ai-parrot 0.3.11__cp311-cp311-manylinux_2_28_x86_64.whl → 0.3.16__cp311-cp311-manylinux_2_28_x86_64.whl

{ai_parrot-0.3.11.dist-info → ai_parrot-0.3.16.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: ai-parrot
-Version: 0.3.11
+Version: 0.3.16
 Summary: Live Chatbots based on Langchain chatbots and Agents     Integrated into Navigator Framework or used into aiohttp applications.
 Home-page: https://github.com/phenobarbital/ai-parrot
 Author: Jesus Lara
@@ -88,6 +88,22 @@ Requires-Dist: streamlit==1.37.1; extra == "analytics"
 Provides-Extra: anthropic
 Requires-Dist: langchain-anthropic==0.1.11; extra == "anthropic"
 Requires-Dist: anthropic==0.25.2; extra == "anthropic"
+Provides-Extra: basic_loaders
+Requires-Dist: youtube-transcript-api==0.6.2; extra == "basic-loaders"
+Requires-Dist: pymupdf==1.24.4; extra == "basic-loaders"
+Requires-Dist: pymupdf4llm==0.0.1; extra == "basic-loaders"
+Requires-Dist: pdf4llm==0.0.6; extra == "basic-loaders"
+Requires-Dist: pytube==15.0.0; extra == "basic-loaders"
+Requires-Dist: pydub==0.25.1; extra == "basic-loaders"
+Requires-Dist: markdownify==0.12.1; extra == "basic-loaders"
+Requires-Dist: yt-dlp==2024.4.9; extra == "basic-loaders"
+Requires-Dist: moviepy==1.0.3; extra == "basic-loaders"
+Requires-Dist: rapidocr-onnxruntime==1.3.15; extra == "basic-loaders"
+Requires-Dist: pytesseract==0.3.10; extra == "basic-loaders"
+Requires-Dist: python-docx==1.1.0; extra == "basic-loaders"
+Requires-Dist: python-pptx==0.6.23; extra == "basic-loaders"
+Requires-Dist: docx2txt==0.8; extra == "basic-loaders"
+Requires-Dist: mammoth==1.7.1; extra == "basic-loaders"
 Provides-Extra: crew
 Requires-Dist: colbert-ai==0.2.19; extra == "crew"
 Requires-Dist: vanna==0.3.4; extra == "crew"
@@ -104,26 +120,11 @@ Requires-Dist: llama-index-llms-huggingface==0.2.7; extra == "hunggingfaces"
 Provides-Extra: loaders
 Requires-Dist: unstructured==0.14.3; extra == "loaders"
 Requires-Dist: unstructured-client==0.18.0; extra == "loaders"
-Requires-Dist: youtube-transcript-api==0.6.2; extra == "loaders"
-Requires-Dist: pymupdf==1.24.4; extra == "loaders"
-Requires-Dist: pymupdf4llm==0.0.1; extra == "loaders"
-Requires-Dist: pdf4llm==0.0.6; extra == "loaders"
 Requires-Dist: PyPDF2==3.0.1; extra == "loaders"
 Requires-Dist: pdfminer.six==20231228; extra == "loaders"
 Requires-Dist: pdfplumber==0.11.0; extra == "loaders"
 Requires-Dist: GitPython==3.1.42; extra == "loaders"
 Requires-Dist: opentelemetry-sdk==1.24.0; extra == "loaders"
-Requires-Dist: rapidocr-onnxruntime==1.3.15; extra == "loaders"
-Requires-Dist: pytesseract==0.3.10; extra == "loaders"
-Requires-Dist: python-docx==1.1.0; extra == "loaders"
-Requires-Dist: python-pptx==0.6.23; extra == "loaders"
-Requires-Dist: docx2txt==0.8; extra == "loaders"
-Requires-Dist: pytube==15.0.0; extra == "loaders"
-Requires-Dist: pydub==0.25.1; extra == "loaders"
-Requires-Dist: markdownify==0.12.1; extra == "loaders"
-Requires-Dist: yt-dlp==2024.4.9; extra == "loaders"
-Requires-Dist: moviepy==1.0.3; extra == "loaders"
-Requires-Dist: mammoth==1.7.1; extra == "loaders"
 Requires-Dist: paddlepaddle==2.6.1; extra == "loaders"
 Requires-Dist: paddlepaddle-gpu==2.6.1; extra == "loaders"
 Requires-Dist: paddleocr==2.8.1; extra == "loaders"

{ai_parrot-0.3.11.dist-info → ai_parrot-0.3.16.dist-info}/RECORD

@@ -1,18 +1,18 @@
 parrot/__init__.py,sha256=eTkAkHeJ5BBDG2fxrXA4M37ODBJoS1DQYpeBAWL2xeI,387
-parrot/conf.py,sha256=-9bVGC7Rf-6wpIg6-ojvU4S_G1wBLUCVDt46KEGHEhM,4257
+parrot/conf.py,sha256=andrPREuR_BHiXA_Q0Utyb5xSb1ct_uKnjMzEOa1ftE,4373
 parrot/exceptions.cpython-311-x86_64-linux-gnu.so,sha256=VNyBh3uLxGQgB0l1bkWjQDqYUN2ZAvRmV12AqQijV9Q,361184
 parrot/manager.py,sha256=NhzXoWxSgtoWHpmYP8cV2Ujq_SlvCbQYQBaohAeL2TM,5935
 parrot/models.py,sha256=RsVQCqhSXBKRPcu-BCga9Y1wyvENFXDCuq3_ObIKvAo,13452
 parrot/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-parrot/version.py,sha256=iXWwg_05dEqR8n13PRWz5QqlmSq6_og3HT3moP37RAg,374
+parrot/version.py,sha256=gt_HAcyWEN5lqHTmPKg0Nn-BgX6aq3F6Ah64q2Dfc7s,374
 parrot/chatbots/__init__.py,sha256=ypskCnME0xUv6psBEGCEyXCrD0J0ULHSllpVmSxqb4A,200
 parrot/chatbots/abstract.py,sha256=CmDn3k4r9uKImOZRN4L9zxLbCdC-1MPUAorDlfZT-kA,26421
 parrot/chatbots/asktroc.py,sha256=gyWzyvpAnmXwXd-3DEKoIJtAxt6NnP5mUZdZbkFky8s,604
-parrot/chatbots/base.py,sha256=cRw7k5FRKOfLdXQJeQvACVE5ZgE1NUWf3IY7OsSsxuo,12912
+parrot/chatbots/base.py,sha256=5QX5-VPAOM-I8o0ktBt3_JEDPRQ_-iK0fFLRMUlvs_s,13396
 parrot/chatbots/basic.py,sha256=DIMTPoGc90BRSlokeOdnjlEXAAfZlIFqxXWaMyAX9uk,232
 parrot/chatbots/bose.py,sha256=z8rm8G_tAwHjDUodXfrAKnhaMzufQyf-GrhxwHeHle4,757
 parrot/chatbots/cody.py,sha256=Z0LNiNtZjEe7bA3hwexclBZK5zEF9m2ODVmrzZjC3Bw,623
-parrot/chatbots/copilot.py,sha256=JTnc-fdszwZ2nLmpNu-tVe6Al8z9PNIYHxv8fd42YQU,2051
+parrot/chatbots/copilot.py,sha256=Q_CwoPm1M0loa7N3DLSLK8eq4m99z1CeU5FI9iqF9XI,2767
 parrot/chatbots/dataframe.py,sha256=CfZiLKIwnaku52nl2PNjciqRlH8m2lM4buO6xI7P408,3914
 parrot/chatbots/hragents.py,sha256=PyNIBJ2OH5CtfVydccgpY50V6GI3cLKuVdOMaa7sQz0,574
 parrot/chatbots/oddie.py,sha256=RMbANmJZP1_vLVGKRNBKmA8otyAiWPkvpA0rJ0U3tZk,796
@@ -45,11 +45,11 @@ parrot/llms/hf.py,sha256=f2HhHCICaSHp0y3KRhqNcYXNO-amYTxDXJ_2_9L5Bk8,1594
 parrot/llms/openai.py,sha256=NgWv6IwJ1DborlYhTyureBBdgHfAPc_lGHQRGt80ca8,1759
 parrot/llms/pipes.py,sha256=Ns_wh-alkKocZKlbQyQLKOSBxqfRC_hCbz34vpOOyP8,3798
 parrot/llms/vertex.py,sha256=a0UsH9sa_GiMkg31E52cWE8pXFZjyMtIanr7eAA7iyE,2615
-parrot/loaders/__init__.py,sha256=LGEaj54DP3FA5-C2IDaA8u-MF4lj-Lbd_Mx5R19qHYY,665
-parrot/loaders/abstract.py,sha256=_tsGDb7TracwkL20J2VYd5hC9MR262c2mmS9VvYB4vM,15870
+parrot/loaders/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+parrot/loaders/abstract.py,sha256=Mx6BtihuwvFkl-Ali_84949BfVXFB0JZjmSKnJ-gXSg,17272
 parrot/loaders/audio.py,sha256=P2tWYKxWLM5TMLMm-5qR35cD_pGQWmf8-UZUTiK4R0o,3698
 parrot/loaders/basepdf.py,sha256=Qh_hzR0JArQEVP31SgWt9utt7qWmbfwVoCzUDyBHcXw,3243
-parrot/loaders/basevideo.py,sha256=WcX-q0Rn_E1dYurbA1eH5NOcUBdOye2iWiFTCY_DVgo,10292
+parrot/loaders/basevideo.py,sha256=xLAMfIhjGR10T3-Pdx8MLq5Bp6woCOuq5Jr6yUZ6LCU,11992
 parrot/loaders/csv.py,sha256=DLcFK3z9boMNH3y9Qca5BWDfYXgXjXsGkzxVN1_2wyo,1103
 parrot/loaders/dir.py,sha256=_CU9kWGCpHnZplUamXLs2yEizA1aCRBASn3F6MggitQ,866
 parrot/loaders/excel.py,sha256=9cTsMfxR_YOpBHz9Ru0LJsxBXDVBh52XM8hHV63QgYo,12445
@@ -67,13 +67,13 @@ parrot/loaders/qa.py,sha256=3K_2yBxUzj-ifDpAbUsIc-v66004fKPzGavUqrhc3Kc,2646
 parrot/loaders/repo.py,sha256=vBqBAnwU6p3_DCvI9DVhi1Bs8iCDYHwFGp0P9zvGRyw,3737
 parrot/loaders/rtd.py,sha256=O0h7LDntP_0IBT8LDQi09u-gYVUO5cuvmGsfZLZ4CoU,1990
 parrot/loaders/txt.py,sha256=-xXVSuvkC2LQ2XZ44Nqwk3V8nE4F6UgXylosMCNgeFo,2804
-parrot/loaders/video.py,sha256=pl5Ho69bp5vrWMqg5tLbsnHUus1LByTDoL6NPk57Ays,2929
-parrot/loaders/videolocal.py,sha256=NwFB6n9sQZxh01L6YKIISpG1tuRsg-ME_qXCDS7Vtkk,5143
+parrot/loaders/video.py,sha256=9zKUFFROSIbWjWFOvxDrW4uOewrMzD7-xADmszOpP4k,2930
+parrot/loaders/videolocal.py,sha256=cRYv3KvKKHltMY4QbnvEMCOLHlEY9ZmWeXTL23fy-gA,9669
 parrot/loaders/vimeo.py,sha256=Cs7FkL2Cr8yV44-Tv5wWkveKzqhOeAIP6kF93SCr_Lk,4118
 parrot/loaders/web.py,sha256=kTi-NtAsbQLKi3wD_2o15Z0HHnYzsEEEGjH0RdvyQqQ,8869
 parrot/loaders/web_base.py,sha256=ZwSFXtJR71cpFGN1WCLUC2W6JjEUV865tRKf8isbJ5M,4382
 parrot/loaders/word.py,sha256=jZdHSL5CtAEn1otBYLNSqKLtO3BNcTObDPgqhzk5-4M,4533
-parrot/loaders/youtube.py,sha256=fVnBBw4IfK6NWP7mO66TgxOzJEcGwE3-3S1WMUApJYg,7751
+parrot/loaders/youtube.py,sha256=DzH9bD5ZrLaTG_6GMjHsy1cHoTBR712yUC8tJiAYbNM,9607
 parrot/loaders/handlers/__init__.py,sha256=ksEDtUOEJELmyCIi0KNv7tR2fCUyADBVkwCcyqN_sVE,70
 parrot/loaders/handlers/data.py,sha256=olZ2p-wyUMGoazah7tgHY7V9buGX1FOeJ-cv2vGEoH8,7386
 parrot/loaders/utils/__init__.py,sha256=SkDyK3MuPGhp0NM6kHvaxQDe97Gcl3n9t5A741OVh1c,28
@@ -87,6 +87,7 @@ parrot/tools/abstract.py,sha256=pVSZw8MDpbVcQ-CHaGwP6CpqXHIs8hH8Oy1AqUuMmrw,1706
 parrot/tools/asknews.py,sha256=hEpPJMyNBVfj2maHbqnumn3VkY45oFvrjkE3Rq8EdGA,1039
 parrot/tools/bing.py,sha256=BtmFD66OIuCaOue5U2_yIqtjWf24IhEgNOX1LAVvHtA,464
 parrot/tools/duck.py,sha256=UAAZzlF-Q0sZh0_IcS96dwSgCuBPdeepkwRrMM5cJPY,1920
+parrot/tools/execute.py,sha256=fTMQAsXuUzVyIWmZxL22LrSj2eQ-Rh-ncyUZ9gY-d-A,1687
 parrot/tools/google.py,sha256=NjijcUWH6Crk5Uty_x3FstjDTGZV8JXfBFDQEtMHhac,6236
 parrot/tools/stack.py,sha256=M-VRWjIDa18bl5p88dSKtxMj4Kn21YB76to0u6yXA30,942
 parrot/tools/weather.py,sha256=4v9Ft5lkVzb9Pg7afNs7BK5T3WEcsZbHPlBrF9oXSo8,2541
@@ -103,8 +104,8 @@ resources/users/handlers.py,sha256=BGzqBvPY_OaIF_nONWX4b_B5OyyBrdGuSihIsdlFwjk,2
 resources/users/models.py,sha256=glk7Emv7QCi6i32xRFDrGc8UwK23_LPg0XUOJoHnwRU,6799
 settings/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 settings/settings.py,sha256=9ueEvyLNurUX-AaIeRPV8GKX1c4YjDLbksUAeqEq6Ck,1854
-ai_parrot-0.3.11.dist-info/LICENSE,sha256=vRKOoa7onTsLNvSzJtGtMaNhWWh8B3YAT733Tlu6M4o,1070
-ai_parrot-0.3.11.dist-info/METADATA,sha256=zOYFernNOneZdQH-EIXCsmT83ATrzxGA4x9WmOhciX8,9838
-ai_parrot-0.3.11.dist-info/WHEEL,sha256=UQ-0qXN3LQUffjrV43_e_ZXj2pgORBqTmXipnkj0E8I,113
-ai_parrot-0.3.11.dist-info/top_level.txt,sha256=qHoO4BhYDfeTkyKnciZSQtn5FSLN3Q-P5xCTkyvbuxg,26
-ai_parrot-0.3.11.dist-info/RECORD,,
+ai_parrot-0.3.16.dist-info/LICENSE,sha256=vRKOoa7onTsLNvSzJtGtMaNhWWh8B3YAT733Tlu6M4o,1070
+ai_parrot-0.3.16.dist-info/METADATA,sha256=wD3EGhVYnjrQtP_j4853HjspuqFjDQvJB_-dvnBTK_A,9958
+ai_parrot-0.3.16.dist-info/WHEEL,sha256=UQ-0qXN3LQUffjrV43_e_ZXj2pgORBqTmXipnkj0E8I,113
+ai_parrot-0.3.16.dist-info/top_level.txt,sha256=qHoO4BhYDfeTkyKnciZSQtn5FSLN3Q-P5xCTkyvbuxg,26
+ai_parrot-0.3.16.dist-info/RECORD,,

parrot/chatbots/base.py

@@ -89,6 +89,9 @@ Whether you need help with a specific question or just want to have a conversati
 - OpenWeatherMap: Get weather information about a location.
 - yahoo_finance_news: Retrieve the latest financial news from Yahoo Finance.
 - python_repl_ast: A Python shell. Use this to execute python commands. Input should be a valid python command. When using this tool, sometimes output is abbreviated - make sure it does not look abbreviated before using it in your answer.
+- executable_python_repl_ast: A Python shell. Use this to execute python commands. Input should be a valid python command. When using this tool, whenever you generate a visual output (like charts with matplotlib), instead of using plt.show(), render the image as a base64-encoded HTML string. Do this by saving the plot to a buffer and encoding it in base64, then return the result as a JSON object formatted as follows: "image": "format": "png", "base64": "base64-encoded-string".
+
+
 - youtube_search: Search for videos on YouTube based on specific keywords.
 
 

parrot/chatbots/copilot.py

@@ -4,17 +4,43 @@ from .base import BaseAgent
 from ..tools import (
     ZipcodeAPIToolkit,
     WikipediaTool,
-    WikidataTool,
+    # WikidataTool,
     GoogleSearchTool,
     GoogleLocationFinder,
     BingSearchTool,
-    AskNewsTool,
+    # AskNewsTool,
     DuckDuckGoSearchTool,
     YouTubeSearchTool,
     OpenWeatherMapTool,
     StackExchangeTool,
 )
+from ..tools.execute import ExecutablePythonREPLTool
 
+# ZipCode API Toolkit
+zpt = ZipcodeAPIToolkit()
+zpt_tools = zpt.get_tools()
+
+wk1 = WikipediaTool()
+# wk12 = WikidataTool()
+
+g1 = GoogleSearchTool()
+g2 = GoogleLocationFinder()
+
+b = BingSearchTool()
+d = DuckDuckGoSearchTool()
+# ask = AskNewsTool()
+
+yt = YouTubeSearchTool()
+stackexchange = StackExchangeTool()
+weather = OpenWeatherMapTool()
+
+tooling = [
+    wk1,
+    g1, g2,
+    b, d, yt,
+    weather,
+    stackexchange
+] + zpt_tools
 
 class CopilotAgent(BaseAgent):
     """CopilotAgent Agent.
@@ -30,17 +56,24 @@ class CopilotAgent(BaseAgent):
         **kwargs
     ):
         super().__init__(name, llm, tools, prompt_template, **kwargs)
+        if not tools:
+            tools = tooling
         self.tools = [
                 PythonAstREPLTool(
                     name='python_repl_ast',
                     globals={},
                     locals={}
+                ),
+                ExecutablePythonREPLTool(
+                    name='executable_python_repl_ast',
+                    globals={},
+                    locals={}
                 )
             ] + list(tools)
         self.prompt = self.get_prompt(
             self.prompt_template
         )
-        # print('PROMPT > ', self.prompt)
+        print('PROMPT > ', self.prompt)
 
     @classmethod
     def default_tools(cls) -> list:

parrot/conf.py

@@ -12,6 +12,8 @@ logging.getLogger(name='h5py').setLevel(logging.INFO)
 logging.getLogger(name='tensorflow').setLevel(logging.INFO)
 logging.getLogger(name='selenium.webdriver').setLevel(logging.WARNING)
 logging.getLogger(name='selenium').setLevel(logging.INFO)
+logging.getLogger(name='matplotlib').setLevel(logging.WARNING)
+logging.getLogger(name='PIL').setLevel(logging.INFO)
 
 
 # Static directory

parrot/loaders/__init__.py

@@ -1,20 +0,0 @@
-from .dir import load_directory
-from .pdf import PDFLoader
-from .web import WebLoader
-from .youtube import YoutubeLoader
-from .vimeo import VimeoLoader
-from .word import MSWordLoader
-from .ppt import PPTXLoader
-from .repo import RepositoryLoader
-from .github import GithubLoader
-from .json import JSONLoader
-from .excel import ExcelLoader
-from .web_base import WebBaseLoader
-from .pdfmark import PDFMarkdownLoader
-from .pdfimages import PDFImageLoader
-from .pdftables import PDFTablesLoader
-from .pdfchapters import PDFChapterLoader
-from .txt import TXTLoader
-from .qa import QAFileLoader
-from .rtd import ReadTheDocsLoader
-from .videolocal import VideoLocalLoader

parrot/loaders/abstract.py

@@ -74,6 +74,7 @@ class AbstractLoader(ABC):
         self,
         tokenizer: Union[str, Callable] = None,
         text_splitter: Union[str, Callable] = None,
+        translation: Optional[str] = None,
         source_type: str = 'file',
         **kwargs
     ):
@@ -114,6 +115,15 @@ class AbstractLoader(ABC):
             )
         # JSON encoder:
         self._encoder = JSONContent()
+        # Traslation
+        self._translation = translation
+        self.translator = None
+        if self._translation:
+            mdl = kwargs.get(
+                'translation_model',
+                f"Helsinki-NLP/opus-mt-en-{self._translation}"
+            )
+            self.translator = self.get_translator(mdl)
 
 
     def __enter__(self):
@@ -159,6 +169,27 @@ class AbstractLoader(ABC):
             use_memory_efficient_attention=True,
         ).to(self._device)
 
+    def get_translator(self, model_name: str = 'Helsinki-NLP/opus-mt-en-es'):
+        if not self._translation:
+            return None
+        trans_model = AutoModelForSeq2SeqLM.from_pretrained(
+            model_name,
+            device_map="auto",
+            torch_dtype=torch.bfloat16,
+            trust_remote_code=True
+        )
+        trans_tokenizer = AutoTokenizer.from_pretrained(model_name)
+        translator = pipeline(
+            "translation",
+            model=trans_model,
+            tokenizer=trans_tokenizer,
+            batch_size=True,
+            max_new_tokens=500,
+            min_new_tokens=300,
+            use_fast=True
+        )
+        return translator
+
     def get_summarization_model(self, model_name: str = 'facebook/bart-large-cnn'):
         if self._no_summarization is True:
             return None
@@ -216,7 +247,7 @@ class AbstractLoader(ABC):
             return ''
         try:
             splitter = TokenTextSplitter(
-                chunk_size=5000,
+                chunk_size=6144,
                 chunk_overlap=100,
             )
             prompt_template = """Write a summary of the following, please also identify the main theme:
@@ -454,3 +485,15 @@ class AbstractLoader(ABC):
         for url in urls:
             documents += cls.load(url, **kwargs)
         return documents
+
+    def saving_file(self, filename: PurePath, data: Any):
+        """Save data to a file.
+
+        Args:
+            filename (PurePath): The path to the file.
+            data (Any): The data to save.
+        """
+        with open(filename, 'wb') as f:
+            f.write(data)
+            f.flush()
+        print(f':: Saved File on {filename}')

parrot/loaders/basevideo.py

@@ -1,8 +1,9 @@
 from collections.abc import Callable
-from typing import Any, Union, List
+from typing import Any, Union, List, Optional
 from abc import abstractmethod
 from pathlib import Path
 from moviepy.editor import VideoFileClip
+from pydub import AudioSegment
 from transformers import (
     pipeline,
     AutoModelForSeq2SeqLM,
@@ -193,25 +194,69 @@ class BaseVideoLoader(AbstractLoader):
             print('ERROR in summarization:', e)
             return ""
 
-    def extract_audio(self, video_path, audio_path):
+    def extract_audio(
+        self,
+        video_path: Path,
+        audio_path: Path,
+        compress_speed: bool = False,
+        output_path: Optional[Path] = None,
+        speed_factor: float = 1.5
+    ):
         """
-        Extracts the audio from a video file and saves it as an audio file.
+        Extracts the audio from a video file and optionally compresses the audio speed.
 
         Args:
             video_path (str): Path to the video file.
             audio_path (str): Path where the extracted audio file will be saved.
+            compress_speed (bool): Whether to compress the audio speed.
+            speed_factor (float): The factor by which to speed up the audio.
         """
+        # Ensure that the paths are valid Path objects
+        video_path = Path(video_path)
+        audio_path = Path(audio_path)
+
+        # Check if the audio file already exists
         if audio_path.exists():
             print(f"Audio already extracted: {audio_path}")
             return
+
+        # Load the video and extract the audio
         video_clip = VideoFileClip(str(video_path))
         audio_clip = video_clip.audio
         if not audio_clip:
+            print("No audio found in video.")
             return
+
+        # Write the extracted audio to the specified path
+        print(f"Extracting audio to: {audio_path}")
         audio_clip.write_audiofile(str(audio_path))
         audio_clip.close()
         video_clip.close()
 
+        # Optionally compress the audio speed
+        if compress_speed:
+            print(f"Compressing audio speed by factor: {speed_factor}")
+
+            # Load the audio file with pydub
+            audio = AudioSegment.from_file(audio_path)
+
+            # Adjust the playback speed by modifying the frame rate
+            sped_up_audio = audio._spawn(audio.raw_data, overrides={
+                "frame_rate": int(audio.frame_rate * speed_factor)
+            })
+
+            # Restore the original frame rate to maintain proper playback speed
+            sped_up_audio = sped_up_audio.set_frame_rate(audio.frame_rate)
+
+            # Overwrite the original file with the sped-up version
+            if not output_path:
+                output_path = audio_path
+            sped_up_audio.export(output_path, format="mp3")
+            print(f"Compressed audio saved to: {audio_path}")
+        else:
+            print(f"Audio extracted: {audio_path}")
+
+
     def get_whisper_transcript(self, audio_path: Path, chunk_length: int = 30):
         # Initialize the Whisper parser
         if self._model_name == 'whisper':

parrot/loaders/video.py

@@ -12,7 +12,7 @@ class VideoLoader(BaseVideoLoader):
     """
     _extension = ['.youtube']
     encoding = 'utf-8'
-    chunk_size = 768
+    chunk_size = 2048
 
     def __init__(
         self,

parrot/loaders/videolocal.py

@@ -1,5 +1,6 @@
 from typing import Any
 from collections.abc import Callable
+import re
 import math
 from pathlib import PurePath
 from langchain.docstore.document import Document
@@ -8,16 +9,35 @@ from .basevideo import BaseVideoLoader
 
 def split_text(text, max_length):
     """Split text into chunks of a maximum length, ensuring not to break words."""
+    # Split the transcript into paragraphs
+    paragraphs = text.split('\n\n')
     chunks = []
-    while len(text) > max_length:
-        # Find the last space before the max_length
-        split_point = text.rfind(' ', 0, max_length)
-        # If no space found, split at max_length
-        if split_point == -1:
-            split_point = max_length
-        chunks.append(text[:split_point])
-        text = text[split_point:].strip()
-    chunks.append(text)
+    current_chunk = ""
+    for paragraph in paragraphs:
+        # If the paragraph is too large, split it into sentences
+        if len(paragraph) > max_length:
+            # Split paragraph into sentences
+            sentences = re.split(r'(?<=[.!?]) +', paragraph)
+            for sentence in sentences:
+                if len(current_chunk) + len(sentence) + 1 > max_length:
+                    # Save the current chunk and start a new one
+                    chunks.append(current_chunk.strip())
+                    current_chunk = sentence
+                else:
+                    # Add sentence to the current chunk
+                    current_chunk += " " + sentence
+        else:
+            # If adding the paragraph exceeds max size, start a new chunk
+            if len(current_chunk) + len(paragraph) + 2 > max_length:
+                chunks.append(current_chunk.strip())
+                current_chunk = paragraph
+            else:
+                # Add paragraph to the current chunk
+                current_chunk += "\n\n" + paragraph
+    # Add any remaining text to chunks
+    if current_chunk.strip():
+        chunks.append(current_chunk.strip())
+
     return chunks
 
 
@@ -37,15 +57,23 @@ class VideoLocalLoader(BaseVideoLoader):
         origin: str = '',
         **kwargs
     ):
-        super().__init__(tokenizer, text_splitter, source_type=source_type, **kwargs)
+        super().__init__(
+            tokenizer,
+            text_splitter,
+            source_type=source_type,
+            **kwargs
+        )
+        self.extract_frames: bool = kwargs.pop('extract_frames', False)
+        self.seconds_per_frame: int = kwargs.pop('seconds_per_frame', 1)
+        self.compress_speed: bool = kwargs.pop('compress_speed', False)
+        self.speed_factor: float = kwargs.pop('speed_factor', 1.5)
         self.path = path
 
     def load_video(self, path: PurePath) -> list:
         metadata = {
-            "url": f"{path.name}",
+            "url": f"{path}",
             "source": f"{path}",
-            "filename": f"{path}",
-            # "index": path.stem,
+            "filename": f"{path.name}",
             "question": '',
             "answer": '',
             'type': 'video_transcript',
@@ -58,10 +86,17 @@ class VideoLocalLoader(BaseVideoLoader):
             }
         }
         documents = []
-        transcript_path = path.with_suffix('.vtt')
+        transcript_path = path.with_suffix('.txt')
+        vtt_path = path.with_suffix('.vtt')
+        summary_path = path.with_suffix('.summary')
         audio_path = path.with_suffix('.mp3')
         # second: extract audio from File
-        self.extract_audio(path, audio_path)
+        self.extract_audio(
+            path,
+            audio_path,
+            compress_speed=self.compress_speed,
+            speed_factor=self.speed_factor
+        )
         # get the Whisper parser
         transcript_whisper = self.get_whisper_transcript(audio_path)
         if transcript_whisper:
@@ -70,35 +105,71 @@ class VideoLocalLoader(BaseVideoLoader):
             transcript = ''
         # Summarize the transcript
         if transcript:
-            # Split transcript into chunks
-            transcript_chunks = split_text(transcript, 32767)
+            # first: extract summary, saving summary as a document:
             summary = self.get_summary_from_text(transcript)
-            # Create Two Documents, one is for transcript, second is VTT:
-            metadata['summary'] = summary
-            for chunk in transcript_chunks:
+            self.saving_file(summary_path, summary.encode('utf-8'))
+            # second: saving transcript to a file:
+            self.saving_file(transcript_path, transcript.encode('utf-8'))
+            # Create Three Documents:
+            # one is for transcript
+            # split document only if size > 65.534
+            if len(transcript) > 65534:
+                # Split transcript into chunks
+                transcript_chunks = split_text(transcript, 32767)
+                for chunk in transcript_chunks:
+                    doc = Document(
+                        page_content=chunk,
+                        metadata=metadata
+                    )
+                    documents.append(doc)
+            else:
                 doc = Document(
-                    page_content=chunk,
+                    page_content=transcript,
                     metadata=metadata
                 )
                 documents.append(doc)
+            # second is Summary
+            if summary:
+                _meta = {
+                    **metadata,
+                    "type": 'video summary'
+                }
+                doc = Document(
+                    page_content=summary,
+                    metadata=_meta
+                )
+            # Third is VTT:
         if transcript_whisper:
             # VTT version:
-            transcript = self.transcript_to_vtt(transcript_whisper, transcript_path)
-            transcript_chunks = split_text(transcript, 65535)
-            for chunk in transcript_chunks:
+            transcript = self.transcript_to_vtt(transcript_whisper, vtt_path)
+            _meta = {
+                **metadata,
+                "type": 'video subte vtt'
+            }
+            if len(transcript) > 65535:
+                transcript_chunks = split_text(transcript, 65535)
+                for chunk in transcript_chunks:
+                    doc = Document(
+                        page_content=chunk,
+                        metadata=_meta
+                    )
+                    documents.append(doc)
+            else:
                 doc = Document(
-                    page_content=chunk,
-                    metadata=metadata
+                    page_content=transcript,
+                    metadata=_meta
                 )
                 documents.append(doc)
             # Saving every dialog chunk as a separate document
             dialogs = self.transcript_to_blocks(transcript_whisper)
             docs = []
             for chunk in dialogs:
+                start_time = chunk['start_time']
                 _meta = {
-                    # "index": f"{path.stem}:{chunk['id']}",
+                    "source": f"{path.name}: min. {start_time}",
+                    "type": "video dialog",
                     "document_meta": {
-                        "start": f"{chunk['start_time']}",
+                        "start": f"{start_time}",
                         "end": f"{chunk['end_time']}",
                         "id": f"{chunk['id']}",
                         "language": self._language,
@@ -128,15 +199,61 @@ class VideoLocalLoader(BaseVideoLoader):
                         documents.extend(self.load_video(item))
         return self.split_documents(documents)
 
+    def extract_video(self, path: PurePath) -> list:
+        metadata = {
+            "url": f"{path}",
+            "source": f"{path}",
+            "filename": f"{path.name}",
+            'type': 'video_transcript',
+            "source_type": self._source_type,
+            "transcript": None,
+            "summary": None,
+            "vtt": None
+        }
+        transcript_path = path.with_suffix('.txt')
+        vtt_path = path.with_suffix('.vtt')
+        summary_path = path.with_suffix('.summary')
+        audio_path = path.with_suffix('.mp3')
+        # second: extract audio from File
+        self.extract_audio(
+            path,
+            audio_path,
+            compress_speed=self.compress_speed,
+            speed_factor=self.speed_factor
+        )
+        # get the Whisper parser
+        transcript_whisper = self.get_whisper_transcript(audio_path)
+        if transcript_whisper:
+            transcript = transcript_whisper['text']
+        else:
+            transcript = ''
+        # Summarize the transcript
+        if transcript:
+            # first: extract summary, saving summary as a document:
+            summary = self.get_summary_from_text(transcript)
+            self.saving_file(summary_path, summary.encode('utf-8'))
+            # second: saving transcript to a file:
+            self.saving_file(transcript_path, transcript.encode('utf-8'))
+            metadata['transcript'] = transcript_path
+            metadata["summary"] = summary
+            metadata['summary_file'] = summary_path
+            metadata["vtt"] = vtt_path
+            # Third is VTT:
+        if transcript_whisper:
+            # VTT version:
+            transcript = self.transcript_to_vtt(transcript_whisper, vtt_path)
+        return metadata
+
     def extract(self) -> list:
+        # Adding also Translation to other language.
         documents = []
         if self.path.is_file():
-            docs = self.load_video(self.path)
-            documents.extend(docs)
-        if self.path.is_dir():
+            doc = self.extract_video(self.path)
+            documents.append(doc)
+        elif self.path.is_dir():
             # iterate over the files in the directory
             for ext in self._extension:
                 for item in self.path.glob(f'*{ext}'):
                     if set(item.parts).isdisjoint(self.skip_directories):
-                        documents.extend(self.load_video(item))
+                        documents.append(self.extract_video(item))
         return documents

parrot/loaders/youtube.py

@@ -1,7 +1,6 @@
 from typing import Optional, Union
 from pytube import YouTube
 from youtube_transcript_api import NoTranscriptFound
-import torch
 from langchain.docstore.document import Document
 from langchain_community.document_loaders.parsers.audio import (
     OpenAIWhisperParserLocal
@@ -190,3 +189,53 @@ class YoutubeLoader(VideoLoader):
                     metadata=metadata
                 )
                 return [doc]
+
+    def extract_video(
+        self,
+        url: str
+    ) -> list:
+        # first: load video metadata:
+        video_info = self.get_video_info(url)
+        # first: download video
+        file_path = self.download_video(url, self._video_path)
+        audio_path = file_path.with_suffix('.mp3')
+        transcript_path = file_path.with_suffix('.txt')
+        vtt_path = file_path.with_suffix('.vtt')
+        summary_path = file_path.with_suffix('.summary')
+        # second: extract audio
+        self.extract_audio(file_path, audio_path)
+        transcript_whisper = self.get_whisper_transcript(audio_path)
+        transcript = transcript_whisper['text']
+        # Summarize the transcript
+        try:
+            summary = self.get_summary_from_text(transcript)
+            self.saving_file(summary_path, summary.encode('utf-8'))
+        except Exception:
+            summary = ''
+        # Create Meta of Video Document
+        metadata = {
+            "url": f"{url}",
+            "source": f"{url}",
+            "source_type": self._source_type,
+            'type': 'video_transcript',
+            "summary": f"{summary!s}",
+            "video_info": video_info
+        }
+        # VTT version:
+        transcript = self.transcript_to_vtt(transcript_whisper, vtt_path)
+        # second: saving transcript to a file:
+        self.saving_file(transcript_path, transcript.encode('utf-8'))
+        metadata['transcript'] = transcript_path
+        metadata["summary"] = summary
+        metadata['summary_file'] = summary_path
+        metadata["vtt"] = vtt_path
+        metadata['audio'] = audio_path
+        return metadata
+
+    def extract(self) -> list:
+        # Adding also Translation to other language.
+        documents = []
+        for url in self.urls:
+            doc = self.extract_video(url)
+            documents.append(doc)
+        return documents

parrot/tools/execute.py

@@ -0,0 +1,56 @@
+"""
+Executable Python REPL Tool.
+"""
+import io
+import base64
+import json
+import matplotlib.pyplot as plt
+from langchain_experimental.tools.python.tool import PythonAstREPLTool
+
+
+class ExecutablePythonREPLTool(PythonAstREPLTool):
+    """
+    Executable Python REPL Tool.
+    """
+    def execute_code(self, code: str) -> str:
+        """
+        Execute the provided Python code and return the output.
+
+        Args:
+            code (str): The Python code to execute.
+
+        Returns:
+            str: The output of the executed code.
+        """
+        try:
+            # Set up a namespace for execution
+            namespace = {}
+            exec(code, namespace)
+
+            # Check if a plot was created
+            if 'plt' in namespace:
+                buf = io.BytesIO()
+                plt.savefig(buf, format='png')
+                plt.close()
+                buf.seek(0)
+                # Encode the image in base64
+                # Encode the image in base64
+                img_str = base64.b64encode(buf.read()).decode('utf-8')
+
+                # Prepare the JSON output
+                result = {
+                    "image": {
+                        "format": "png",
+                        "base64": img_str
+                    }
+                }
+                # Return both the code and the JSON result
+                return f"**Code Executed**:\n```python\n{code}\n```\n\n**Result**:\n{json.dumps(result)}"
+            else:
+                return f"**Code Executed**:\n```python\n{code}\n```\n\n"
+
+        except Exception as e:
+            return f"Error executing code: {e}"
+
+    def __call__(self, code: str) -> str:
+        return self.execute_code(code)

parrot/version.py

@@ -3,7 +3,7 @@
 __title__ = "ai-parrot"
 __description__ = "Live Chatbots based on Langchain chatbots and Agents \
     Integrated into Navigator Framework or used into aiohttp applications."
-__version__ = "0.3.11"
+__version__ = "0.3.16"
 __author__ = "Jesus Lara"
 __author_email__ = "jesuslarag@gmail.com"
 __license__ = "MIT"