=0.11.0; extra == 'scope-guard-vllm'
 Requires-Dist: xgrammar; extra == 'scope-guard-vllm'
+Provides-Extra: serving
+Requires-Dist: fastapi[standard]>=0.119.1; extra == 'serving'
+Requires-Dist: uvicorn>=0.29.0; extra == 'serving'
 Description-Content-Type: text/markdown
 
 <div align="center">

{orbitals-0.0.3 → orbitals-0.1.1}/README.scope-guard.md

@@ -282,24 +282,15 @@ pip install orbitals[scope-guard-serve]
 orbitals scope-guard serve scope-guard --port 8000
 ```
 
-Alternatively, we also release a pre-built Docker image:
-
-```bash
-docker run --runtime nvidia --gpus all \
-    -p 8000:8000 \
-    --ipc=host \
-    principled-intelligence/scope-guard:0.1.0-scope-guard-4B-q-2601
-```
-
 Once the server is running, you can interact with it as follows:
 
 #### 1. Direct HTTP Requests
 
-Send requests to the `/api/in/scope-classifier/classify` (or `/api/in/scope-classifier/batch-classify`) endpoint using cURL or any HTTP client:
+Send requests to the `/orbitals/scope-guard/validate` (or `/orbitals/scope-guard/batch-validatevalidate`) endpoint using cURL or any HTTP client:
 
 ```bash
 curl -X 'POST' \
-    'http://localhost:8000/api/in/scope-classifier/classify' \
+    'http://localhost:8000/orbitals/scope-guard/validate' \
     -H 'accept: application/json' \
     -H 'Content-Type: application/json' \
     -d '{

{orbitals-0.0.3 → orbitals-0.1.1}/pyproject.toml

@@ -4,11 +4,11 @@ build-backend = "hatchling.build"
 
 [project]
 name = "orbitals"
-version = "0.0.3"
+version = "0.1.1"
 description = "LLM Guardrails tailored to your Principles"
 readme = "README.md"
 
-requires-python = ">=3.10"
+requires-python = "==3.13"
 license = { text = "Apache-2.0" }
 
 authors = [
@@ -29,19 +29,19 @@ classifiers = [
 dependencies = ["pydantic>=2.0.0", "typer>=0.12.3", "requests", "aiohttp"]
 
 [project.optional-dependencies]
-scope-guard-hf = ["transformers>=4.47.0,<5.0.0", "accelerate>=1.11.0", "pynvml"]
+serving = ["uvicorn>=0.29.0", "fastapi[standard]>=0.119.1"]
+scope-guard-hf = [
+    "transformers>=4.47.0,<5.0.0",
+    "accelerate>=1.11.0",
+    "nvidia-ml-py",
+]
 scope-guard-vllm = [
     "transformers>=4.47.0,<5.0.0",
-    "vllm>=0.11.0",
+    "vllm>=0.11.0,<0.13.0",
     "xgrammar",
-    "pynvml",
-]
-scope-guard-serve = [
-    "orbitals[scope-guard-vllm]",
-    "fastapi[standard]>=0.119.1",
-    "uvicorn>=0.29.0",
-    "vllm>=0.11.0",
+    "nvidia-ml-py",
 ]
+scope-guard-serve = ["orbitals[scope-guard-vllm]", "orbitals[serving]"]
 scope-guard-all = [
     "orbitals[scope-guard-vllm]",
     "orbitals[scope-guard-hf]",

{orbitals-0.0.3 → orbitals-0.1.1}/src/orbitals/scope_guard/serving/main.py

@@ -54,19 +54,13 @@ class ScopeGuardResponse(BaseModel):
     time_taken: float
 
 
-@app.post("/api/orbitals/scope-guard/validate", response_model=ScopeGuardResponse)
-@app.post("/api/in/scope-classifier/classify", response_model=ScopeGuardResponse)
+@app.post("/orbitals/scope-guard/validate", response_model=ScopeGuardResponse)
 async def validate(
     request: Request,
     conversation: ScopeGuardInput,
     ai_service_description: Annotated[str | AIServiceDescription, Body()],
     skip_evidences: bool | None = Body(None),
 ) -> ScopeGuardResponse:
-    if request.url.path.endswith("/api/in/scope-classifier/classify"):
-        logging.warning(
-            "The /api/in/scope-classifier/classify endpoint is deprecated. Please use /api/in/scope-guard/validate instead."
-        )
-
     global scope_guard
 
     start_time = time.time()
@@ -85,10 +79,9 @@ async def validate(
 
 
 @app.post(
-    "/api/orbitals/scope-guard/batch-validate",
+    "/orbitals/scope-guard/batch-validate",
     response_model=list[ScopeGuardResponse],
 )
-@app.post("/api/in/scope-classifier/batch-classify", response_model=ScopeGuardResponse)
 async def batch_validate(
     request: Request,
     conversations: list[ScopeGuardInput],
@@ -96,11 +89,6 @@ async def batch_validate(
     ai_service_descriptions: list[str] | list[AIServiceDescription] | None = None,
     skip_evidences: bool | None = Body(None),
 ) -> list[ScopeGuardResponse]:
-    if request.url.path.endswith("/api/in/scope-classifier/batch-classify"):
-        logging.warning(
-            "The /api/in/scope-classifier/batch-classify endpoint is deprecated. Please use /api/in/scope-guard/batch-validate instead."
-        )
-
     global scope_guard
 
     start_time = time.time()

{orbitals-0.0.3 → orbitals-0.1.1}/src/orbitals/utils.py

@@ -8,7 +8,7 @@ def maybe_configure_gpu_usage():
     optimal usage: search for the gpu with the most free memory, and
     set CUDA_VISIBLE_DEVICES to that GPU only.
 
-    Uses pynvml to avoid triggering CUDA initialization from torch.
+    Uses nvidia-ml-py (pynvml) to avoid triggering CUDA initialization from torch.
     """
     if "CUDA_VISIBLE_DEVICES" in os.environ:
         logging.info(
@@ -17,15 +17,15 @@ def maybe_configure_gpu_usage():
         return
 
     try:
-        import pynvml  # ty: ignore[unresolved-import]
+        import pynvml  # ty: ignore[unresolved-import]  # provided by nvidia-ml-py
     except ModuleNotFoundError:
-        logging.debug("pynvml not available, skipping GPU auto-configuration")
+        logging.debug("nvidia-ml-py not available, skipping GPU auto-configuration")
         return
 
     try:
         pynvml.nvmlInit()
     except pynvml.NVMLError:
-        logging.debug("NVML initialization failed, skipping GPU auto-configuration")
+        logging.error("NVML initialization failed, skipping GPU auto-configuration")
         return
 
     try:

{orbitals-0.0.3 → orbitals-0.1.1}/src/scripts/push_hf_pipeline.py

@@ -5,22 +5,15 @@ from hf_pipeline.scope_guard import ScopeGuardPipeline
 
 
 def main():
-    # CUDA_VISIBLE_DEVICES="" PYTHONPATH=$(pwd) python src/scripts/push_hf_pipeline.py
-
     PIPELINE_REGISTRY.register_pipeline(
         "scope-guard",
         pipeline_class=ScopeGuardPipeline,
         pt_model=AutoModelForSequenceClassification,
-        default={
-            "pt": (
-                "Qwen/Qwen3-0.6B",
-                "c1899de",
-            )
-        },
+        default={"pt": ("principled-intelligence/scope-guard-4B-q-2601", "631addd")},
     )
 
     p = pipeline(task="scope-guard")
-    p.push_to_hub("principled-intelligence/scope-guard", private=True)
+    p.push_to_hub("principled-intelligence/scope-guard", private=False)
 
 
 if __name__ == "__main__":