nvidia-haystack 0.0.4__tar.gz → 0.1.0__tar.gz

nvidia_haystack-0.1.0/CHANGELOG.md

@@ -0,0 +1,63 @@
+# Changelog
+
+## [unreleased]
+
+### 🚀 Features
+
+- Update default embedding model to nvidia/nv-embedqa-e5-v5 (#1015)
+- Add NVIDIA NIM ranker support (#1023)
+
+### 🐛 Bug Fixes
+
+- Lints in `nvidia-haystack` (#993)
+
+### 🚜 Refactor
+
+- Remove deprecated Nvidia Cloud Functions backend and related code. (#803)
+
+### 📚 Documentation
+
+- Update Nvidia API docs (#1031)
+
+### 🧪 Testing
+
+- Do not retry tests in `hatch run test` command (#954)
+
+### ⚙️ Miscellaneous Tasks
+
+- Retry tests to reduce flakyness (#836)
+- Update ruff invocation to include check parameter (#853)
+
+### Docs
+
+- Update NvidiaGenerator docstrings (#966)
+
+## [integrations/nvidia-v0.0.3] - 2024-05-22
+
+### 📚 Documentation
+
+- Update docstrings of Nvidia integrations (#599)
+
+### ⚙️ Miscellaneous Tasks
+
+- Add generate docs to Nvidia workflow (#603)
+
+## [integrations/nvidia-v0.0.2] - 2024-03-18
+
+### 📚 Documentation
+
+- Disable-class-def (#556)
+
+## [integrations/nvidia-v0.0.1] - 2024-03-07
+
+### 🚀 Features
+
+- Add `NvidiaTextEmbedder`, `NvidiaDocumentEmbedder` and co. (#537)
+
+### 🐛 Bug Fixes
+
+- `nvidia-haystack`- Handle non-strict env var secrets correctly (#543)
+
+## [integrations/nvidia-v0.0.0] - 2024-03-01
+
+<!-- generated by git-cliff -->

{nvidia_haystack-0.0.4 → nvidia_haystack-0.1.0}/PKG-INFO

@@ -1,12 +1,11 @@
 Metadata-Version: 2.3
 Name: nvidia-haystack
-Version: 0.0.4
+Version: 0.1.0
 Project-URL: Documentation, https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/nvidia#readme
 Project-URL: Issues, https://github.com/deepset-ai/haystack-core-integrations/issues
 Project-URL: Source, https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/nvidia
 Author-email: deepset GmbH <info@deepset.ai>
-License-Expression: Apache-2.0
-License-File: LICENSE.txt
+License: Apache-2.0
 Classifier: Development Status :: 4 - Beta
 Classifier: License :: OSI Approved :: Apache Software License
 Classifier: Programming Language :: Python

{nvidia_haystack-0.0.4 → nvidia_haystack-0.1.0}/pydoc/config.yml

@@ -5,7 +5,10 @@ loaders:
       [
         "haystack_integrations.components.embedders.nvidia.document_embedder",
         "haystack_integrations.components.embedders.nvidia.text_embedder",
+        "haystack_integrations.components.embedders.nvidia.truncate",
         "haystack_integrations.components.generators.nvidia.generator",
+        "haystack_integrations.components.rankers.nvidia.ranker",
+        "haystack_integrations.components.rankers.nvidia.truncate",
       ]
     ignore_when_discovered: ["__init__"]
 processors:

{nvidia_haystack-0.0.4 → nvidia_haystack-0.1.0}/pyproject.toml

@@ -42,24 +42,37 @@ root = "../.."
 git_describe_command = 'git describe --tags --match="integrations/nvidia-v[0-9]*"'
 
 [tool.hatch.envs.default]
-dependencies = ["coverage[toml]>=6.5", "pytest", "pytest-rerunfailures", "haystack-pydoc-tools"]
+installer = "uv"
+dependencies = [
+  "coverage[toml]>=6.5",
+  "pytest",
+  "pytest-rerunfailures",
+  "haystack-pydoc-tools",
+  "requests_mock",
+]
 [tool.hatch.envs.default.scripts]
-test = "pytest --reruns 3 --reruns-delay 30 -x {args:tests}"
-test-cov = "coverage run -m pytest --reruns 3 --reruns-delay 30 -x {args:tests}"
+test = "pytest {args:tests}"
+test-cov = "coverage run -m pytest {args:tests}"
+test-cov-retry = "test-cov --reruns 3 --reruns-delay 30 -x"
 cov-report = ["- coverage combine", "coverage report"]
 cov = ["test-cov", "cov-report"]
+cov-retry = ["test-cov-retry", "cov-report"]
 docs = ["pydoc-markdown pydoc/config.yml"]
 
 [[tool.hatch.envs.all.matrix]]
 python = ["3.8", "3.9", "3.10", "3.11"]
 
 [tool.hatch.envs.lint]
+installer = "uv"
 detached = true
-dependencies = ["black>=23.1.0", "mypy>=1.0.0", "ruff>=0.0.243"]
+dependencies = ["pip", "black>=23.1.0", "mypy>=1.0.0", "ruff>=0.0.243"]
 [tool.hatch.envs.lint.scripts]
 typing = "mypy --install-types --non-interactive --explicit-package-bases {args:src/ tests}"
-style = ["ruff check {args:.}", "black --check --diff {args:.}"]
-fmt = ["black {args:.}", "ruff --fix {args:.}", "style"]
+style = [
+  "ruff check {args:}",
+  "black --check --diff {args:.}",
+]
+fmt = ["black {args:.}", "ruff check --fix {args:}", "style"]
 all = ["style", "typing"]
 
 [tool.black]
@@ -70,6 +83,8 @@ skip-string-normalization = true
 [tool.ruff]
 target-version = "py38"
 line-length = 120
+
+[tool.ruff.lint]
 select = [
   "A",
   "ARG",
@@ -118,13 +133,13 @@ unfixable = [
   "F401",
 ]
 
-[tool.ruff.isort]
-known-first-party = ["src"]
+[tool.ruff.lint.isort]
+known-first-party = ["haystack_integrations"]
 
-[tool.ruff.flake8-tidy-imports]
+[tool.ruff.lint.flake8-tidy-imports]
 ban-relative-imports = "parents"
 
-[tool.ruff.per-file-ignores]
+[tool.ruff.lint.per-file-ignores]
 # Tests can use magic values, assertions, and relative imports
 "tests/**/*" = ["PLR2004", "S101", "TID252"]
 
@@ -147,6 +162,8 @@ module = [
   "haystack_integrations.*",
   "pytest.*",
   "numpy.*",
+  "requests_mock.*",
+  "pydantic.*",
 ]
 ignore_missing_imports = true
 

{nvidia_haystack-0.0.4 → nvidia_haystack-0.1.0}/src/haystack_integrations/components/embedders/nvidia/document_embedder.py

@@ -1,13 +1,16 @@
+import warnings
 from typing import Any, Dict, List, Optional, Tuple, Union
 
 from haystack import Document, component, default_from_dict, default_to_dict
 from haystack.utils import Secret, deserialize_secrets_inplace
 from tqdm import tqdm
 
-from ._nim_backend import NimBackend
-from .backend import EmbedderBackend
+from haystack_integrations.utils.nvidia import NimBackend, is_hosted, url_validation
+
 from .truncate import EmbeddingTruncateMode
 
+_DEFAULT_API_URL = "https://ai.api.nvidia.com/v1/retrieval/nvidia"
+
 
 @component
 class NvidiaDocumentEmbedder:
@@ -31,9 +34,9 @@ class NvidiaDocumentEmbedder:
 
     def __init__(
         self,
-        model: str = "NV-Embed-QA",
+        model: Optional[str] = None,
         api_key: Optional[Secret] = Secret.from_env_var("NVIDIA_API_KEY"),
-        api_url: str = "https://ai.api.nvidia.com/v1/retrieval/nvidia",
+        api_url: str = _DEFAULT_API_URL,
         prefix: str = "",
         suffix: str = "",
         batch_size: int = 32,
@@ -47,10 +50,13 @@ class NvidiaDocumentEmbedder:
 
         :param model:
             Embedding model to use.
+            If no specific model along with locally hosted API URL is provided,
+            the system defaults to the available model found using /models API.
         :param api_key:
             API key for the NVIDIA NIM.
         :param api_url:
             Custom API URL for the NVIDIA NIM.
+            Format for API URL is http://host:port
         :param prefix:
             A string to add to the beginning of each text.
         :param suffix:
@@ -71,7 +77,7 @@ class NvidiaDocumentEmbedder:
 
         self.api_key = api_key
         self.model = model
-        self.api_url = api_url
+        self.api_url = url_validation(api_url, _DEFAULT_API_URL, ["v1/embeddings"])
         self.prefix = prefix
         self.suffix = suffix
         self.batch_size = batch_size
@@ -83,9 +89,31 @@ class NvidiaDocumentEmbedder:
             truncate = EmbeddingTruncateMode.from_str(truncate)
         self.truncate = truncate
 
-        self.backend: Optional[EmbedderBackend] = None
+        self.backend: Optional[Any] = None
         self._initialized = False
 
+        if is_hosted(api_url) and not self.model:  # manually set default model
+            self.model = "nvidia/nv-embedqa-e5-v5"
+
+    def default_model(self):
+        """Set default model in local NIM mode."""
+        valid_models = [
+            model.id for model in self.backend.models() if not model.base_model or model.base_model == model.id
+        ]
+        name = next(iter(valid_models), None)
+        if name:
+            warnings.warn(
+                f"Default model is set as: {name}. \n"
+                "Set model using model parameter. \n"
+                "To get available models use available_models property.",
+                UserWarning,
+                stacklevel=2,
+            )
+            self.model = self.backend.model = name
+        else:
+            error_message = "No locally hosted model was found."
+            raise ValueError(error_message)
+
     def warm_up(self):
         """
         Initializes the component.
@@ -105,6 +133,9 @@ class NvidiaDocumentEmbedder:
 
         self._initialized = True
 
+        if not self.model:
+            self.default_model()
+
     def to_dict(self) -> Dict[str, Any]:
         """
         Serializes the component to a dictionary.
@@ -200,6 +231,11 @@ class NvidiaDocumentEmbedder:
             )
             raise TypeError(msg)
 
+        for doc in documents:
+            if not doc.content:
+                msg = f"Document '{doc.id}' has no content to embed."
+                raise ValueError(msg)
+
         texts_to_embed = self._prepare_texts_to_embed(documents)
         embeddings, metadata = self._embed_batch(texts_to_embed, self.batch_size)
         for doc, emb in zip(documents, embeddings):

{nvidia_haystack-0.0.4 → nvidia_haystack-0.1.0}/src/haystack_integrations/components/embedders/nvidia/text_embedder.py

@@ -1,12 +1,15 @@
+import warnings
 from typing import Any, Dict, List, Optional, Union
 
 from haystack import component, default_from_dict, default_to_dict
 from haystack.utils import Secret, deserialize_secrets_inplace
 
-from ._nim_backend import NimBackend
-from .backend import EmbedderBackend
+from haystack_integrations.utils.nvidia import NimBackend, is_hosted, url_validation
+
 from .truncate import EmbeddingTruncateMode
 
+_DEFAULT_API_URL = "https://ai.api.nvidia.com/v1/retrieval/nvidia"
+
 
 @component
 class NvidiaTextEmbedder:
@@ -32,9 +35,9 @@ class NvidiaTextEmbedder:
 
     def __init__(
         self,
-        model: str = "NV-Embed-QA",
+        model: Optional[str] = None,
         api_key: Optional[Secret] = Secret.from_env_var("NVIDIA_API_KEY"),
-        api_url: str = "https://ai.api.nvidia.com/v1/retrieval/nvidia",
+        api_url: str = _DEFAULT_API_URL,
         prefix: str = "",
         suffix: str = "",
         truncate: Optional[Union[EmbeddingTruncateMode, str]] = None,
@@ -44,10 +47,13 @@ class NvidiaTextEmbedder:
 
         :param model:
             Embedding model to use.
+            If no specific model along with locally hosted API URL is provided,
+            the system defaults to the available model found using /models API.
         :param api_key:
             API key for the NVIDIA NIM.
         :param api_url:
             Custom API URL for the NVIDIA NIM.
+            Format for API URL is http://host:port
         :param prefix:
             A string to add to the beginning of each text.
         :param suffix:
@@ -59,7 +65,7 @@ class NvidiaTextEmbedder:
 
         self.api_key = api_key
         self.model = model
-        self.api_url = api_url
+        self.api_url = url_validation(api_url, _DEFAULT_API_URL, ["v1/embeddings"])
         self.prefix = prefix
         self.suffix = suffix
 
@@ -67,9 +73,31 @@ class NvidiaTextEmbedder:
             truncate = EmbeddingTruncateMode.from_str(truncate)
         self.truncate = truncate
 
-        self.backend: Optional[EmbedderBackend] = None
+        self.backend: Optional[Any] = None
         self._initialized = False
 
+        if is_hosted(api_url) and not self.model:  # manually set default model
+            self.model = "nvidia/nv-embedqa-e5-v5"
+
+    def default_model(self):
+        """Set default model in local NIM mode."""
+        valid_models = [
+            model.id for model in self.backend.models() if not model.base_model or model.base_model == model.id
+        ]
+        name = next(iter(valid_models), None)
+        if name:
+            warnings.warn(
+                f"Default model is set as: {name}. \n"
+                "Set model using model parameter. \n"
+                "To get available models use available_models property.",
+                UserWarning,
+                stacklevel=2,
+            )
+            self.model = self.backend.model = name
+        else:
+            error_message = "No locally hosted model was found."
+            raise ValueError(error_message)
+
     def warm_up(self):
         """
         Initializes the component.
@@ -89,6 +117,9 @@ class NvidiaTextEmbedder:
 
         self._initialized = True
 
+        if not self.model:
+            self.default_model()
+
     def to_dict(self) -> Dict[str, Any]:
         """
         Serializes the component to a dictionary.
@@ -144,6 +175,9 @@ class NvidiaTextEmbedder:
                 "In case you want to embed a list of Documents, please use the NvidiaDocumentEmbedder."
             )
             raise TypeError(msg)
+        elif not text:
+            msg = "Cannot embed an empty string."
+            raise ValueError(msg)
 
         assert self.backend is not None
         text_to_embed = self.prefix + text + self.suffix

{nvidia_haystack-0.0.4 → nvidia_haystack-0.1.0}/src/haystack_integrations/components/embedders/nvidia/truncate.py

@@ -6,10 +6,12 @@ class EmbeddingTruncateMode(Enum):
     Specifies how inputs to the NVIDIA embedding components are truncated.
     If START, the input will be truncated from the start.
     If END, the input will be truncated from the end.
+    If NONE, an error will be returned (if the input is too long).
     """
 
     START = "START"
     END = "END"
+    NONE = "NONE"
 
     def __str__(self):
         return self.value

{nvidia_haystack-0.0.4 → nvidia_haystack-0.1.0}/src/haystack_integrations/components/generators/nvidia/generator.py

@@ -1,13 +1,13 @@
 # SPDX-FileCopyrightText: 2024-present deepset GmbH <info@deepset.ai>
 #
 # SPDX-License-Identifier: Apache-2.0
+import warnings
 from typing import Any, Dict, List, Optional
 
 from haystack import component, default_from_dict, default_to_dict
 from haystack.utils.auth import Secret, deserialize_secrets_inplace
 
-from ._nim_backend import NimBackend
-from .backend import GeneratorBackend
+from haystack_integrations.utils.nvidia import NimBackend, is_hosted, url_validation
 
 _DEFAULT_API_URL = "https://integrate.api.nvidia.com/v1"
 
@@ -15,10 +15,11 @@ _DEFAULT_API_URL = "https://integrate.api.nvidia.com/v1"
 @component
 class NvidiaGenerator:
     """
-    A component for generating text using generative models provided by
-    [NVIDIA NIMs](https://ai.nvidia.com).
+    Generates text using generative models hosted with
+    [NVIDIA NIM](https://ai.nvidia.com) on on the [NVIDIA API Catalog](https://build.nvidia.com/explore/discover).
+
+    ### Usage example
 
-    Usage example:
     ```python
     from haystack_integrations.components.generators.nvidia import NvidiaGenerator
 
@@ -37,11 +38,13 @@ class NvidiaGenerator:
     print(result["meta"])
     print(result["usage"])
     ```
+
+    You need an NVIDIA API key for this component to work.
     """
 
     def __init__(
         self,
-        model: str,
+        model: Optional[str] = None,
         api_url: str = _DEFAULT_API_URL,
         api_key: Optional[Secret] = Secret.from_env_var("NVIDIA_API_KEY"),
         model_arguments: Optional[Dict[str, Any]] = None,
@@ -53,21 +56,47 @@ class NvidiaGenerator:
             Name of the model to use for text generation.
             See the [NVIDIA NIMs](https://ai.nvidia.com)
             for more information on the supported models.
+            `Note`: If no specific model along with locally hosted API URL is provided,
+            the system defaults to the available model found using /models API.
+            Check supported models at [NVIDIA NIM](https://ai.nvidia.com).
         :param api_key:
-            API key for the NVIDIA NIM.
+            API key for the NVIDIA NIM. Set it as the `NVIDIA_API_KEY` environment
+            variable or pass it here.
         :param api_url:
             Custom API URL for the NVIDIA NIM.
         :param model_arguments:
-            Additional arguments to pass to the model provider. Different models accept different arguments.
-            Search your model in the [NVIDIA NIMs](https://ai.nvidia.com)
-            to know the supported arguments.
+            Additional arguments to pass to the model provider. These arguments are
+            specific to a model.
+            Search your model in the [NVIDIA NIM](https://ai.nvidia.com)
+            to find the arguments it accepts.
         """
         self._model = model
-        self._api_url = api_url
+        self._api_url = url_validation(api_url, _DEFAULT_API_URL, ["v1/chat/completions"])
         self._api_key = api_key
         self._model_arguments = model_arguments or {}
 
-        self._backend: Optional[GeneratorBackend] = None
+        self._backend: Optional[Any] = None
+
+        self.is_hosted = is_hosted(api_url)
+
+    def default_model(self):
+        """Set default model in local NIM mode."""
+        valid_models = [
+            model.id for model in self._backend.models() if not model.base_model or model.base_model == model.id
+        ]
+        name = next(iter(valid_models), None)
+        if name:
+            warnings.warn(
+                f"Default model is set as: {name}. \n"
+                "Set model using model parameter. \n"
+                "To get available models use available_models property.",
+                UserWarning,
+                stacklevel=2,
+            )
+            self._model = self._backend.model_name = name
+        else:
+            error_message = "No locally hosted model was found."
+            raise ValueError(error_message)
 
     def warm_up(self):
         """
@@ -86,6 +115,9 @@ class NvidiaGenerator:
             model_kwargs=self._model_arguments,
         )
 
+        if not self.is_hosted and not self._model:
+            self.default_model()
+
     def to_dict(self) -> Dict[str, Any]:
         """
         Serializes the component to a dictionary.

nvidia_haystack-0.1.0/src/haystack_integrations/components/rankers/nvidia/__init__.py

@@ -0,0 +1,3 @@
+from .ranker import NvidiaRanker
+
+__all__ = ["NvidiaRanker"]