Notebook Documentation prompt completions fixes

CHANGELOG.md

@@ -1,5 +1,33 @@
 # ChangeLog
 
+## [2024-09-04]
+
+### `llama-index-core` [0.11.5]
+
+- remove unneeded assert in property graph retriever (#15832)
+- make simple property graphs serialize again (#15833)
+- fix json schema for fastapi return types on core components (#15816)
+
+### `llama-index-llms-nvidia` [0.2.2]
+
+- NVIDIA llm: Add Completion for starcoder models (#15802)
+
+### `llama-index-llms-ollama` [0.3.1]
+
+- add ollama response usage (#15773)
+
+### `llama-index-readers-dashscope` [0.2.1]
+
+- fix pydantic v2 validation errors (#15800)
+
+### `llama-index-readers-discord` [0.2.1]
+
+- fix: convert Document id from int to string in DiscordReader (#15806)
+
+### `llama-index-vector-stores-mariadb` [0.1.0]
+
+- Add MariaDB vector store integration package (#15564)
+
 ## [2024-09-02]
 
 ### `llama-index-core` [0.11.4]

docs/docs/CHANGELOG.md

@@ -1,5 +1,33 @@
 # ChangeLog
 
+## [2024-09-04]
+
+### `llama-index-core` [0.11.5]
+
+- remove unneeded assert in property graph retriever (#15832)
+- make simple property graphs serialize again (#15833)
+- fix json schema for fastapi return types on core components (#15816)
+
+### `llama-index-llms-nvidia` [0.2.2]
+
+- NVIDIA llm: Add Completion for starcoder models (#15802)
+
+### `llama-index-llms-ollama` [0.3.1]
+
+- add ollama response usage (#15773)
+
+### `llama-index-readers-dashscope` [0.2.1]
+
+- fix pydantic v2 validation errors (#15800)
+
+### `llama-index-readers-discord` [0.2.1]
+
+- fix: convert Document id from int to string in DiscordReader (#15806)
+
+### `llama-index-vector-stores-mariadb` [0.1.0]
+
+- Add MariaDB vector store integration package (#15564)
+
 ## [2024-09-02]
 
 ### `llama-index-core` [0.11.4]

docs/docs/api_reference/storage/vector_store/mariadb.md

@@ -0,0 +1,4 @@
+::: llama_index.vector_stores.mariadb
+    options:
+      members:
+        - MariaDBVectorStore

docs/mkdocs.yml

@@ -338,6 +338,7 @@ nav:
           - ./examples/llm/nvidia.ipynb
           - ./examples/llm/nvidia_nim.ipynb
           - ./examples/llm/nvidia_tensorrt.ipynb
+          - ./examples/llm/nvidia_text_completion.ipynb
           - ./examples/llm/nvidia_triton.ipynb
           - ./examples/llm/oci_genai.ipynb
           - ./examples/llm/octoai.ipynb
@@ -1470,6 +1471,7 @@ nav:
               - ./api_reference/storage/vector_store/lancedb.md
               - ./api_reference/storage/vector_store/lantern.md
               - ./api_reference/storage/vector_store/lindorm.md
+              - ./api_reference/storage/vector_store/mariadb.md
               - ./api_reference/storage/vector_store/metal.md
               - ./api_reference/storage/vector_store/milvus.md
               - ./api_reference/storage/vector_store/mongodb.md
@@ -2151,6 +2153,7 @@ plugins:
             - ../llama-index-integrations/embeddings/llama-index-embeddings-xinference
             - ../llama-index-integrations/postprocessor/llama-index-postprocessor-xinference-rerank
             - ../llama-index-integrations/selectors/llama-index-selectors-notdiamond
+            - ../llama-index-integrations/vector_stores/llama-index-vector-stores-mariadb
   - redirects:
       redirect_maps:
         ./api/llama_index.vector_stores.MongoDBAtlasVectorSearch.html: api_reference/storage/vector_store/mongodb.md

llama-index-core/llama_index/core/__init__.py

@@ -1,6 +1,6 @@
 """Init file of LlamaIndex."""
 
-__version__ = "0.11.4"
+__version__ = "0.11.5"
 
 import logging
 from logging import NullHandler

llama-index-core/pyproject.toml

@@ -46,7 +46,7 @@ name = "llama-index-core"
 packages = [{include = "llama_index"}]
 readme = "README.md"
 repository = "https://github.com/run-llama/llama_index"
-version = "0.11.4"
+version = "0.11.5"
 
 [tool.poetry.dependencies]
 SQLAlchemy = {extras = ["asyncio"], version = ">=1.4.49"}

llama-index-integrations/embeddings/llama-index-embeddings-nvidia/llama_index/embeddings/nvidia/base.py

@@ -144,9 +144,14 @@ def __init__(
         )
         self._aclient._custom_headers = {"User-Agent": "llama-index-embeddings-nvidia"}
 
-        if not model:
+        self.model = model
+        if self._is_hosted and not self.model:
+            self.model = DEFAULT_MODEL
+        elif not self._is_hosted and not self.model:
             self.__get_default_model()
 
+        self._validate_model(self.model)  ## validate model
+
     def __get_default_model(self) -> None:
         """Set default model."""
         if not self._is_hosted:
@@ -183,6 +188,29 @@ def _validate_url(self, base_url):
             warnings.warn(f"{expected_format} Rest is ignored")
         return base_url.strip("/")
 
+    def _validate_model(self, model_name: str) -> None:
+        """
+        Validates compatibility of the hosted model with the client.
+
+        Args:
+            model_name (str): The name of the model.
+
+        Raises:
+            ValueError: If the model is incompatible with the client.
+        """
+        if self._is_hosted:
+            if model_name not in MODEL_ENDPOINT_MAP:
+                if model_name in [model.id for model in self._client.models.list()]:
+                    warnings.warn(f"Unable to determine validity of {model_name}")
+                else:
+                    raise ValueError(
+                        f"Model {model_name} is incompatible with client {self.class_name()}. "
+                        f"Please check `{self.class_name()}.available_models()`."
+                    )
+        else:
+            if model_name not in [model.id for model in self.available_models]:
+                raise ValueError(f"No locally hosted {model_name} was found.")
+
     @property
     def available_models(self) -> List[Model]:
         """Get available models."""

llama-index-integrations/embeddings/llama-index-embeddings-nvidia/tests/test_embeddings_nvidia.py

@@ -32,14 +32,14 @@ def test_embedding_class():
 def test_nvidia_embedding_param_setting():
     emb = NVIDIAEmbedding(
         api_key="BOGUS",
-        model="test-model",
+        model="NV-Embed-QA",
         truncate="END",
         timeout=20,
         max_retries=10,
         embed_batch_size=15,
     )
 
-    assert emb.model == "test-model"
+    assert emb.model == "NV-Embed-QA"
     assert emb.truncate == "END"
     assert emb._client.timeout == 20
     assert emb._client.max_retries == 10
@@ -90,3 +90,28 @@ def test_nvidia_embedding_callback(mock_integration_api):
 def test_nvidia_embedding_throws_with_invalid_key(mock_integration_api):
     emb = NVIDIAEmbedding(api_key="invalid")
     emb.get_text_embedding("hi")
+
+
+# @pytest.mark.parametrize("model", list(MODEL_ENDPOINT_MAP.keys()))
+# def test_model_compatible_client_model(model: str) -> None:
+#     NVIDIAEmbedding(api_key="BOGUS", model=model)
+
+
+def test_model_incompatible_client_model() -> None:
+    model_name = "x"
+    err_msg = (
+        f"Model {model_name} is incompatible with client NVIDIAEmbedding. "
+        f"Please check `NVIDIAEmbedding.available_models()`."
+    )
+    with pytest.raises(ValueError) as msg:
+        NVIDIAEmbedding(api_key="BOGUS", model=model_name)
+    assert err_msg == str(msg.value)
+
+
+def test_model_incompatible_client_known_model() -> None:
+    model_name = "google/deplot"
+    warn_msg = f"Unable to determine validity"
+    with pytest.warns(UserWarning) as msg:
+        NVIDIAEmbedding(api_key="BOGUS", model=model_name)
+    assert len(msg) == 1
+    assert warn_msg in str(msg[0].message)

llama-index-integrations/llms/llama-index-llms-nvidia/llama_index/llms/nvidia/base.py

@@ -11,6 +11,7 @@
 from llama_index.llms.nvidia.utils import (
     is_nvidia_function_calling_model,
     is_chat_model,
+    ALL_MODELS,
 )
 
 from llama_index.llms.openai_like import OpenAILike
@@ -110,17 +111,20 @@ def __init__(
             is_function_calling_model=is_nvidia_function_calling_model(model),
             **kwargs,
         )
-        self.model = model
         self._is_hosted = base_url in KNOWN_URLS
 
         if self._is_hosted and api_key == "NO_API_KEY_PROVIDED":
             warnings.warn(
                 "An API key is required for the hosted NIM. This will become an error in 0.2.0.",
             )
-
-        if not model:
+        self.model = model
+        if self._is_hosted and not self.model:
+            self.model = DEFAULT_MODEL
+        elif not self._is_hosted and not self.model:
             self.__get_default_model()
 
+        self._validate_model(self.model)  ## validate model
+
     def __get_default_model(self):
         """Set default model."""
         if not self._is_hosted:
@@ -163,6 +167,29 @@ def _validate_url(self, base_url):
                 raise ValueError(f"Invalid base_url, {expected_format}")
         return urlunparse((result.scheme, result.netloc, "v1", "", "", ""))
 
+    def _validate_model(self, model_name: str) -> None:
+        """
+        Validates compatibility of the hosted model with the client.
+
+        Args:
+            model_name (str): The name of the model.
+
+        Raises:
+            ValueError: If the model is incompatible with the client.
+        """
+        if self._is_hosted:
+            if model_name not in ALL_MODELS:
+                if model_name in [model.id for model in self.available_models]:
+                    warnings.warn(f"Unable to determine validity of {model_name}")
+                else:
+                    raise ValueError(
+                        f"Model {model_name} is incompatible with client {self.class_name()}. "
+                        f"Please check `{self.class_name()}.available_models()`."
+                    )
+        else:
+            if model_name not in [model.id for model in self.available_models]:
+                raise ValueError(f"No locally hosted {model_name} was found.")
+
     @property
     def available_models(self) -> List[Model]:
         models = [

llama-index-integrations/llms/llama-index-llms-nvidia/llama_index/llms/nvidia/utils.py

@@ -42,6 +42,10 @@
 
 COMPLETION_MODELS = ("bigcode/starcoder2-7b", "bigcode/starcoder2-15b")
 
+ALL_MODELS = (
+    tuple(API_CATALOG_MODELS.keys()) + NVIDIA_FUNTION_CALLING_MODELS + COMPLETION_MODELS
+)
+
 
 def is_chat_model(modelname: str):
     return modelname not in COMPLETION_MODELS

llama-index-integrations/llms/llama-index-llms-nvidia/tests/test_mode_switch.py

@@ -17,7 +17,7 @@ def mock_unknown_urls(httpx_mock: HTTPXMock, base_url: str):
     mock_response = {
         "data": [
             {
-                "id": "dummy",
+                "id": "meta/llama3-8b-instruct",
                 "object": "model",
                 "created": 1234567890,
                 "owned_by": "OWNER",
@@ -63,16 +63,18 @@ def test_mode_switch_nim_with_url_deprecated():
 
 @pytest.mark.parametrize("base_url", ["https://test_url/v1/"])
 def test_mode_switch_param_setting_deprecated(base_url):
-    instance = Interface(model="dummy")
+    instance = Interface(model="meta/llama3-8b-instruct")
 
     with pytest.warns(DeprecationWarning):
         instance1 = instance.mode("nim", base_url=base_url)
-    assert instance1.model == "dummy"
+    assert instance1.model == "meta/llama3-8b-instruct"
     assert str(instance1.api_base) == base_url
 
     with pytest.warns(DeprecationWarning):
-        instance2 = instance1.mode("nvidia", api_key="test", model="dummy-2")
-    assert instance2.model == "dummy-2"
+        instance2 = instance1.mode(
+            "nvidia", api_key="test", model="meta/llama3-15b-instruct"
+        )
+    assert instance2.model == "meta/llama3-15b-instruct"
     assert str(instance2.api_base) == BASE_URL
     assert instance2.api_key == "test"