Merge branch 'main' into azure-dyn-route-fix

README.md

@@ -15,6 +15,7 @@
 
 Semantic Router is a superfast decision-making layer for your LLMs and agents. Rather than waiting for slow LLM generations to make tool-use decisions, we use the magic of semantic vector space to make those decisions — _routing_ our requests using _semantic_ meaning.
 
+
 ---
 
 ## Quickstart
@@ -25,7 +26,7 @@ To get started with _semantic-router_ we install it like so:
 pip install -qU semantic-router
 ```
 
-❗️ _If wanting to use local embeddings you can use `FastEmbedEncoder` (`pip install -qU "semantic-router[fastembed]`"). To use the `HybridRouteLayer` you must `pip install -qU "semantic-router[hybrid]"`._
+❗️ _If wanting to use a fully local version of semantic router you can use `HuggingFaceEncoder` and `LlamaCppEncoder` (`pip install -qU "semantic-router[local]"`, see [here](https://github.com/aurelio-labs/semantic-router/blob/main/docs/05-local-execution.ipynb)). To use the `HybridRouteLayer` you must `pip install -qU "semantic-router[hybrid]"`._
 
 We begin by defining a set of `Route` objects. These are the decision paths that the semantic router can decide to use, let's try two simple routes for now — one for talk on _politics_ and another for _chitchat_:
 

docs/00-introduction.ipynb

@@ -41,7 +41,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "!pip install -qU semantic-router==0.0.16"
+    "!pip install -qU semantic-router==0.0.17"
    ]
   },
   {

docs/01-save-load-from-file.ipynb

@@ -36,7 +36,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "!pip install -qU semantic-router==0.0.16"
+    "!pip install -qU semantic-router==0.0.17"
    ]
   },
   {

docs/02-dynamic-routes.ipynb

@@ -48,7 +48,7 @@
       },
       "outputs": [],
       "source": [
-        "!pip install -qU semantic-router==0.0.16"
+        "!pip install -qU semantic-router==0.0.17"
       ]
     },
     {

docs/03-basic-langchain-agent.ipynb

@@ -78,7 +78,7 @@
    ],
    "source": [
     "!pip install -qU \\\n",
-    "    semantic-router==0.0.16 \\\n",
+    "    semantic-router==0.0.17 \\\n",
     "    langchain==0.0.352 \\\n",
     "    openai==1.6.1"
    ]

docs/05-local-execution.ipynb

@@ -41,7 +41,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "!pip install -qU \"semantic-router[local]==0.0.16\""
+    "!pip install -qU \"semantic-router[local]==0.0.17\""
    ]
   },
   {
@@ -342,7 +342,7 @@
     "from semantic_router import RouteLayer\n",
     "\n",
     "from llama_cpp import Llama\n",
-    "from semantic_router.llms import LlamaCppLLM\n",
+    "from semantic_router.llms.llamacpp import LlamaCppLLM\n",
     "\n",
     "enable_gpu = True  # offload LLM layers to the GPU (must fit in memory)\n",
     "\n",

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "semantic-router"
-version = "0.0.16"
+version = "0.0.17"
 description = "Super fast semantic router for AI decision making"
 authors = [
     "James Briggs <james@aurelio.ai>",
@@ -16,7 +16,7 @@ packages = [{include = "semantic_router"}]
 
 [tool.poetry.dependencies]
 python = "^3.9"
-pydantic = "^1.8.2"
+pydantic = "^2.5.3"
 openai = "^1.3.9"
 cohere = "^4.32"
 numpy = "^1.25.2"

semantic_router/__init__.py

@@ -3,3 +3,5 @@
 from semantic_router.route import Route
 
 __all__ = ["RouteLayer", "HybridRouteLayer", "Route", "LayerConfig"]
+
+__version__ = "0.0.17"

semantic_router/encoders/base.py

@@ -1,6 +1,6 @@
 from typing import List
 
-from pydantic import BaseModel, Field
+from pydantic.v1 import BaseModel, Field
 
 
 class BaseEncoder(BaseModel):

semantic_router/encoders/cohere.py

@@ -25,7 +25,9 @@ def __init__(
         try:
             self.client = cohere.Client(cohere_api_key)
         except Exception as e:
-            raise ValueError(f"Cohere API client failed to initialize. Error: {e}")
+            raise ValueError(
+                f"Cohere API client failed to initialize. Error: {e}"
+            ) from e
 
     def __call__(self, docs: List[str]) -> List[List[float]]:
         if self.client is None:
@@ -34,4 +36,4 @@ def __call__(self, docs: List[str]) -> List[List[float]]:
             embeds = self.client.embed(docs, input_type="search_query", model=self.name)
             return embeds.embeddings
         except Exception as e:
-            raise ValueError(f"Cohere API call failed. Error: {e}")
+            raise ValueError(f"Cohere API call failed. Error: {e}") from e

semantic_router/encoders/fastembed.py

@@ -1,7 +1,7 @@
 from typing import Any, List, Optional
 
 import numpy as np
-from pydantic import PrivateAttr
+from pydantic.v1 import PrivateAttr
 
 from semantic_router.encoders import BaseEncoder
 
@@ -48,4 +48,4 @@ def __call__(self, docs: List[str]) -> List[List[float]]:
             embeddings: List[List[float]] = [e.tolist() for e in embeds]
             return embeddings
         except Exception as e:
-            raise ValueError(f"FastEmbed embed failed. Error: {e}")
+            raise ValueError(f"FastEmbed embed failed. Error: {e}") from e

semantic_router/encoders/huggingface.py

@@ -1,6 +1,6 @@
 from typing import Any, List, Optional
 
-from pydantic import PrivateAttr
+from pydantic.v1 import PrivateAttr
 
 from semantic_router.encoders import BaseEncoder
 

semantic_router/encoders/openai.py

@@ -29,7 +29,9 @@ def __init__(
         try:
             self.client = openai.Client(api_key=api_key)
         except Exception as e:
-            raise ValueError(f"OpenAI API client failed to initialize. Error: {e}")
+            raise ValueError(
+                f"OpenAI API client failed to initialize. Error: {e}"
+            ) from e
 
     def __call__(self, docs: List[str]) -> List[List[float]]:
         if self.client is None:
@@ -49,7 +51,7 @@ def __call__(self, docs: List[str]) -> List[List[float]]:
                 logger.warning(f"Retrying in {2**j} seconds...")
             except Exception as e:
                 logger.error(f"OpenAI API call failed. Error: {error_message}")
-                raise ValueError(f"OpenAI API call failed. Error: {e}")
+                raise ValueError(f"OpenAI API call failed. Error: {e}") from e
 
         if (
             not embeds

semantic_router/encoders/zure.py

@@ -74,7 +74,9 @@ def __init__(
                 # _strict_response_validation=True,
             )
         except Exception as e:
-            raise ValueError(f"OpenAI API client failed to initialize. Error: {e}")
+            raise ValueError(
+                f"OpenAI API client failed to initialize. Error: {e}"
+            ) from e
 
     def __call__(self, docs: List[str]) -> List[List[float]]:
         if self.client is None:
@@ -100,7 +102,7 @@ def __call__(self, docs: List[str]) -> List[List[float]]:
                 logger.warning(f"Retrying in {2**j} seconds...")
             except Exception as e:
                 logger.error(f"Azure OpenAI API call failed. Error: {error_message}")
-                raise ValueError(f"Azure OpenAI API call failed. Error: {e}")
+                raise ValueError(f"Azure OpenAI API call failed. Error: {e}") from e
 
         if (
             not embeds

semantic_router/llms/__init__.py

@@ -1,15 +1,7 @@
 from semantic_router.llms.base import BaseLLM
 from semantic_router.llms.cohere import CohereLLM
-from semantic_router.llms.llamacpp import LlamaCppLLM
 from semantic_router.llms.openai import OpenAILLM
 from semantic_router.llms.openrouter import OpenRouterLLM
 from semantic_router.llms.zure import AzureOpenAILLM
 
-__all__ = [
-    "BaseLLM",
-    "OpenAILLM",
-    "OpenRouterLLM",
-    "CohereLLM",
-    "LlamaCppLLM",
-    "AzureOpenAILLM",
-]
+__all__ = ["BaseLLM", "OpenAILLM", "OpenRouterLLM", "CohereLLM"]

semantic_router/llms/base.py

@@ -1,7 +1,7 @@
 import json
 from typing import Any, List, Optional
 
-from pydantic import BaseModel
+from pydantic.v1 import BaseModel
 
 from semantic_router.schema import Message
 from semantic_router.utils.logger import logger

semantic_router/llms/cohere.py

@@ -24,7 +24,9 @@ def __init__(
         try:
             self.client = cohere.Client(cohere_api_key)
         except Exception as e:
-            raise ValueError(f"Cohere API client failed to initialize. Error: {e}")
+            raise ValueError(
+                f"Cohere API client failed to initialize. Error: {e}"
+            ) from e
 
     def __call__(self, messages: List[Message]) -> str:
         if self.client is None:
@@ -43,4 +45,4 @@ def __call__(self, messages: List[Message]) -> str:
             return output
 
         except Exception as e:
-            raise ValueError(f"Cohere API call failed. Error: {e}")
+            raise ValueError(f"Cohere API call failed. Error: {e}") from e

semantic_router/llms/openai.py

@@ -29,7 +29,9 @@ def __init__(
         try:
             self.client = openai.OpenAI(api_key=api_key)
         except Exception as e:
-            raise ValueError(f"OpenAI API client failed to initialize. Error: {e}")
+            raise ValueError(
+                f"OpenAI API client failed to initialize. Error: {e}"
+            ) from e
         self.temperature = temperature
         self.max_tokens = max_tokens
 
@@ -51,4 +53,4 @@ def __call__(self, messages: List[Message]) -> str:
             return output
         except Exception as e:
             logger.error(f"LLM error: {e}")
-            raise Exception(f"LLM error: {e}")
+            raise Exception(f"LLM error: {e}") from e

semantic_router/llms/openrouter.py

@@ -34,7 +34,9 @@ def __init__(
         try:
             self.client = openai.OpenAI(api_key=api_key, base_url=self.base_url)
         except Exception as e:
-            raise ValueError(f"OpenRouter API client failed to initialize. Error: {e}")
+            raise ValueError(
+                f"OpenRouter API client failed to initialize. Error: {e}"
+            ) from e
         self.temperature = temperature
         self.max_tokens = max_tokens
 
@@ -56,4 +58,4 @@ def __call__(self, messages: List[Message]) -> str:
             return output
         except Exception as e:
             logger.error(f"LLM error: {e}")
-            raise Exception(f"LLM error: {e}")
+            raise Exception(f"LLM error: {e}") from e

semantic_router/route.py

@@ -2,7 +2,7 @@
 import re
 from typing import Any, Callable, Dict, List, Optional, Union
 
-from pydantic import BaseModel
+from pydantic.v1 import BaseModel
 
 from semantic_router.llms import BaseLLM
 from semantic_router.schema import Message, RouteChoice

semantic_router/schema.py

@@ -1,8 +1,8 @@
 from enum import Enum
 from typing import Dict, List, Literal, Optional
 
-from pydantic import BaseModel
-from pydantic.dataclasses import dataclass
+from pydantic.v1 import BaseModel
+from pydantic.v1.dataclasses import dataclass
 
 from semantic_router.encoders import (
     BaseEncoder,

semantic_router/utils/function_call.py

@@ -1,7 +1,7 @@
 import inspect
 from typing import Any, Callable, Dict, List, Union
 
-from pydantic import BaseModel
+from pydantic.v1 import BaseModel
 
 from semantic_router.llms import BaseLLM
 from semantic_router.schema import Message, RouteChoice

semantic_router/utils/llm.py

@@ -32,7 +32,7 @@ def llm(prompt: str) -> Optional[str]:
         return output
     except Exception as e:
         logger.error(f"LLM error: {e}")
-        raise Exception(f"LLM error: {e}")
+        raise Exception(f"LLM error: {e}") from e
 
 
 # TODO integrate async LLM function
@@ -62,4 +62,4 @@ def llm(prompt: str) -> Optional[str]:
 #         return output
 #     except Exception as e:
 #         logger.error(f"LLM error: {e}")
-#         raise Exception(f"LLM error: {e}")
+#         raise Exception(f"LLM error: {e}") from e

tests/unit/llms/test_llm_llamacpp.py

@@ -1,7 +1,7 @@
 import pytest
 from llama_cpp import Llama
 
-from semantic_router.llms import LlamaCppLLM
+from semantic_router.llms.llamacpp import LlamaCppLLM
 from semantic_router.schema import Message
 
 

tests/unit/test_schema.py

@@ -1,5 +1,5 @@
 import pytest
-from pydantic import ValidationError
+from pydantic.v1 import ValidationError
 
 from semantic_router.schema import (
     CohereEncoder,