lfx-nightly 0.1.11.dev0__py3-none-any.whl

lfx/components/huggingface/__init__.py

@@ -0,0 +1,37 @@
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, Any
+
+from lfx.components._importing import import_mod
+
+if TYPE_CHECKING:
+    from .huggingface import HuggingFaceEndpointsComponent
+    from .huggingface_inference_api import HuggingFaceInferenceAPIEmbeddingsComponent
+
+_dynamic_imports = {
+    "HuggingFaceEndpointsComponent": "huggingface",
+    "HuggingFaceInferenceAPIEmbeddingsComponent": "huggingface_inference_api",
+}
+
+__all__ = [
+    "HuggingFaceEndpointsComponent",
+    "HuggingFaceInferenceAPIEmbeddingsComponent",
+]
+
+
+def __getattr__(attr_name: str) -> Any:
+    """Lazily import huggingface components on attribute access."""
+    if attr_name not in _dynamic_imports:
+        msg = f"module '{__name__}' has no attribute '{attr_name}'"
+        raise AttributeError(msg)
+    try:
+        result = import_mod(attr_name, _dynamic_imports[attr_name], __spec__.parent)
+    except (ModuleNotFoundError, ImportError, AttributeError) as e:
+        msg = f"Could not import '{attr_name}' from '{__name__}': {e}"
+        raise AttributeError(msg) from e
+    globals()[attr_name] = result
+    return result
+
+
+def __dir__() -> list[str]:
+    return list(__all__)

lfx/components/huggingface/huggingface.py

@@ -0,0 +1,197 @@
+from typing import Any
+
+from langchain_community.llms.huggingface_endpoint import HuggingFaceEndpoint
+from tenacity import retry, stop_after_attempt, wait_fixed
+
+from lfx.base.models.model import LCModelComponent
+from lfx.field_typing import LanguageModel
+from lfx.field_typing.range_spec import RangeSpec
+from lfx.io import DictInput, DropdownInput, FloatInput, IntInput, SecretStrInput, SliderInput, StrInput
+
+# TODO: langchain_community.llms.huggingface_endpoint is depreciated.
+#  Need to update to langchain_huggingface, but have dependency with langchain_core 0.3.0
+
+# Constants
+DEFAULT_MODEL = "meta-llama/Llama-3.3-70B-Instruct"
+
+
+class HuggingFaceEndpointsComponent(LCModelComponent):
+    display_name: str = "Hugging Face"
+    description: str = "Generate text using Hugging Face Inference APIs."
+    icon = "HuggingFace"
+    name = "HuggingFaceModel"
+
+    inputs = [
+        *LCModelComponent.get_base_inputs(),
+        DropdownInput(
+            name="model_id",
+            display_name="Model ID",
+            info="Select a model from Hugging Face Hub",
+            options=[
+                DEFAULT_MODEL,
+                "mistralai/Mixtral-8x7B-Instruct-v0.1",
+                "mistralai/Mistral-7B-Instruct-v0.3",
+                "meta-llama/Llama-3.1-8B-Instruct",
+                "Qwen/Qwen2.5-Coder-32B-Instruct",
+                "Qwen/QwQ-32B-Preview",
+                "openai-community/gpt2",
+                "custom",
+            ],
+            value=DEFAULT_MODEL,
+            required=True,
+            real_time_refresh=True,
+        ),
+        StrInput(
+            name="custom_model",
+            display_name="Custom Model ID",
+            info="Enter a custom model ID from Hugging Face Hub",
+            value="",
+            show=False,
+            required=True,
+        ),
+        IntInput(
+            name="max_new_tokens", display_name="Max New Tokens", value=512, info="Maximum number of generated tokens"
+        ),
+        IntInput(
+            name="top_k",
+            display_name="Top K",
+            advanced=True,
+            info="The number of highest probability vocabulary tokens to keep for top-k-filtering",
+        ),
+        FloatInput(
+            name="top_p",
+            display_name="Top P",
+            value=0.95,
+            advanced=True,
+            info=(
+                "If set to < 1, only the smallest set of most probable tokens with "
+                "probabilities that add up to `top_p` or higher are kept for generation"
+            ),
+        ),
+        FloatInput(
+            name="typical_p",
+            display_name="Typical P",
+            value=0.95,
+            advanced=True,
+            info="Typical Decoding mass.",
+        ),
+        SliderInput(
+            name="temperature",
+            display_name="Temperature",
+            value=0.8,
+            range_spec=RangeSpec(min=0, max=2, step=0.01),
+            info="The value used to module the logits distribution",
+            advanced=True,
+        ),
+        FloatInput(
+            name="repetition_penalty",
+            display_name="Repetition Penalty",
+            info="The parameter for repetition penalty. 1.0 means no penalty.",
+            advanced=True,
+        ),
+        StrInput(
+            name="inference_endpoint",
+            display_name="Inference Endpoint",
+            value="https://api-inference.huggingface.co/models/",
+            info="Custom inference endpoint URL.",
+            required=True,
+        ),
+        DropdownInput(
+            name="task",
+            display_name="Task",
+            options=["text2text-generation", "text-generation", "summarization", "translation"],
+            value="text-generation",
+            advanced=True,
+            info="The task to call the model with. Should be a task that returns `generated_text` or `summary_text`.",
+        ),
+        SecretStrInput(name="huggingfacehub_api_token", display_name="API Token", password=True, required=True),
+        DictInput(name="model_kwargs", display_name="Model Keyword Arguments", advanced=True),
+        IntInput(name="retry_attempts", display_name="Retry Attempts", value=1, advanced=True),
+    ]
+
+    def get_api_url(self) -> str:
+        if "huggingface" in self.inference_endpoint.lower():
+            if self.model_id == "custom":
+                if not self.custom_model:
+                    error_msg = "Custom model ID is required when 'custom' is selected"
+                    raise ValueError(error_msg)
+                return f"{self.inference_endpoint}{self.custom_model}"
+            return f"{self.inference_endpoint}{self.model_id}"
+        return self.inference_endpoint
+
+    async def update_build_config(self, build_config: dict, field_value: Any, field_name: str | None = None) -> dict:
+        """Update build configuration based on field updates."""
+        try:
+            if field_name is None or field_name == "model_id":
+                # If model_id is custom, show custom model field
+                if field_value == "custom":
+                    build_config["custom_model"]["show"] = True
+                    build_config["custom_model"]["required"] = True
+                else:
+                    build_config["custom_model"]["show"] = False
+                    build_config["custom_model"]["value"] = ""
+
+        except (KeyError, AttributeError) as e:
+            self.log(f"Error updating build config: {e!s}")
+        return build_config
+
+    def create_huggingface_endpoint(
+        self,
+        task: str | None,
+        huggingfacehub_api_token: str | None,
+        model_kwargs: dict[str, Any],
+        max_new_tokens: int,
+        top_k: int | None,
+        top_p: float,
+        typical_p: float | None,
+        temperature: float | None,
+        repetition_penalty: float | None,
+    ) -> HuggingFaceEndpoint:
+        retry_attempts = self.retry_attempts
+        endpoint_url = self.get_api_url()
+
+        @retry(stop=stop_after_attempt(retry_attempts), wait=wait_fixed(2))
+        def _attempt_create():
+            return HuggingFaceEndpoint(
+                endpoint_url=endpoint_url,
+                task=task,
+                huggingfacehub_api_token=huggingfacehub_api_token,
+                model_kwargs=model_kwargs,
+                max_new_tokens=max_new_tokens,
+                top_k=top_k,
+                top_p=top_p,
+                typical_p=typical_p,
+                temperature=temperature,
+                repetition_penalty=repetition_penalty,
+            )
+
+        return _attempt_create()
+
+    def build_model(self) -> LanguageModel:
+        task = self.task or None
+        huggingfacehub_api_token = self.huggingfacehub_api_token
+        model_kwargs = self.model_kwargs or {}
+        max_new_tokens = self.max_new_tokens
+        top_k = self.top_k or None
+        top_p = self.top_p
+        typical_p = self.typical_p or None
+        temperature = self.temperature or 0.8
+        repetition_penalty = self.repetition_penalty or None
+
+        try:
+            llm = self.create_huggingface_endpoint(
+                task=task,
+                huggingfacehub_api_token=huggingfacehub_api_token,
+                model_kwargs=model_kwargs,
+                max_new_tokens=max_new_tokens,
+                top_k=top_k,
+                top_p=top_p,
+                typical_p=typical_p,
+                temperature=temperature,
+                repetition_penalty=repetition_penalty,
+            )
+        except Exception as e:
+            msg = "Could not connect to Hugging Face Endpoints API."
+            raise ValueError(msg) from e
+
+        return llm

lfx/components/huggingface/huggingface_inference_api.py

@@ -0,0 +1,106 @@
+from urllib.parse import urlparse
+
+import requests
+from langchain_community.embeddings.huggingface import HuggingFaceInferenceAPIEmbeddings
+
+# Next update: use langchain_huggingface
+from pydantic import SecretStr
+from tenacity import retry, stop_after_attempt, wait_fixed
+
+from lfx.base.embeddings.model import LCEmbeddingsModel
+from lfx.field_typing import Embeddings
+from lfx.io import MessageTextInput, Output, SecretStrInput
+
+
+class HuggingFaceInferenceAPIEmbeddingsComponent(LCEmbeddingsModel):
+    display_name = "Hugging Face Embeddings Inference"
+    description = "Generate embeddings using Hugging Face Text Embeddings Inference (TEI)"
+    documentation = "https://huggingface.co/docs/text-embeddings-inference/index"
+    icon = "HuggingFace"
+    name = "HuggingFaceInferenceAPIEmbeddings"
+
+    inputs = [
+        SecretStrInput(
+            name="api_key",
+            display_name="API Key",
+            advanced=False,
+            info="Required for non-local inference endpoints. Local inference does not require an API Key.",
+        ),
+        MessageTextInput(
+            name="inference_endpoint",
+            display_name="Inference Endpoint",
+            required=True,
+            value="https://api-inference.huggingface.co/models/",
+            info="Custom inference endpoint URL.",
+        ),
+        MessageTextInput(
+            name="model_name",
+            display_name="Model Name",
+            value="BAAI/bge-large-en-v1.5",
+            info="The name of the model to use for text embeddings.",
+            required=True,
+        ),
+    ]
+
+    outputs = [
+        Output(display_name="Embeddings", name="embeddings", method="build_embeddings"),
+    ]
+
+    def validate_inference_endpoint(self, inference_endpoint: str) -> bool:
+        parsed_url = urlparse(inference_endpoint)
+        if not all([parsed_url.scheme, parsed_url.netloc]):
+            msg = (
+                f"Invalid inference endpoint format: '{self.inference_endpoint}'. "
+                "Please ensure the URL includes both a scheme (e.g., 'http://' or 'https://') and a domain name. "
+                "Example: 'http://localhost:8080' or 'https://api.example.com'"
+            )
+            raise ValueError(msg)
+
+        try:
+            response = requests.get(f"{inference_endpoint}/health", timeout=5)
+        except requests.RequestException as e:
+            msg = (
+                f"Inference endpoint '{inference_endpoint}' is not responding. "
+                "Please ensure the URL is correct and the service is running."
+            )
+            raise ValueError(msg) from e
+
+        if response.status_code != requests.codes.ok:
+            msg = f"Hugging Face health check failed: {response.status_code}"
+            raise ValueError(msg)
+        # returning True to solve linting error
+        return True
+
+    def get_api_url(self) -> str:
+        if "huggingface" in self.inference_endpoint.lower():
+            return f"{self.inference_endpoint}"
+        return self.inference_endpoint
+
+    @retry(stop=stop_after_attempt(3), wait=wait_fixed(2))
+    def create_huggingface_embeddings(
+        self, api_key: SecretStr, api_url: str, model_name: str
+    ) -> HuggingFaceInferenceAPIEmbeddings:
+        return HuggingFaceInferenceAPIEmbeddings(api_key=api_key, api_url=api_url, model_name=model_name)
+
+    def build_embeddings(self) -> Embeddings:
+        api_url = self.get_api_url()
+
+        is_local_url = (
+            api_url.startswith(("http://localhost", "http://127.0.0.1", "http://0.0.0.0", "http://docker"))
+            or "huggingface.co" not in api_url.lower()
+        )
+
+        if not self.api_key and is_local_url:
+            self.validate_inference_endpoint(api_url)
+            api_key = SecretStr("APIKeyForLocalDeployment")
+        elif not self.api_key:
+            msg = "API Key is required for non-local inference endpoints"
+            raise ValueError(msg)
+        else:
+            api_key = SecretStr(self.api_key).get_secret_value()
+
+        try:
+            return self.create_huggingface_embeddings(api_key, api_url, self.model_name)
+        except Exception as e:
+            msg = "Could not connect to Hugging Face Inference API."
+            raise ValueError(msg) from e

lfx/components/ibm/__init__.py

@@ -0,0 +1,34 @@
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, Any
+
+from lfx.components._importing import import_mod
+
+if TYPE_CHECKING:
+    from lfx.components.ibm.watsonx import WatsonxAIComponent
+    from lfx.components.ibm.watsonx_embeddings import WatsonxEmbeddingsComponent
+
+_dynamic_imports = {
+    "WatsonxAIComponent": "watsonx",
+    "WatsonxEmbeddingsComponent": "watsonx_embeddings",
+}
+
+__all__ = ["WatsonxAIComponent", "WatsonxEmbeddingsComponent"]
+
+
+def __getattr__(attr_name: str) -> Any:
+    """Lazily import ibm components on attribute access."""
+    if attr_name not in _dynamic_imports:
+        msg = f"module '{__name__}' has no attribute '{attr_name}'"
+        raise AttributeError(msg)
+    try:
+        result = import_mod(attr_name, _dynamic_imports[attr_name], __spec__.parent)
+    except (ModuleNotFoundError, ImportError, AttributeError) as e:
+        msg = f"Could not import '{attr_name}' from '{__name__}': {e}"
+        raise AttributeError(msg) from e
+    globals()[attr_name] = result
+    return result
+
+
+def __dir__() -> list[str]:
+    return list(__all__)

lfx/components/ibm/watsonx.py

@@ -0,0 +1,203 @@
+import json
+from typing import Any
+
+import requests
+from langchain_ibm import ChatWatsonx
+from pydantic.v1 import SecretStr
+
+from lfx.base.models.model import LCModelComponent
+from lfx.field_typing import LanguageModel
+from lfx.field_typing.range_spec import RangeSpec
+from lfx.inputs.inputs import BoolInput, DropdownInput, IntInput, SecretStrInput, SliderInput, StrInput
+from lfx.log.logger import logger
+from lfx.schema.dotdict import dotdict
+
+
+class WatsonxAIComponent(LCModelComponent):
+    display_name = "IBM watsonx.ai"
+    description = "Generate text using IBM watsonx.ai foundation models."
+    icon = "WatsonxAI"
+    name = "IBMwatsonxModel"
+    beta = False
+
+    _default_models = ["ibm/granite-3-2b-instruct", "ibm/granite-3-8b-instruct", "ibm/granite-13b-instruct-v2"]
+
+    inputs = [
+        *LCModelComponent.get_base_inputs(),
+        DropdownInput(
+            name="url",
+            display_name="watsonx API Endpoint",
+            info="The base URL of the API.",
+            value=None,
+            options=[
+                "https://us-south.ml.cloud.ibm.com",
+                "https://eu-de.ml.cloud.ibm.com",
+                "https://eu-gb.ml.cloud.ibm.com",
+                "https://au-syd.ml.cloud.ibm.com",
+                "https://jp-tok.ml.cloud.ibm.com",
+                "https://ca-tor.ml.cloud.ibm.com",
+            ],
+            real_time_refresh=True,
+        ),
+        StrInput(
+            name="project_id",
+            display_name="watsonx Project ID",
+            required=True,
+            info="The project ID or deployment space ID that is associated with the foundation model.",
+        ),
+        SecretStrInput(
+            name="api_key",
+            display_name="API Key",
+            info="The API Key to use for the model.",
+            required=True,
+        ),
+        DropdownInput(
+            name="model_name",
+            display_name="Model Name",
+            options=[],
+            value=None,
+            dynamic=True,
+            required=True,
+        ),
+        IntInput(
+            name="max_tokens",
+            display_name="Max Tokens",
+            advanced=True,
+            info="The maximum number of tokens to generate.",
+            range_spec=RangeSpec(min=1, max=4096),
+            value=1000,
+        ),
+        StrInput(
+            name="stop_sequence",
+            display_name="Stop Sequence",
+            advanced=True,
+            info="Sequence where generation should stop.",
+            field_type="str",
+        ),
+        SliderInput(
+            name="temperature",
+            display_name="Temperature",
+            info="Controls randomness, higher values increase diversity.",
+            value=0.1,
+            range_spec=RangeSpec(min=0, max=2, step=0.01),
+            advanced=True,
+        ),
+        SliderInput(
+            name="top_p",
+            display_name="Top P",
+            info="The cumulative probability cutoff for token selection. "
+            "Lower values mean sampling from a smaller, more top-weighted nucleus.",
+            value=0.9,
+            range_spec=RangeSpec(min=0, max=1, step=0.01),
+            advanced=True,
+        ),
+        SliderInput(
+            name="frequency_penalty",
+            display_name="Frequency Penalty",
+            info="Penalty for frequency of token usage.",
+            value=0.5,
+            range_spec=RangeSpec(min=-2.0, max=2.0, step=0.01),
+            advanced=True,
+        ),
+        SliderInput(
+            name="presence_penalty",
+            display_name="Presence Penalty",
+            info="Penalty for token presence in prior text.",
+            value=0.3,
+            range_spec=RangeSpec(min=-2.0, max=2.0, step=0.01),
+            advanced=True,
+        ),
+        IntInput(
+            name="seed",
+            display_name="Random Seed",
+            advanced=True,
+            info="The random seed for the model.",
+            value=8,
+        ),
+        BoolInput(
+            name="logprobs",
+            display_name="Log Probabilities",
+            advanced=True,
+            info="Whether to return log probabilities of the output tokens.",
+            value=True,
+        ),
+        IntInput(
+            name="top_logprobs",
+            display_name="Top Log Probabilities",
+            advanced=True,
+            info="Number of most likely tokens to return at each position.",
+            value=3,
+            range_spec=RangeSpec(min=1, max=20),
+        ),
+        StrInput(
+            name="logit_bias",
+            display_name="Logit Bias",
+            advanced=True,
+            info='JSON string of token IDs to bias or suppress (e.g., {"1003": -100, "1004": 100}).',
+            field_type="str",
+        ),
+    ]
+
+    @staticmethod
+    def fetch_models(base_url: str) -> list[str]:
+        """Fetch available models from the watsonx.ai API."""
+        try:
+            endpoint = f"{base_url}/ml/v1/foundation_model_specs"
+            params = {"version": "2024-09-16", "filters": "function_text_chat,!lifecycle_withdrawn"}
+            response = requests.get(endpoint, params=params, timeout=10)
+            response.raise_for_status()
+            data = response.json()
+            models = [model["model_id"] for model in data.get("resources", [])]
+            return sorted(models)
+        except Exception:  # noqa: BLE001
+            logger.exception("Error fetching models. Using default models.")
+            return WatsonxAIComponent._default_models
+
+    def update_build_config(self, build_config: dotdict, field_value: Any, field_name: str | None = None):
+        """Update model options when URL or API key changes."""
+        logger.info("Updating build config. Field name: %s, Field value: %s", field_name, field_value)
+
+        if field_name == "url" and field_value:
+            try:
+                models = self.fetch_models(base_url=build_config.url.value)
+                build_config.model_name.options = models
+                if build_config.model_name.value:
+                    build_config.model_name.value = models[0]
+                info_message = f"Updated model options: {len(models)} models found in {build_config.url.value}"
+                logger.info(info_message)
+            except Exception:  # noqa: BLE001
+                logger.exception("Error updating model options.")
+
+    def build_model(self) -> LanguageModel:
+        # Parse logit_bias from JSON string if provided
+        logit_bias = None
+        if hasattr(self, "logit_bias") and self.logit_bias:
+            try:
+                logit_bias = json.loads(self.logit_bias)
+            except json.JSONDecodeError:
+                logger.warning("Invalid logit_bias JSON format. Using default instead.")
+                logit_bias = {"1003": -100, "1004": -100}
+
+        chat_params = {
+            "max_tokens": getattr(self, "max_tokens", None),
+            "temperature": getattr(self, "temperature", None),
+            "top_p": getattr(self, "top_p", None),
+            "frequency_penalty": getattr(self, "frequency_penalty", None),
+            "presence_penalty": getattr(self, "presence_penalty", None),
+            "seed": getattr(self, "seed", None),
+            "stop": [self.stop_sequence] if self.stop_sequence else [],
+            "n": 1,
+            "logprobs": getattr(self, "logprobs", True),
+            "top_logprobs": getattr(self, "top_logprobs", None),
+            "time_limit": 600000,
+            "logit_bias": logit_bias,
+        }
+
+        return ChatWatsonx(
+            apikey=SecretStr(self.api_key).get_secret_value(),
+            url=self.url,
+            project_id=self.project_id,
+            model_id=self.model_name,
+            params=chat_params,
+            streaming=self.stream,
+        )