ccs-llmconnector 1.0.0__py3-none-any.whl

llmconnector/grok_client.py

@@ -0,0 +1,139 @@
+"""Thin wrapper around the xAI Grok chat API via the xai-sdk package."""
+
+from __future__ import annotations
+
+import base64
+import mimetypes
+from pathlib import Path
+from typing import Optional, Sequence, Union
+
+from xai_sdk import Client
+from xai_sdk.chat import image as chat_image
+from xai_sdk.chat import user
+
+ImageInput = Union[str, Path]
+
+
+class GrokClient:
+    """Convenience wrapper around the xAI Grok chat API."""
+
+    def generate_response(
+        self,
+        *,
+        api_key: str,
+        prompt: str,
+        model: str,
+        max_tokens: int = 32000,
+        reasoning_effort: Optional[str] = None,
+        images: Optional[Sequence[ImageInput]] = None,
+    ) -> str:
+        """Generate a response from the specified Grok model.
+
+        Args:
+            api_key: API key used to authenticate with xAI.
+            prompt: Natural-language instruction or query for the model.
+            model: Identifier of the Grok model to target (for example, ``"grok-3"``).
+            max_tokens: Cap for tokens in the generated response, defaults to 32000.
+            reasoning_effort: Optional hint for reasoning-focused models (``"low"`` or ``"high"``).
+            images: Optional collection of image references (local paths, URLs, or data URLs).
+
+        Returns:
+            The text output produced by the model.
+
+        Raises:
+            ValueError: If required arguments are missing or the request payload is empty.
+            RuntimeError: If the Grok response does not contain any textual content.
+        """
+        if not api_key:
+            raise ValueError("api_key must be provided.")
+        if not prompt and not images:
+            raise ValueError("At least one of prompt or images must be provided.")
+        if not model:
+            raise ValueError("model must be provided.")
+
+        message_parts = []
+        if prompt:
+            message_parts.append(prompt)
+
+        if images:
+            for image in images:
+                message_parts.append(chat_image(self._to_image_url(image)))
+
+        if not message_parts:
+            raise ValueError("No content provided for response generation.")
+
+        grok_client = Client(api_key=api_key)
+
+        create_kwargs = {
+            "model": model,
+            "max_tokens": max_tokens,
+            "messages": [user(*message_parts)],
+        }
+
+        normalized_effort = (reasoning_effort or "").strip().lower()
+        if normalized_effort in {"low", "high"}:
+            create_kwargs["reasoning_effort"] = normalized_effort
+
+        chat = grok_client.chat.create(**create_kwargs)
+        response = chat.sample()
+
+        content = getattr(response, "content", None)
+        if content:
+            return content
+
+        reasoning_content = getattr(response, "reasoning_content", None)
+        if reasoning_content:
+            return reasoning_content
+
+        raise RuntimeError("xAI response did not include any text output.")
+
+    def list_models(self, *, api_key: str) -> list[dict[str, Optional[str]]]:
+        """Return the Grok language models available to the authenticated account."""
+        if not api_key:
+            raise ValueError("api_key must be provided.")
+
+        grok_client = Client(api_key=api_key)
+        models: list[dict[str, Optional[str]]] = []
+
+        for model in grok_client.models.list_language_models():
+            model_id = getattr(model, "name", None)
+            if not model_id:
+                continue
+
+            aliases = getattr(model, "aliases", None)
+            display_name: Optional[str] = None
+            if aliases:
+                try:
+                    display_name = next(iter(aliases)) or None
+                except (StopIteration, TypeError):
+                    display_name = None
+
+            models.append(
+                {
+                    "id": model_id,
+                    "display_name": display_name,
+                }
+            )
+
+        return models
+
+    @staticmethod
+    def _to_image_url(image: ImageInput) -> str:
+        """Convert an image reference into a URL or data URL suitable for the xAI SDK."""
+        if isinstance(image, Path):
+            return _encode_image_path(image)
+
+        if image.startswith(("http://", "https://", "data:")):
+            return image
+
+        return _encode_image_path(Path(image))
+
+
+def _encode_image_path(path: Path) -> str:
+    """Generate a data URL for the provided image path."""
+    expanded = path.expanduser()
+    data = expanded.read_bytes()
+    mime_type = mimetypes.guess_type(expanded.name)[0] or "application/octet-stream"
+    encoded = base64.b64encode(data).decode("utf-8")
+    return f"data:{mime_type};base64,{encoded}"
+

llmconnector/openai_client.py

@@ -0,0 +1,139 @@
+"""Thin wrapper around the OpenAI Responses API."""
+
+from __future__ import annotations
+
+import base64
+import mimetypes
+from pathlib import Path
+from typing import Optional, Sequence, Union
+
+from openai import OpenAI
+
+try:
+    from openai import APIError as OpenAIError  # type: ignore
+except ImportError:  # pragma: no cover - fallback for older SDKs
+    from openai.error import OpenAIError  # type: ignore
+
+ImageInput = Union[str, Path]
+
+
+class OpenAIResponsesClient:
+    """Convenience wrapper around the OpenAI Responses API."""
+
+    def generate_response(
+        self,
+        *,
+        api_key: str,
+        prompt: str,
+        model: str,
+        max_tokens: int = 32000,
+        reasoning_effort: Optional[str] = None,
+        images: Optional[Sequence[ImageInput]] = None,
+    ) -> str:
+        """Generate a response from the specified model.
+
+        Args:
+            api_key: Secret key used to authenticate with OpenAI.
+            prompt: Natural-language instruction or query for the model.
+            model: Identifier of the OpenAI model to target (for example, ``"gpt-4o"``).
+            max_tokens: Cap for tokens across the entire exchange, defaults to 32000.
+            reasoning_effort: Optional reasoning effort hint (``"low"``, ``"medium"``, or ``"high"``).
+            images: Optional collection of image references (local paths or URLs).
+
+        Returns:
+            The text output produced by the model.
+
+        Raises:
+            ValueError: If required arguments are missing or the request payload is empty.
+            OpenAIError: If the underlying OpenAI request fails.
+        """
+        if not api_key:
+            raise ValueError("api_key must be provided.")
+        if not prompt and not images:
+            raise ValueError("At least one of prompt or images must be provided.")
+        if not model:
+            raise ValueError("model must be provided.")
+
+        client = OpenAI(api_key=api_key)
+
+        content_blocks = []
+        if prompt:
+            content_blocks.append({"type": "input_text", "text": prompt})
+
+        if images:
+            for image in images:
+                content_blocks.append(self._to_image_block(image))
+
+        if not content_blocks:
+            raise ValueError("No content provided for response generation.")
+
+        request_payload = {
+            "model": model,
+            "input": [
+                {
+                    "role": "user",
+                    "content": content_blocks,
+                }
+            ],
+            "max_output_tokens": max_tokens,
+        }
+
+        if reasoning_effort:
+            request_payload["reasoning"] = {"effort": reasoning_effort}
+
+        response = client.responses.create(**request_payload)
+
+        return response.output_text
+
+    @staticmethod
+    def _to_image_block(image: ImageInput) -> dict:
+        """Convert an image reference into an OpenAI Responses API block."""
+        if isinstance(image, Path):
+            return {
+                "type": "input_image",
+                "image_url": _encode_image_path(image),
+            }
+
+        if image.startswith(("http://", "https://", "data:")):
+            return {
+                "type": "input_image",
+                "image_url": image,
+            }
+
+        return {
+            "type": "input_image",
+            "image_url": _encode_image_path(Path(image)),
+        }
+
+    def list_models(self, *, api_key: str) -> list[dict[str, Optional[str]]]:
+        """Return the models available to the authenticated OpenAI account."""
+        if not api_key:
+            raise ValueError("api_key must be provided.")
+
+        client = OpenAI(api_key=api_key)
+        response = client.models.list()
+        data = getattr(response, "data", []) or []
+
+        models: list[dict[str, Optional[str]]] = []
+        for model in data:
+            model_id = getattr(model, "id", None)
+            if model_id is None and isinstance(model, dict):
+                model_id = model.get("id")
+            if not model_id:
+                continue
+
+            display_name = getattr(model, "display_name", None)
+            if display_name is None and isinstance(model, dict):
+                display_name = model.get("display_name")
+
+            models.append({"id": model_id, "display_name": display_name})
+
+        return models
+
+
+def _encode_image_path(path: Path) -> str:
+    """Generate a data URL for the provided image path."""
+    data = path.read_bytes()
+    encoded = base64.b64encode(data).decode("utf-8")
+    mime_type = mimetypes.guess_type(path.name)[0] or "image/png"
+    return f"data:{mime_type};base64,{encoded}"

llmconnector/py.typed

@@ -0,0 +1 @@
+