agno 1.8.1__py3-none-any.whl → 2.0.0a1__py3-none-any.whl

agno/{embedder → knowledge/embedder}/google.py

@@ -2,7 +2,7 @@ from dataclasses import dataclass
 from os import getenv
 from typing import Any, Dict, List, Optional, Tuple
 
-from agno.embedder.base import Embedder
+from agno.knowledge.embedder.base import Embedder
 from agno.utils.log import log_error, log_info
 
 try:
@@ -56,6 +56,11 @@ class GeminiEmbedder(Embedder):
 
         return self.gemini_client
 
+    @property
+    def aclient(self) -> GeminiClient:
+        """Returns the same client instance since Google GenAI Client supports both sync and async operations."""
+        return self.client
+
     def _response(self, text: str) -> EmbedContentResponse:
         # If a user provides a model id with the `models/` prefix, we need to remove it
         _id = self.id
@@ -105,3 +110,71 @@ class GeminiEmbedder(Embedder):
         except Exception as e:
             log_error(f"Error extracting embeddings: {e}")
             return [], usage
+
+    async def async_get_embedding(self, text: str) -> List[float]:
+        """Async version of get_embedding using client.aio."""
+        # If a user provides a model id with the `models/` prefix, we need to remove it
+        _id = self.id
+        if _id.startswith("models/"):
+            _id = _id.split("/")[-1]
+
+        _request_params: Dict[str, Any] = {"contents": text, "model": _id, "config": {}}
+        if self.dimensions:
+            _request_params["config"]["output_dimensionality"] = self.dimensions
+        if self.task_type:
+            _request_params["config"]["task_type"] = self.task_type
+        if self.title:
+            _request_params["config"]["title"] = self.title
+        if not _request_params["config"]:
+            del _request_params["config"]
+
+        if self.request_params:
+            _request_params.update(self.request_params)
+
+        try:
+            response = await self.aclient.aio.models.embed_content(**_request_params)
+            if response.embeddings and len(response.embeddings) > 0:
+                values = response.embeddings[0].values
+                if values is not None:
+                    return values
+            log_info("No embeddings found in response")
+            return []
+        except Exception as e:
+            log_error(f"Error extracting embeddings: {e}")
+            return []
+
+    async def async_get_embedding_and_usage(self, text: str) -> Tuple[List[float], Optional[Dict[str, Any]]]:
+        """Async version of get_embedding_and_usage using client.aio."""
+        # If a user provides a model id with the `models/` prefix, we need to remove it
+        _id = self.id
+        if _id.startswith("models/"):
+            _id = _id.split("/")[-1]
+
+        _request_params: Dict[str, Any] = {"contents": text, "model": _id, "config": {}}
+        if self.dimensions:
+            _request_params["config"]["output_dimensionality"] = self.dimensions
+        if self.task_type:
+            _request_params["config"]["task_type"] = self.task_type
+        if self.title:
+            _request_params["config"]["title"] = self.title
+        if not _request_params["config"]:
+            del _request_params["config"]
+
+        if self.request_params:
+            _request_params.update(self.request_params)
+
+        try:
+            response = await self.aclient.aio.models.embed_content(**_request_params)
+            usage = None
+            if response.metadata and hasattr(response.metadata, "billable_character_count"):
+                usage = {"billable_character_count": response.metadata.billable_character_count}
+
+            if response.embeddings and len(response.embeddings) > 0:
+                values = response.embeddings[0].values
+                if values is not None:
+                    return values, usage
+            log_info("No embeddings found in response")
+            return [], usage
+        except Exception as e:
+            log_error(f"Error extracting embeddings: {e}")
+            return [], usage

agno/{embedder → knowledge/embedder}/huggingface.py

@@ -2,11 +2,11 @@ from dataclasses import dataclass
 from os import getenv
 from typing import Any, Dict, List, Optional, Tuple
 
-from agno.embedder.base import Embedder
+from agno.knowledge.embedder.base import Embedder
 from agno.utils.log import logger
 
 try:
-    from huggingface_hub import InferenceClient
+    from huggingface_hub import AsyncInferenceClient, InferenceClient
 except ImportError:
     logger.error("`huggingface-hub` not installed, please run `pip install huggingface-hub`")
     raise
@@ -20,6 +20,7 @@ class HuggingfaceCustomEmbedder(Embedder):
     api_key: Optional[str] = getenv("HUGGINGFACE_API_KEY")
     client_params: Optional[Dict[str, Any]] = None
     huggingface_client: Optional[InferenceClient] = None
+    async_client: Optional[AsyncInferenceClient] = None
 
     @property
     def client(self) -> InferenceClient:
@@ -33,6 +34,18 @@ class HuggingfaceCustomEmbedder(Embedder):
         self.huggingface_client = InferenceClient(**_client_params)
         return self.huggingface_client
 
+    @property
+    def aclient(self) -> AsyncInferenceClient:
+        if self.async_client:
+            return self.async_client
+        _client_params: Dict[str, Any] = {}
+        if self.api_key:
+            _client_params["api_key"] = self.api_key
+        if self.client_params:
+            _client_params.update(self.client_params)
+        self.async_client = AsyncInferenceClient(**_client_params)
+        return self.async_client
+
     def _response(self, text: str):
         return self.client.feature_extraction(text=text, model=self.id)
 
@@ -53,3 +66,24 @@ class HuggingfaceCustomEmbedder(Embedder):
 
     def get_embedding_and_usage(self, text: str) -> Tuple[List[float], Optional[Dict]]:
         return self.get_embedding(text=text), None
+
+    async def async_get_embedding(self, text: str) -> List[float]:
+        """Async version of get_embedding using AsyncInferenceClient."""
+        response = await self.aclient.feature_extraction(text=text, model=self.id)
+        try:
+            # If already a list, return directly
+            if isinstance(response, list):
+                return response
+            # If numpy array, convert to list
+            elif hasattr(response, "tolist"):
+                return response.tolist()
+            else:
+                return list(response)
+        except Exception as e:
+            logger.warning(f"Failed to process embeddings: {e}")
+            return []
+
+    async def async_get_embedding_and_usage(self, text: str) -> Tuple[List[float], Optional[Dict]]:
+        """Async version of get_embedding_and_usage."""
+        embedding = await self.async_get_embedding(text=text)
+        return embedding, None

agno/{embedder → knowledge/embedder}/jina.py

@@ -4,13 +4,18 @@ from typing import Any, Dict, List, Optional, Tuple
 
 from typing_extensions import Literal
 
-from agno.embedder.base import Embedder
+from agno.knowledge.embedder.base import Embedder
 from agno.utils.log import logger
 
 try:
     import requests
 except ImportError:
-    raise ImportError("requests not installed, use pip install requests")
+    raise ImportError("`requests` not installed, use pip install requests")
+
+try:
+    import aiohttp
+except ImportError:
+    raise ImportError("`aiohttp` not installed, use pip install aiohttp")
 
 
 @dataclass
@@ -71,3 +76,44 @@ class JinaEmbedder(Embedder):
         except Exception as e:
             logger.warning(f"Failed to get embedding and usage: {e}")
             return [], None
+
+    async def _async_response(self, text: str) -> Dict[str, Any]:
+        """Async version of _response using aiohttp."""
+        data = {
+            "model": self.id,
+            "late_chunking": self.late_chunking,
+            "dimensions": self.dimensions,
+            "embedding_type": self.embedding_type,
+            "input": [text],  # Jina API expects a list
+        }
+        if self.user is not None:
+            data["user"] = self.user
+        if self.request_params:
+            data.update(self.request_params)
+
+        timeout = aiohttp.ClientTimeout(total=self.timeout) if self.timeout else None
+
+        async with aiohttp.ClientSession(timeout=timeout) as session:
+            async with session.post(self.base_url, headers=self._get_headers(), json=data) as response:
+                response.raise_for_status()
+                return await response.json()
+
+    async def async_get_embedding(self, text: str) -> List[float]:
+        """Async version of get_embedding."""
+        try:
+            result = await self._async_response(text)
+            return result["data"][0]["embedding"]
+        except Exception as e:
+            logger.warning(f"Failed to get embedding: {e}")
+            return []
+
+    async def async_get_embedding_and_usage(self, text: str) -> Tuple[List[float], Optional[Dict]]:
+        """Async version of get_embedding_and_usage."""
+        try:
+            result = await self._async_response(text)
+            embedding = result["data"][0]["embedding"]
+            usage = result.get("usage")
+            return embedding, usage
+        except Exception as e:
+            logger.warning(f"Failed to get embedding and usage: {e}")
+            return [], None

agno/knowledge/embedder/langdb.py

@@ -0,0 +1,22 @@
+from dataclasses import dataclass
+from os import getenv
+from typing import Optional
+
+from agno.knowledge.embedder.openai import OpenAIEmbedder
+
+
+@dataclass
+class LangDBEmbedder(OpenAIEmbedder):
+    id: str = "text-embedding-ada-002"
+    dimensions: int = 1536
+    api_key: Optional[str] = getenv("LANGDB_API_KEY")
+    project_id: Optional[str] = getenv("LANGDB_PROJECT_ID")
+    base_url: Optional[str] = None
+
+    def __post_init__(self):
+        """Set the base_url based on project_id if not provided."""
+        if not self.project_id:
+            raise ValueError("LANGDB_PROJECT_ID not set in the environment")
+
+        if not self.base_url:
+            self.base_url = f"https://api.us-east-1.langdb.ai/{self.project_id}/v1"

agno/knowledge/embedder/mistral.py

@@ -0,0 +1,139 @@
+from dataclasses import dataclass
+from os import getenv
+from typing import Any, Dict, List, Optional, Tuple
+
+from agno.knowledge.embedder.base import Embedder
+from agno.utils.log import logger
+
+try:
+    from mistralai import Mistral  # type: ignore
+    from mistralai.models.embeddingresponse import EmbeddingResponse  # type: ignore
+except ImportError:
+    logger.error("`mistralai` not installed")
+    raise
+
+
+@dataclass
+class MistralEmbedder(Embedder):
+    id: str = "mistral-embed"
+    dimensions: int = 1024
+    # -*- Request parameters
+    request_params: Optional[Dict[str, Any]] = None
+    # -*- Client parameters
+    api_key: Optional[str] = getenv("MISTRAL_API_KEY")
+    endpoint: Optional[str] = None
+    max_retries: Optional[int] = None
+    timeout: Optional[int] = None
+    client_params: Optional[Dict[str, Any]] = None
+    # -*- Provide the Mistral Client manually
+    mistral_client: Optional[Mistral] = None
+
+    @property
+    def client(self) -> Mistral:
+        if self.mistral_client:
+            return self.mistral_client
+
+        _client_params: Dict[str, Any] = {
+            "api_key": self.api_key,
+            "endpoint": self.endpoint,
+            "max_retries": self.max_retries,
+            "timeout": self.timeout,
+        }
+        _client_params = {k: v for k, v in _client_params.items() if v is not None}
+
+        if self.client_params:
+            _client_params.update(self.client_params)
+
+        self.mistral_client = Mistral(**_client_params)
+
+        return self.mistral_client
+
+    def _response(self, text: str) -> EmbeddingResponse:
+        _request_params: Dict[str, Any] = {
+            "inputs": text,
+            "model": self.id,
+        }
+        if self.request_params:
+            _request_params.update(self.request_params)
+        response = self.client.embeddings.create(**_request_params)
+        if response is None:
+            raise ValueError("Failed to get embedding response")
+        return response
+
+    def get_embedding(self, text: str) -> List[float]:
+        try:
+            response: EmbeddingResponse = self._response(text=text)
+            if response.data and response.data[0].embedding:
+                return response.data[0].embedding
+            return []
+        except Exception as e:
+            logger.warning(f"Error getting embedding: {e}")
+            return []
+
+    def get_embedding_and_usage(self, text: str) -> Tuple[List[float], Dict[str, Any]]:
+        try:
+            response: EmbeddingResponse = self._response(text=text)
+            embedding: List[float] = (
+                response.data[0].embedding if (response.data and response.data[0].embedding) else []
+            )
+            usage: Dict[str, Any] = response.usage.model_dump() if response.usage else {}
+            return embedding, usage
+        except Exception as e:
+            logger.warning(f"Error getting embedding and usage: {e}")
+            return [], {}
+
+    async def async_get_embedding(self, text: str) -> List[float]:
+        """Async version of get_embedding."""
+        try:
+            # Check if the client has an async version of embeddings.create
+            if hasattr(self.client.embeddings, "create_async"):
+                response: EmbeddingResponse = await self.client.embeddings.create_async(
+                    inputs=text, model=self.id, **self.request_params if self.request_params else {}
+                )
+            else:
+                # Fallback to running sync method in thread executor
+                import asyncio
+
+                loop = asyncio.get_running_loop()
+                response: EmbeddingResponse = await loop.run_in_executor(  # type: ignore
+                    None,
+                    lambda: self.client.embeddings.create(
+                        inputs=text, model=self.id, **self.request_params if self.request_params else {}
+                    ),
+                )
+
+            if response.data and response.data[0].embedding:
+                return response.data[0].embedding
+            return []
+        except Exception as e:
+            logger.warning(f"Error getting embedding: {e}")
+            return []
+
+    async def async_get_embedding_and_usage(self, text: str) -> Tuple[List[float], Dict[str, Any]]:
+        """Async version of get_embedding_and_usage."""
+        try:
+            # Check if the client has an async version of embeddings.create
+            if hasattr(self.client.embeddings, "create_async"):
+                response: EmbeddingResponse = await self.client.embeddings.create_async(
+                    inputs=text, model=self.id, **self.request_params if self.request_params else {}
+                )
+            else:
+                # Fallback to running sync method in thread executor
+                import asyncio
+
+                loop = asyncio.get_running_loop()
+                response: EmbeddingResponse = await loop.run_in_executor(  # type: ignore
+                    None,
+                    lambda: self.client.embeddings.create(
+                        inputs=text, model=self.id, **self.request_params if self.request_params else {}
+                    ),
+                )
+
+            embedding: List[float] = (
+                response.data[0].embedding if (response.data and response.data[0].embedding) else []
+            )
+            usage: Dict[str, Any] = response.usage.model_dump() if response.usage else {}
+            return embedding, usage
+        except Exception as e:
+            logger.warning(f"Error getting embedding and usage: {e}")
+            return [], {}

agno/{embedder → knowledge/embedder}/nebius.py

@@ -2,7 +2,7 @@ from dataclasses import dataclass
 from os import getenv
 from typing import Optional
 
-from agno.embedder.openai import OpenAIEmbedder
+from agno.knowledge.embedder.openai import OpenAIEmbedder
 
 
 @dataclass

agno/{embedder → knowledge/embedder}/ollama.py

@@ -1,12 +1,13 @@
 from dataclasses import dataclass
 from typing import Any, Dict, List, Optional, Tuple
 
-from agno.embedder.base import Embedder
-from agno.utils.log import logger
+from agno.knowledge.embedder.base import Embedder
+from agno.utils.log import log_error, logger
 
 try:
     import importlib.metadata as metadata
 
+    from ollama import AsyncClient as AsyncOllamaClient
     from ollama import Client as OllamaClient
     from packaging import version
 
@@ -30,7 +31,7 @@ except ImportError as e:
 
 except Exception as e:
     # Catch-all for unexpected errors
-    print(f"An unexpected error occurred: {e}")
+    log_error(f"An unexpected error occurred: {e}")
 
 
 @dataclass
@@ -42,6 +43,7 @@ class OllamaEmbedder(Embedder):
     options: Optional[Any] = None
     client_kwargs: Optional[Dict[str, Any]] = None
     ollama_client: Optional[OllamaClient] = None
+    async_client: Optional[AsyncOllamaClient] = None
 
     @property
     def client(self) -> OllamaClient:
@@ -58,6 +60,21 @@ class OllamaEmbedder(Embedder):
         self.ollama_client = OllamaClient(**_ollama_params)
         return self.ollama_client
 
+    @property
+    def aclient(self) -> AsyncOllamaClient:
+        if self.async_client:
+            return self.async_client
+
+        _ollama_params: Dict[str, Any] = {
+            "host": self.host,
+            "timeout": self.timeout,
+        }
+        _ollama_params = {k: v for k, v in _ollama_params.items() if v is not None}
+        if self.client_kwargs:
+            _ollama_params.update(self.client_kwargs)
+        self.async_client = AsyncOllamaClient(**_ollama_params)
+        return self.async_client
+
     def _response(self, text: str) -> Dict[str, Any]:
         kwargs: Dict[str, Any] = {}
         if self.options is not None:
@@ -88,3 +105,37 @@ class OllamaEmbedder(Embedder):
         embedding = self.get_embedding(text=text)
         usage = None
         return embedding, usage
+
+    async def _async_response(self, text: str) -> Dict[str, Any]:
+        """Async version of _response using AsyncOllamaClient."""
+        kwargs: Dict[str, Any] = {}
+        if self.options is not None:
+            kwargs["options"] = self.options
+
+        response = await self.aclient.embed(input=text, model=self.id, **kwargs)
+        if response and "embeddings" in response:
+            embeddings = response["embeddings"]
+            if isinstance(embeddings, list) and len(embeddings) > 0 and isinstance(embeddings[0], list):
+                return {"embeddings": embeddings[0]}  # Use the first element
+            elif isinstance(embeddings, list) and all(isinstance(x, (int, float)) for x in embeddings):
+                return {"embeddings": embeddings}  # Return as-is if already flat
+        return {"embeddings": []}  # Return an empty list if no valid embedding is found
+
+    async def async_get_embedding(self, text: str) -> List[float]:
+        """Async version of get_embedding."""
+        try:
+            response = await self._async_response(text=text)
+            embedding = response.get("embeddings", [])
+            if len(embedding) != self.dimensions:
+                logger.warning(f"Expected embedding dimension {self.dimensions}, but got {len(embedding)}")
+                return []
+            return embedding
+        except Exception as e:
+            logger.warning(f"Error getting embedding: {e}")
+            return []
+
+    async def async_get_embedding_and_usage(self, text: str) -> Tuple[List[float], Optional[Dict]]:
+        """Async version of get_embedding_and_usage."""
+        embedding = await self.async_get_embedding(text=text)
+        usage = None
+        return embedding, usage