agno 1.7.8__py3-none-any.whl → 1.7.9__py3-none-any.whl

agno/agent/agent.py

@@ -893,7 +893,7 @@ class Agent:
             ):
                 yield event
         else:
-            from agno.utils.events import RunResponseContentEvent
+            from agno.run.response import IntermediateRunResponseContentEvent, RunResponseContentEvent
 
             for event in self._handle_model_response_stream(
                 run_response=run_response,
@@ -903,7 +903,10 @@ class Agent:
             ):
                 if isinstance(event, RunResponseContentEvent):
                     if stream_intermediate_steps:
-                        yield event
+                        yield IntermediateRunResponseContentEvent(
+                            content=event.content,
+                            content_type=event.content_type,
+                        )
                 else:
                     yield event
 
@@ -1331,7 +1334,7 @@ class Agent:
             ):
                 yield event
         else:
-            from agno.utils.events import RunResponseContentEvent
+            from agno.run.response import IntermediateRunResponseContentEvent, RunResponseContentEvent
 
             async for event in self._ahandle_model_response_stream(
                 run_response=run_response,
@@ -1341,7 +1344,10 @@ class Agent:
             ):
                 if isinstance(event, RunResponseContentEvent):
                     if stream_intermediate_steps:
-                        yield event
+                        yield IntermediateRunResponseContentEvent(
+                            content=event.content,
+                            content_type=event.content_type,
+                        )
                 else:
                     yield event
 
@@ -4973,9 +4979,26 @@ class Agent:
 
                 run_messages.messages += history_copy
 
-        # 4.Add user message to run_messages
+        # 4. Add messages to run_messages if provided
+        if messages is not None and len(messages) > 0:
+            for _m in messages:
+                if isinstance(_m, Message):
+                    run_messages.messages.append(_m)
+                    if run_messages.extra_messages is None:
+                        run_messages.extra_messages = []
+                    run_messages.extra_messages.append(_m)
+                elif isinstance(_m, dict):
+                    try:
+                        run_messages.messages.append(Message.model_validate(_m))
+                        if run_messages.extra_messages is None:
+                            run_messages.extra_messages = []
+                        run_messages.extra_messages.append(Message.model_validate(_m))
+                    except Exception as e:
+                        log_warning(f"Failed to validate message: {e}")
+
+        # 5. Add user message to run_messages
         user_message: Optional[Message] = None
-        # 4.1 Build user message if message is None, str or list
+        # 5.1 Build user message if message is None, str or list
         if message is None or isinstance(message, str) or isinstance(message, list):
             user_message = self.get_user_message(
                 message=message,
@@ -4986,16 +5009,16 @@ class Agent:
                 knowledge_filters=knowledge_filters,
                 **kwargs,
             )
-        # 4.2 If message is provided as a Message, use it directly
+        # 5.2 If message is provided as a Message, use it directly
         elif isinstance(message, Message):
             user_message = message
-        # 4.3 If message is provided as a dict, try to validate it as a Message
+        # 5.3 If message is provided as a dict, try to validate it as a Message
         elif isinstance(message, dict):
             try:
                 user_message = Message.model_validate(message)
             except Exception as e:
                 log_warning(f"Failed to validate message: {e}")
-        # 4.4 If message is provided as a BaseModel, convert it to a Message
+        # 5.4 If message is provided as a BaseModel, convert it to a Message
         elif isinstance(message, BaseModel):
             try:
                 # Create a user message with the BaseModel content
@@ -5008,23 +5031,6 @@ class Agent:
             run_messages.user_message = user_message
             run_messages.messages.append(user_message)
 
-        # 5. Add messages to run_messages if provided
-        if messages is not None and len(messages) > 0:
-            for _m in messages:
-                if isinstance(_m, Message):
-                    run_messages.messages.append(_m)
-                    if run_messages.extra_messages is None:
-                        run_messages.extra_messages = []
-                    run_messages.extra_messages.append(_m)
-                elif isinstance(_m, dict):
-                    try:
-                        run_messages.messages.append(Message.model_validate(_m))
-                        if run_messages.extra_messages is None:
-                            run_messages.extra_messages = []
-                        run_messages.extra_messages.append(Message.model_validate(_m))
-                    except Exception as e:
-                        log_warning(f"Failed to validate message: {e}")
-
         return run_messages
 
     def get_continue_run_messages(

agno/document/reader/pdf_reader.py

@@ -1,6 +1,7 @@
 import asyncio
+import re
 from pathlib import Path
-from typing import IO, Any, List, Optional, Union
+from typing import IO, Any, List, Optional, Tuple, Union
 from uuid import uuid4
 
 from agno.document.base import Document
@@ -15,7 +16,13 @@ except ImportError:
     raise ImportError("`pypdf` not installed. Please install it via `pip install pypdf`.")
 
 
-def process_image_page(doc_name: str, page_number: int, page: Any) -> Document:
+PAGE_START_NUMBERING_FORMAT_DEFAULT = "<start page {page_nr}>"
+PAGE_END_NUMBERING_FORMAT_DEFAULT = "<end page {page_nr}>"
+PAGE_NUMBERING_CORRECTNESS_RATIO_FOR_REMOVAL = 0.4
+
+
+def _ocr_reader(page: Any) -> str:
+    """A single PDF page object."""
     try:
         import rapidocr_onnxruntime as rapidocr
     except ImportError:
@@ -23,7 +30,6 @@ def process_image_page(doc_name: str, page_number: int, page: Any) -> Document:
             "`rapidocr_onnxruntime` not installed. Please install it via `pip install rapidocr_onnxruntime`."
         )
     ocr = rapidocr.RapidOCR()
-    page_text = page.extract_text() or ""
     images_text_list = []
 
     # Extract and process images
@@ -34,22 +40,13 @@ def process_image_page(doc_name: str, page_number: int, page: Any) -> Document:
         ocr_result, elapse = ocr(image_data)
 
         # Extract text from OCR result
-        if ocr_result:
-            images_text_list += [item[1] for item in ocr_result]
+        images_text_list += [item[1] for item in ocr_result] if ocr_result else []
 
-    images_text = "\n".join(images_text_list)
-    content = page_text + "\n" + images_text
-
-    # Append the document
-    return Document(
-        name=doc_name,
-        id=str(uuid4()),
-        meta_data={"page": page_number},
-        content=content,
-    )
+    return "\n".join(images_text_list)
 
 
-async def async_process_image_page(doc_name: str, page_number: int, page: Any) -> Document:
+async def _async_ocr_reader(page: Any) -> str:
+    """page: A single PDF page object."""
     try:
         import rapidocr_onnxruntime as rapidocr
     except ImportError:
@@ -58,9 +55,6 @@ async def async_process_image_page(doc_name: str, page_number: int, page: Any) -
         )
     ocr = rapidocr.RapidOCR()
 
-    page_text = page.extract_text() or ""
-    images_text_list: List = []
-
     # Process images in parallel
     async def process_image(image_data: bytes) -> List[str]:
         ocr_result, _ = ocr(image_data)
@@ -69,27 +63,221 @@ async def async_process_image_page(doc_name: str, page_number: int, page: Any) -
     image_tasks = [process_image(image.data) for image in page.images]
     images_results = await asyncio.gather(*image_tasks)
 
+    images_text_list: List = []
     for result in images_results:
         images_text_list.extend(result)
 
     images_text = "\n".join(images_text_list)
-    content = page_text + "\n" + images_text
-
-    return Document(
-        name=doc_name,
-        id=str(uuid4()),
-        meta_data={"page": page_number},
-        content=content,
+    return images_text
+
+
+def _clean_page_numbers(
+    page_content_list: List[str],
+    extra_content: List[str] = [],
+    page_start_numbering_format: str = PAGE_START_NUMBERING_FORMAT_DEFAULT,
+    page_end_numbering_format: str = PAGE_END_NUMBERING_FORMAT_DEFAULT,
+) -> Tuple[List[str], Optional[int]]:
+    f"""
+    Identifies and removes or reformats page numbers from a list of PDF page contents, based on the most consistent sequential numbering.
+
+    Args:
+        page_content_list (List[str]): A list of strings where each string represents the content of a PDF page.
+        extra_content (List[str]): A list of strings where each string will be appended after the main content. Can be used for appending image information.
+        page_start_numbering_format (str): A format string to prepend to the page content, with `{{page_nr}}` as a placeholder for the page number.
+            Defaults to {PAGE_START_NUMBERING_FORMAT_DEFAULT}. Make it an empty string to remove the page number.
+        page_end_numbering_format (str): A format string to append to the page content, with `{{page_nr}}` as a placeholder for the page number.
+            Defaults to {PAGE_END_NUMBERING_FORMAT_DEFAULT}. Make it an empty string to remove the page number.
+
+    Returns:
+        List[str]: The list of page contents with page numbers removed or reformatted based on the detected sequence.
+        Optional[Int]: The shift for the page numbering. Can be (-2, -1, 0, 1, 2).
+
+    Notes:
+        - The function scans for page numbers using a regular expression that matches digits at the start or end of a string.
+        - It evaluates several potential starting points for numbering (-2, -1, 0, 1, 2 shifts) to determine the most consistent sequence.
+        - If at least a specified ratio of pages (defined by `PAGE_NUMBERING_CORRECTNESS_RATIO_FOR_REMOVAL`) has correct sequential numbering, 
+          the page numbers are processed.
+        - If page numbers are found, the function will add formatted page numbers to each page's content if `page_start_numbering_format` or 
+          `page_end_numbering_format` is provided.
+    """
+    assert len(extra_content) == 0 or len(extra_content) == len(page_content_list), (
+        "Please provide an equally sized list of extra content if provided."
     )
 
+    # Regex to match potential page numbers at the start or end of a string
+    page_number_regex = re.compile(r"^\s*(\d+)\s*|\s*(\d+)\s*$")
+
+    def find_page_number(content):
+        match = page_number_regex.search(content)
+        if match:
+            return int(match.group(1) or match.group(2))
+        return None
+
+    page_numbers = [find_page_number(content) for content in page_content_list]
+    if all(x is None or x > 5 for x in page_numbers):
+        # This approach won't work reliably for higher page numbers.
+        return page_content_list, None
+
+    # Possible range shifts to detect page numbering
+    range_shifts = [-2, -1, 0, 1, 2]
+    best_match, best_correct_count, best_shift = _identify_best_page_sequence(page_numbers, range_shifts)
+
+    # Check if at least ..% of the pages have correct sequential numbering
+    if best_match and best_correct_count / len(page_numbers) >= PAGE_NUMBERING_CORRECTNESS_RATIO_FOR_REMOVAL:
+        # Remove the page numbers from the content
+        for i, expected_number in enumerate(best_match):
+            page_content_list[i] = re.sub(
+                rf"^\s*{expected_number}\s*|\s*{expected_number}\s*$", "", page_content_list[i]
+            )
+
+            page_start = (
+                page_start_numbering_format.format(page_nr=expected_number) + "\n"
+                if page_start_numbering_format
+                else ""
+            )
+            page_end = (
+                "\n" + page_end_numbering_format.format(page_nr=expected_number) if page_end_numbering_format else ""
+            )
+            extra_info = "\n" + extra_content[i] if extra_content else ""
+
+            # Add formatted page numbering if configured.
+            page_content_list[i] = page_start + page_content_list[i] + extra_info + page_end
+    else:
+        best_shift = None
+
+    return page_content_list, best_shift
+
+
+def _identify_best_page_sequence(page_numbers, range_shifts):
+    best_match = None
+    best_shift: Optional[int] = None
+    best_correct_count = 0
+
+    for shift in range_shifts:
+        expected_numbers = [i + shift for i in range(len(page_numbers))]
+        # Check if expected number occurs (or that the expected "2" occurs in an incorrectly merged number like 25,
+        # where 2 is the page number and 5 is part of the PDF content).
+        correct_count = sum(
+            1
+            for actual, expected in zip(page_numbers, expected_numbers)
+            if actual == expected or str(actual).startswith(str(expected)) or str(actual).endswith(str(expected))
+        )
+
+        if correct_count > best_correct_count:
+            best_correct_count = correct_count
+            best_match = expected_numbers
+            best_shift = shift
+
+    return best_match, best_correct_count, best_shift
+
 
 class BasePDFReader(Reader):
+    def __init__(
+        self,
+        split_on_pages: bool = True,
+        page_start_numbering_format: Optional[str] = None,
+        page_end_numbering_format: Optional[str] = None,
+        **kwargs,
+    ):
+        if page_start_numbering_format is None:
+            page_start_numbering_format = PAGE_START_NUMBERING_FORMAT_DEFAULT
+        if page_end_numbering_format is None:
+            page_end_numbering_format = PAGE_END_NUMBERING_FORMAT_DEFAULT
+
+        self.split_on_pages = split_on_pages
+        self.page_start_numbering_format = page_start_numbering_format
+        self.page_end_numbering_format = page_end_numbering_format
+
+        super().__init__(**kwargs)
+
     def _build_chunked_documents(self, documents: List[Document]) -> List[Document]:
         chunked_documents: List[Document] = []
         for document in documents:
             chunked_documents.extend(self.chunk_document(document))
         return chunked_documents
 
+    def _create_documents(self, pdf_content: List[str], doc_name: str, use_uuid_for_id: bool, page_number_shift):
+        if self.split_on_pages:
+            shift = page_number_shift if page_number_shift is not None else 1
+            documents: List[Document] = []
+            for page_number, page_content in enumerate(pdf_content, start=shift):
+                documents.append(
+                    Document(
+                        name=doc_name,
+                        id=(str(uuid4()) if use_uuid_for_id else f"{doc_name}_{page_number}"),
+                        meta_data={"page": page_number},
+                        content=page_content,
+                    )
+                )
+        else:
+            pdf_content_str = "\n".join(pdf_content)
+            document = Document(
+                name=doc_name,
+                id=str(uuid4()) if use_uuid_for_id else doc_name,
+                meta_data={},
+                content=pdf_content_str,
+            )
+            documents = [document]
+
+        if self.chunk:
+            return self._build_chunked_documents(documents)
+
+        return documents
+
+    def _pdf_reader_to_documents(
+        self,
+        doc_reader: DocumentReader,
+        doc_name,
+        read_images=False,
+        use_uuid_for_id=False,
+    ):
+        pdf_content = []
+        pdf_images_text = []
+        for page in doc_reader.pages:
+            pdf_content.append(page.extract_text())
+            if read_images:
+                pdf_images_text.append(_ocr_reader(page))
+
+        pdf_content, shift = _clean_page_numbers(
+            page_content_list=pdf_content,
+            extra_content=pdf_images_text,
+            page_start_numbering_format=self.page_start_numbering_format,
+            page_end_numbering_format=self.page_end_numbering_format,
+        )
+        return self._create_documents(pdf_content, doc_name, use_uuid_for_id, shift)
+
+    async def _async_pdf_reader_to_documents(
+        self,
+        doc_reader: DocumentReader,
+        doc_name: str,
+        read_images=False,
+        use_uuid_for_id=False,
+    ):
+        async def _read_pdf_page(page, read_images) -> Tuple[str, str]:
+            # We tried "asyncio.to_thread(page.extract_text)", but it maintains state internally, which leads to issues.
+            page_text = page.extract_text()
+
+            if read_images:
+                pdf_images_text = await _async_ocr_reader(page)
+            else:
+                pdf_images_text = ""
+
+            return page_text, pdf_images_text
+
+        # Process pages in parallel using asyncio.gather
+        pdf_content: List[Tuple[str, str]] = await asyncio.gather(
+            *[_read_pdf_page(page, read_images) for page in doc_reader.pages]
+        )
+
+        pdf_content_clean, shift = _clean_page_numbers(
+            page_content_list=[x[0] for x in pdf_content],
+            extra_content=[x[1] for x in pdf_content],
+            page_start_numbering_format=self.page_start_numbering_format,
+            page_end_numbering_format=self.page_end_numbering_format,
+        )
+
+        return self._create_documents(pdf_content_clean, doc_name, use_uuid_for_id, shift)
+
 
 class PDFReader(BasePDFReader):
     """Reader for PDF files"""
@@ -106,24 +294,13 @@ class PDFReader(BasePDFReader):
         log_info(f"Reading: {doc_name}")
 
         try:
-            doc_reader = DocumentReader(pdf)
+            pdf_reader = DocumentReader(pdf)
         except PdfStreamError as e:
             logger.error(f"Error reading PDF: {e}")
             return []
 
-        documents = []
-        for page_number, page in enumerate(doc_reader.pages, start=1):
-            documents.append(
-                Document(
-                    name=doc_name,
-                    id=str(uuid4()),
-                    meta_data={"page": page_number},
-                    content=page.extract_text(),
-                )
-            )
-        if self.chunk:
-            return self._build_chunked_documents(documents)
-        return documents
+        # Read and chunk.
+        return self._pdf_reader_to_documents(pdf_reader, doc_name, use_uuid_for_id=True)
 
     async def async_read(self, pdf: Union[str, Path, IO[Any]]) -> List[Document]:
         try:
@@ -137,30 +314,13 @@ class PDFReader(BasePDFReader):
         log_info(f"Reading: {doc_name}")
 
         try:
-            doc_reader = DocumentReader(pdf)
+            pdf_reader = DocumentReader(pdf)
         except PdfStreamError as e:
             logger.error(f"Error reading PDF: {e}")
             return []
 
-        async def _process_document(doc_name: str, page_number: int, page: Any) -> Document:
-            return Document(
-                name=doc_name,
-                id=str(uuid4()),
-                meta_data={"page": page_number},
-                content=page.extract_text(),
-            )
-
-        # Process pages in parallel using asyncio.gather
-        documents = await asyncio.gather(
-            *[
-                _process_document(doc_name, page_number, page)
-                for page_number, page in enumerate(doc_reader.pages, start=1)
-            ]
-        )
-
-        if self.chunk:
-            return self._build_chunked_documents(documents)
-        return documents
+        # Read and chunk.
+        return await self._async_pdf_reader_to_documents(pdf_reader, doc_name, use_uuid_for_id=True)
 
 
 class PDFUrlReader(BasePDFReader):
@@ -182,21 +342,10 @@ class PDFUrlReader(BasePDFReader):
         response = fetch_with_retry(url, proxy=self.proxy)
 
         doc_name = url.split("/")[-1].split(".")[0].replace("/", "_").replace(" ", "_")
-        doc_reader = DocumentReader(BytesIO(response.content))
-
-        documents = []
-        for page_number, page in enumerate(doc_reader.pages, start=1):
-            documents.append(
-                Document(
-                    name=doc_name,
-                    id=f"{doc_name}_{page_number}",
-                    meta_data={"page": page_number},
-                    content=page.extract_text(),
-                )
-            )
-        if self.chunk:
-            return self._build_chunked_documents(documents)
-        return documents
+        pdf_reader = DocumentReader(BytesIO(response.content))
+
+        # Read and chunk.
+        return self._pdf_reader_to_documents(pdf_reader, doc_name, use_uuid_for_id=False)
 
     async def async_read(self, url: str) -> List[Document]:
         if not url:
@@ -213,27 +362,10 @@ class PDFUrlReader(BasePDFReader):
             response = await async_fetch_with_retry(url, client=client)
 
         doc_name = url.split("/")[-1].split(".")[0].replace("/", "_").replace(" ", "_")
-        doc_reader = DocumentReader(BytesIO(response.content))
-
-        async def _process_document(doc_name: str, page_number: int, page: Any) -> Document:
-            return Document(
-                name=doc_name,
-                id=f"{doc_name}_{page_number}",
-                meta_data={"page": page_number},
-                content=page.extract_text(),
-            )
-
-        # Process pages in parallel using asyncio.gather
-        documents = await asyncio.gather(
-            *[
-                _process_document(doc_name, page_number, page)
-                for page_number, page in enumerate(doc_reader.pages, start=1)
-            ]
-        )
+        pdf_reader = DocumentReader(BytesIO(response.content))
 
-        if self.chunk:
-            return self._build_chunked_documents(documents)
-        return documents
+        # Read and chunk.
+        return await self._async_pdf_reader_to_documents(pdf_reader, doc_name, use_uuid_for_id=False)
 
 
 class PDFImageReader(BasePDFReader):
@@ -252,16 +384,10 @@ class PDFImageReader(BasePDFReader):
             doc_name = "pdf"
 
         log_info(f"Reading: {doc_name}")
-        doc_reader = DocumentReader(pdf)
-
-        documents = []
-        for page_number, page in enumerate(doc_reader.pages, start=1):
-            documents.append(process_image_page(doc_name, page_number, page))
-
-        if self.chunk:
-            return self._build_chunked_documents(documents)
+        pdf_reader = DocumentReader(pdf)
 
-        return documents
+        # Read and chunk.
+        return self._pdf_reader_to_documents(pdf_reader, doc_name, read_images=True, use_uuid_for_id=False)
 
     async def async_read(self, pdf: Union[str, Path, IO[Any]]) -> List[Document]:
         if not pdf:
@@ -276,18 +402,10 @@ class PDFImageReader(BasePDFReader):
             doc_name = "pdf"
 
         log_info(f"Reading: {doc_name}")
-        doc_reader = DocumentReader(pdf)
+        pdf_reader = DocumentReader(pdf)
 
-        documents = await asyncio.gather(
-            *[
-                async_process_image_page(doc_name, page_number, page)
-                for page_number, page in enumerate(doc_reader.pages, start=1)
-            ]
-        )
-
-        if self.chunk:
-            return self._build_chunked_documents(documents)
-        return documents
+        # Read and chunk.
+        return await self._async_pdf_reader_to_documents(pdf_reader, doc_name, read_images=True, use_uuid_for_id=False)
 
 
 class PDFUrlImageReader(BasePDFReader):
@@ -310,17 +428,10 @@ class PDFUrlImageReader(BasePDFReader):
         response = httpx.get(url, proxy=self.proxy) if self.proxy else httpx.get(url)
 
         doc_name = url.split("/")[-1].split(".")[0].replace(" ", "_")
-        doc_reader = DocumentReader(BytesIO(response.content))
-
-        documents = []
-        for page_number, page in enumerate(doc_reader.pages, start=1):
-            documents.append(process_image_page(doc_name, page_number, page))
-
-        # Optionally chunk documents
-        if self.chunk:
-            return self._build_chunked_documents(documents)
+        pdf_reader = DocumentReader(BytesIO(response.content))
 
-        return documents
+        # Read and chunk.
+        return self._pdf_reader_to_documents(pdf_reader, doc_name, read_images=True, use_uuid_for_id=False)
 
     async def async_read(self, url: str) -> List[Document]:
         if not url:
@@ -338,15 +449,7 @@ class PDFUrlImageReader(BasePDFReader):
             response.raise_for_status()
 
         doc_name = url.split("/")[-1].split(".")[0].replace(" ", "_")
-        doc_reader = DocumentReader(BytesIO(response.content))
+        pdf_reader = DocumentReader(BytesIO(response.content))
 
-        documents = await asyncio.gather(
-            *[
-                async_process_image_page(doc_name, page_number, page)
-                for page_number, page in enumerate(doc_reader.pages, start=1)
-            ]
-        )
-
-        if self.chunk:
-            return self._build_chunked_documents(documents)
-        return documents
+        # Read and chunk.
+        return await self._async_pdf_reader_to_documents(pdf_reader, doc_name, read_images=True, use_uuid_for_id=False)

agno/run/response.py

@@ -17,6 +17,7 @@ class RunEvent(str, Enum):
 
     run_started = "RunStarted"
     run_response_content = "RunResponseContent"
+    run_intermediate_response_content = "RunIntermediateResponseContent"
     run_completed = "RunCompleted"
     run_error = "RunError"
     run_cancelled = "RunCancelled"
@@ -92,6 +93,13 @@ class RunResponseContentEvent(BaseAgentRunResponseEvent):
     extra_data: Optional[RunResponseExtraData] = None
 
 
+@dataclass
+class IntermediateRunResponseContentEvent(BaseAgentRunResponseEvent):
+    event: str = RunEvent.run_intermediate_response_content.value
+    content: Optional[Any] = None
+    content_type: str = "str"
+
+
 @dataclass
 class RunResponseCompletedEvent(BaseAgentRunResponseEvent):
     event: str = RunEvent.run_completed.value
@@ -207,6 +215,7 @@ class OutputModelResponseCompletedEvent(BaseAgentRunResponseEvent):
 RunResponseEvent = Union[
     RunResponseStartedEvent,
     RunResponseContentEvent,
+    IntermediateRunResponseContentEvent,
     RunResponseCompletedEvent,
     RunResponseErrorEvent,
     RunResponseCancelledEvent,
@@ -230,6 +239,7 @@ RunResponseEvent = Union[
 RUN_EVENT_TYPE_REGISTRY = {
     RunEvent.run_started.value: RunResponseStartedEvent,
     RunEvent.run_response_content.value: RunResponseContentEvent,
+    RunEvent.run_intermediate_response_content.value: IntermediateRunResponseContentEvent,
     RunEvent.run_completed.value: RunResponseCompletedEvent,
     RunEvent.run_error.value: RunResponseErrorEvent,
     RunEvent.run_cancelled.value: RunResponseCancelledEvent,

agno/run/team.py

@@ -17,6 +17,7 @@ class TeamRunEvent(str, Enum):
 
     run_started = "TeamRunStarted"
     run_response_content = "TeamRunResponseContent"
+    run_intermediate_response_content = "TeamRunIntermediateResponseContent"
     run_completed = "TeamRunCompleted"
     run_error = "TeamRunError"
     run_cancelled = "TeamRunCancelled"
@@ -94,6 +95,13 @@ class RunResponseContentEvent(BaseTeamRunResponseEvent):
     extra_data: Optional[RunResponseExtraData] = None
 
 
+@dataclass
+class IntermediateRunResponseContentEvent(BaseTeamRunResponseEvent):
+    event: str = TeamRunEvent.run_intermediate_response_content.value
+    content: Optional[Any] = None
+    content_type: str = "str"
+
+
 @dataclass
 class RunResponseCompletedEvent(BaseTeamRunResponseEvent):
     event: str = TeamRunEvent.run_completed.value
@@ -191,6 +199,7 @@ class OutputModelResponseCompletedEvent(BaseTeamRunResponseEvent):
 TeamRunResponseEvent = Union[
     RunResponseStartedEvent,
     RunResponseContentEvent,
+    IntermediateRunResponseContentEvent,
     RunResponseCompletedEvent,
     RunResponseErrorEvent,
     RunResponseCancelledEvent,
@@ -211,6 +220,7 @@ TeamRunResponseEvent = Union[
 TEAM_RUN_EVENT_TYPE_REGISTRY = {
     TeamRunEvent.run_started.value: RunResponseStartedEvent,
     TeamRunEvent.run_response_content.value: RunResponseContentEvent,
+    TeamRunEvent.run_intermediate_response_content.value: IntermediateRunResponseContentEvent,
     TeamRunEvent.run_completed.value: RunResponseCompletedEvent,
     TeamRunEvent.run_error.value: RunResponseErrorEvent,
     TeamRunEvent.run_cancelled.value: RunResponseCancelledEvent,

agno/team/team.py

@@ -1110,16 +1110,21 @@ class Team:
                 response_format=response_format,
                 stream_intermediate_steps=stream_intermediate_steps,
             ):
-                from agno.run.team import RunResponseContentEvent
+                from agno.run.team import IntermediateRunResponseContentEvent, RunResponseContentEvent
 
                 if isinstance(event, RunResponseContentEvent):
                     if stream_intermediate_steps:
-                        yield event
+                        yield IntermediateRunResponseContentEvent(
+                            content=event.content,
+                            content_type=event.content_type,
+                        )
                 else:
                     yield event
 
             yield from self._generate_response_with_output_model_stream(
-                run_response=run_response, run_messages=run_messages
+                run_response=run_response,
+                run_messages=run_messages,
+                stream_intermediate_steps=stream_intermediate_steps,
             )
 
         # If a parser model is provided, structure the response separately
@@ -1519,19 +1524,28 @@ class Team:
             ):
                 yield event
         else:
-            from agno.run.team import RunResponseContentEvent
-
-            async for event in self._agenerate_response_with_output_model_stream(
+            async for event in self._ahandle_model_response_stream(
                 run_response=run_response,
                 run_messages=run_messages,
+                response_format=response_format,
                 stream_intermediate_steps=stream_intermediate_steps,
             ):
+                from agno.run.team import IntermediateRunResponseContentEvent, RunResponseContentEvent
+
                 if isinstance(event, RunResponseContentEvent):
                     if stream_intermediate_steps:
-                        yield event
+                        yield IntermediateRunResponseContentEvent(
+                            content=event.content,
+                            content_type=event.content_type,
+                        )
                 else:
                     yield event
 
+            async for event in self._agenerate_response_with_output_model_stream(
+                run_response=run_response,
+                run_messages=run_messages,
+                stream_intermediate_steps=stream_intermediate_steps,
+            ):
                 yield event
 
         # If a parser model is provided, structure the response separately
@@ -2446,10 +2460,9 @@ class Team:
         model_response.content = output_model_response.content
 
     def _generate_response_with_output_model_stream(
-        self, run_response: TeamRunResponse, run_messages: RunMessages, stream_intermediate_steps: bool = True
+        self, run_response: TeamRunResponse, run_messages: RunMessages, stream_intermediate_steps: bool = False
     ):
         """Parse the model response using the output model stream."""
-
         from agno.utils.events import (
             create_team_output_model_response_completed_event,
             create_team_output_model_response_started_event,
@@ -2497,7 +2510,7 @@ class Team:
         model_response.content = output_model_response.content
 
     async def _agenerate_response_with_output_model_stream(
-        self, run_response: TeamRunResponse, run_messages: RunMessages, stream_intermediate_steps: bool = True
+        self, run_response: TeamRunResponse, run_messages: RunMessages, stream_intermediate_steps: bool = False
     ):
         """Parse the model response using the output model stream."""
         from agno.utils.events import (
@@ -2514,9 +2527,7 @@ class Team:
         messages_for_output_model = self.get_messages_for_output_model(run_messages.messages)
         model_response = ModelResponse(content="")
 
-        model_response_stream = self.output_model.aresponse_stream(messages=messages_for_output_model)
-
-        async for model_response_event in model_response_stream:
+        async for model_response_event in self.output_model.aresponse_stream(messages=messages_for_output_model):
             for event in self._handle_model_response_chunk(
                 run_response=run_response,
                 full_model_response=model_response,

agno/tools/aws_lambda.py

@@ -26,6 +26,9 @@ class AWSLambdaTools(Toolkit):
         super().__init__(name="aws-lambda", tools=tools, **kwargs)
 
     def list_functions(self) -> str:
+        """
+        List all AWS Lambda functions in the configured AWS account.
+        """
         try:
             response = self.client.list_functions()
             functions = [func["FunctionName"] for func in response["Functions"]]
@@ -34,6 +37,13 @@ class AWSLambdaTools(Toolkit):
             return f"Error listing functions: {str(e)}"
 
     def invoke_function(self, function_name: str, payload: str = "{}") -> str:
+        """
+        Invoke a specific AWS Lambda function with an optional JSON payload.
+
+        Args:
+            function_name (str): The name of the Lambda function to invoke.
+            payload (str): The JSON payload to send to the function. Defaults to "{}".
+        """
         try:
             response = self.client.invoke(FunctionName=function_name, Payload=payload)
             return f"Function invoked successfully. Status code: {response['StatusCode']}, Payload: {response['Payload'].read().decode('utf-8')}"

agno/vectordb/lancedb/lance_db.py

@@ -1,5 +1,6 @@
 import json
 from hashlib import md5
+from os import getenv
 from typing import Any, Dict, List, Optional
 
 try:
@@ -74,9 +75,12 @@ class LanceDb(VectorDb):
         # Distance metric
         self.distance: Distance = distance
 
+        # Remote LanceDB connection details
+        self.api_key: Optional[str] = api_key
+
         # LanceDB connection details
         self.uri: lancedb.URI = uri
-        self.connection: lancedb.LanceDBConnection = connection or lancedb.connect(uri=self.uri, api_key=api_key)
+        self.connection: lancedb.DBConnection = connection or lancedb.connect(uri=self.uri, api_key=api_key)
         self.table: Optional[lancedb.db.LanceTable] = table
 
         self.async_connection: Optional[lancedb.AsyncConnection] = async_connection
@@ -168,7 +172,11 @@ class LanceDb(VectorDb):
         schema = self._base_schema()
 
         log_info(f"Creating table: {self.table_name}")
-        tbl = self.connection.create_table(self.table_name, schema=schema, mode="overwrite", exist_ok=True)  # type: ignore
+        if self.api_key or getenv("LANCEDB_API_KEY"):
+            log_info("API key found, creating table in remote LanceDB")
+            tbl = self.connection.create_table(name=self.table_name, schema=schema, mode="overwrite")  # type: ignore
+        else:
+            tbl = self.connection.create_table(name=self.table_name, schema=schema, mode="overwrite", exist_ok=True)  # type: ignore
         return tbl  # type: ignore
 
     def doc_exists(self, document: Document) -> bool:

agno/vectordb/pgvector/pgvector.py

@@ -766,6 +766,9 @@ class PgVector(VectorDb):
                     )
                 )
 
+            if self.reranker:
+                search_results = self.reranker.rerank(query=query, documents=search_results)
+
             log_info(f"Found {len(search_results)} documents")
             return search_results
         except Exception as e:

{agno-1.7.8.dist-info → agno-1.7.9.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agno
-Version: 1.7.8
+Version: 1.7.9
 Summary: Agno: a lightweight library for building Multi-Agent Systems
 Author-email: Ashpreet Bedi <ashpreet@agno.com>
 License: Copyright (c) Agno, Inc.

{agno-1.7.8.dist-info → agno-1.7.9.dist-info}/RECORD

@@ -5,7 +5,7 @@ agno/exceptions.py,sha256=HWuuNFS5J0l1RYJsdUrSx51M22aFEoh9ltoeonXBoBw,2891
 agno/media.py,sha256=lXJuylmhuIEWThKZkQ9pUZPp8Kms7EdT4N_U4YN9I00,12656
 agno/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 agno/agent/__init__.py,sha256=Ai6GVyw-0rkA2eYAfoEQIvbi_mrWQUxuPFaFbSDJYCQ,1306
-agno/agent/agent.py,sha256=pvXeME41lUvYBALizxWnawLcIhBwKG33fYJbw-MNY8s,381013
+agno/agent/agent.py,sha256=jNCP1lMtzKmDliOgPhItQTrx6HJG5ZHNo3P4g41Jy9k,381426
 agno/agent/metrics.py,sha256=Lf7JYgPPdqRCyPfCDVUjnmUZ1SkWXrJClL80aW2ffEw,4379
 agno/api/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 agno/api/agent.py,sha256=J-Y4HI-J0Bu6r9gxRYCM3U7SnVBGwLIouDy806KSIIw,2821
@@ -92,7 +92,7 @@ agno/document/reader/docx_reader.py,sha256=fNSZNzBlROQow7nagouEfN8E4KgVp3hTcSj3d
 agno/document/reader/firecrawl_reader.py,sha256=4CFACPA22t0gI1YgYL1mZLZQN_T6LrLj8V9mAmU2In8,5748
 agno/document/reader/json_reader.py,sha256=TrE14YAPkEd3q1e1dFf1ZX-GJPlXadsbeCzNh6EGpgg,2189
 agno/document/reader/markdown_reader.py,sha256=SX0Zydj_0AmBr8Grk8z9jDiNmebU7rnPg2aAJoFEHAk,3546
-agno/document/reader/pdf_reader.py,sha256=UQpKO41JtmA1lZaVe4gyZERsxglcZWK2qlnpYNp1_fI,11253
+agno/document/reader/pdf_reader.py,sha256=KgY_V7dI0XC-hoYIKYg_nGszzsthWF58ejocegkY-1I,17269
 agno/document/reader/text_reader.py,sha256=jtCuhWHkC5QNPmRF3DyXcXVKHM1jnqqxXVxHKZIpkfQ,3315
 agno/document/reader/url_reader.py,sha256=dQmuO1NkJg9A9m2YeiVMuR7GikLl_K1FudE0UO62Um4,1988
 agno/document/reader/website_reader.py,sha256=Ei-FIsdw0FWhGOBVge24JcNMfiAHL8nUA1tawrfBZPM,17307
@@ -291,8 +291,8 @@ agno/reranker/sentence_transformer.py,sha256=Pqv38uZ5MhwYPQkei8ixs3jANoBOIKqVwI3
 agno/run/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 agno/run/base.py,sha256=cueGFKz_TOCNO9aMe8IBFwohax4z6IgL0e3HAK7XqRQ,7546
 agno/run/messages.py,sha256=rAC4CLW-xBA6qFS1BOvcjJ9j_qYf0a7sX1mcdY04zMU,1126
-agno/run/response.py,sha256=4Y1VPfRVIQdhacqW_N41RaM_y0RoDvJ-D32cACv2Tvo,15348
-agno/run/team.py,sha256=3VSdOrlGhRu_zJ9pE623eSqb0MDZiYSZ43GBAl0juQw,16075
+agno/run/response.py,sha256=CafLgeLszVbdre7jm1A5TNG4WMpN0OiPMPk6ioeN-cg,15766
+agno/run/team.py,sha256=WT5HbqVA1fvKPQg8gr2tpdeaNq2aM3WoWpas93P07ck,16504
 agno/run/workflow.py,sha256=2eBvNXbGini3-cWwl3dOf-o-VgGq51ZEl4GCbQFblaw,1581
 agno/run/v2/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 agno/run/v2/workflow.py,sha256=FiZY5eWl6oEJCEfBUIX2q1WDHLI8mqVmaWWwKBA1PL8,18903
@@ -328,14 +328,14 @@ agno/storage/workflow/mongodb.py,sha256=x-0Jl2WovupTfwuVNOSndE9-7V4U7BBIjejtJ1Wa
 agno/storage/workflow/postgres.py,sha256=66bvx6eT7PtFvd4EtTCfI2smynAyvpjvAPYtPo-PCNg,91
 agno/storage/workflow/sqlite.py,sha256=PLqEA1YC8AtIklINr6wy8lzK6KABEqvlJW-nz5KacWM,85
 agno/team/__init__.py,sha256=OSkwJhm4uSoOwpHLeDdcH4q2R_BmfS-7a9_aPxB-Skw,967
-agno/team/team.py,sha256=XmA2r9LAqgYVGINIj0hanfJH-iFtfY4Bn4pRl7TK82s,376693
+agno/team/team.py,sha256=9rZuwihREck6SU6-IZQMOOjZfzxGeQVulsYTMGtb4Rw,377428
 agno/tools/__init__.py,sha256=jNll2sELhPPbqm5nPeT4_uyzRO2_KRTW-8Or60kioS0,210
 agno/tools/agentql.py,sha256=w6FlCfhuS0cc2BHa9K6dZjqO1ycA66fSZbR_nvXiVSo,3813
 agno/tools/airflow.py,sha256=2ZCwx65w_tSXm4xEzZQR_teOiXJlnEgIqU9AgQTQemI,2493
 agno/tools/api.py,sha256=gd86Fvk_3kHcaNWepTQ_Bmzs9RwajwiwcsEHJKdngB8,4238
 agno/tools/apify.py,sha256=WWs9NWiiyMW5jKRRYLDHBoPPX4ScGMMt87VCXHowTy0,13555
 agno/tools/arxiv.py,sha256=xnfbzOBDuHRj5SHImlz_q-pdRT9o_Vnna-VtDt1JuDU,5202
-agno/tools/aws_lambda.py,sha256=GL4McjdbzveteVzeYHKHcwTCLA_B6DJmK4J5tLSNfFk,1459
+agno/tools/aws_lambda.py,sha256=CbVQpoLudpmle1KkgecuVDqE_JdapBBf26OdIj8fUsY,1831
 agno/tools/aws_ses.py,sha256=fl5NTRWcljzxp4WxTg2gIAjLMNcuEWs_vnjeRtEKRHY,2090
 agno/tools/baidusearch.py,sha256=HBdhLz1HUtKXJjIQru21jKiSonG9jEjNB_W6FPjklew,2883
 agno/tools/bitbucket.py,sha256=CLhYdB_HTriy44VHtsRtqCY-BhnHpZMBHQ9vgL4vpWg,11207
@@ -511,14 +511,14 @@ agno/vectordb/clickhouse/index.py,sha256=_YW-8AuEYy5kzOHi0zIzjngpQPgJOBdSrn9BfEL
 agno/vectordb/couchbase/__init__.py,sha256=dKZkcQLFN4r2_NIdXby4inzAAn4BDMlb9T2BW_i0_gQ,93
 agno/vectordb/couchbase/couchbase.py,sha256=I-nWRPAEyEoww-1FpQJLM6CyxWXWPtR9dRjpAyzLWyo,48814
 agno/vectordb/lancedb/__init__.py,sha256=tb9qvinKyWMTLjJYMwW_lhYHFvrfWTfHODtBfMj-NLE,111
-agno/vectordb/lancedb/lance_db.py,sha256=0zKFBJ-RNx-BBxmHfKxzeird6JrYxM2beJd5ySOqmpg,22511
+agno/vectordb/lancedb/lance_db.py,sha256=PBLtn0AkAVEEYJywYKzdf5gWbQWvka6yJXeAfsJQdtI,22886
 agno/vectordb/milvus/__init__.py,sha256=I9V-Rm-rIYxWdRVIs6bKI-6JSJsyOd1-vvasvVpYHuE,127
 agno/vectordb/milvus/milvus.py,sha256=SwLPmgP80dPWFeN9ehj3D1yseyVXbHJDIBNCu-NPs9s,30002
 agno/vectordb/mongodb/__init__.py,sha256=yYwaWdxZRnFTd87Hfgs8_DO4QxcJxy1iL3__bnxP71I,73
 agno/vectordb/mongodb/mongodb.py,sha256=yG_bs7iJrXatrmI2BMLALxkVT1qVx4BYoiUcXWXlSsg,48537
 agno/vectordb/pgvector/__init__.py,sha256=Lui0HBzoHPIsKh5QuiT0eyTvYW88nQPfd_723jjHFCk,288
 agno/vectordb/pgvector/index.py,sha256=qfGgPP33SwZkXLfUcAC_XgQsyZIyggpGS2bfIkjjs-E,495
-agno/vectordb/pgvector/pgvector.py,sha256=IRtsuHknq5OyHK81QRXl_LMJsBmfbPS2cWob6jRkvrU,44181
+agno/vectordb/pgvector/pgvector.py,sha256=JE52MgXSXRqbySPMzgy1DcvJxZhabiw1ufFHf6QImWA,44305
 agno/vectordb/pineconedb/__init__.py,sha256=D7iThXtUCxNO0Nyjunv5Z91Jc1vHG1pgAFXthqD1I_w,92
 agno/vectordb/pineconedb/pineconedb.py,sha256=NaYmkskY1eLAWIM-HSBa-DVHEMQ6ZuZDqqUmmHZhlro,18399
 agno/vectordb/qdrant/__init__.py,sha256=x1ReQt79f9aI_T4JUWb36KNFnvdd-kVwZ1sLsU4sW7Q,76
@@ -550,9 +550,9 @@ agno/workspace/enums.py,sha256=MxF1CUMXBaZMTKLEfiR-7kEhTki2Gfz6W7u49RdYYaE,123
 agno/workspace/helpers.py,sha256=Mp-VlRsPVhW10CfDWYVhc9ANLk9RjNurDfCgXmycZCg,2066
 agno/workspace/operator.py,sha256=CNLwVR45eE5dSRjto2o0c9NgCi2xD-JZR5uLt9kfIt8,30758
 agno/workspace/settings.py,sha256=bcyHHN7lH1LPSMt4i_20XpTjZLoNXdzwyW-G9nHYV40,5703
-agno-1.7.8.dist-info/licenses/LICENSE,sha256=m2rfTWFUfIwCaQqgT2WeBjuKzMKEJRwnaiofg9n8MsQ,16751
-agno-1.7.8.dist-info/METADATA,sha256=Y5HQs_UEHkulA08VeiMi7w6cTig4B-C2DykiMQETwb4,44277
-agno-1.7.8.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-agno-1.7.8.dist-info/entry_points.txt,sha256=Be-iPnPVabMohESsuUdV5w6IAYEIlpc2emJZbyNnfGI,88
-agno-1.7.8.dist-info/top_level.txt,sha256=MKyeuVesTyOKIXUhc-d_tPa2Hrh0oTA4LM0izowpx70,5
-agno-1.7.8.dist-info/RECORD,,
+agno-1.7.9.dist-info/licenses/LICENSE,sha256=m2rfTWFUfIwCaQqgT2WeBjuKzMKEJRwnaiofg9n8MsQ,16751
+agno-1.7.9.dist-info/METADATA,sha256=F9YGHpMsxVonyK77LMJ7vPXygE1klxiUtekDCh58O6g,44277
+agno-1.7.9.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+agno-1.7.9.dist-info/entry_points.txt,sha256=Be-iPnPVabMohESsuUdV5w6IAYEIlpc2emJZbyNnfGI,88
+agno-1.7.9.dist-info/top_level.txt,sha256=MKyeuVesTyOKIXUhc-d_tPa2Hrh0oTA4LM0izowpx70,5
+agno-1.7.9.dist-info/RECORD,,