Spaces:
Sleeping
Sleeping
Fix audio tools and add browsing tool instead of visit web page tool
Browse files- community_tools.py +10 -2
- mini_agents.py +5 -5
- tools.py +1 -1
community_tools.py
CHANGED
@@ -3,6 +3,10 @@ from langchain_community.agent_toolkits.load_tools import load_tools
|
|
3 |
from smolagents.tools import Tool, tool
|
4 |
from youtube_transcript_api import YouTubeTranscriptApi
|
5 |
from youtube_transcript_api.proxies import WebshareProxyConfig
|
|
|
|
|
|
|
|
|
6 |
import os
|
7 |
|
8 |
google_map_tool = Tool.from_langchain(GooglePlacesTool())
|
@@ -10,11 +14,15 @@ google_map_tool = Tool.from_langchain(GooglePlacesTool())
|
|
10 |
from langchain_community.tools import WikipediaQueryRun
|
11 |
from langchain_community.utilities import WikipediaAPIWrapper
|
12 |
|
13 |
-
wikipedia_tool = Tool.from_langchain(WikipediaQueryRun(api_wrapper=WikipediaAPIWrapper()))
|
14 |
|
15 |
arxiv_tools = [Tool.from_langchain(tool) for tool in load_tools(["arxiv"])]
|
16 |
|
17 |
-
|
|
|
|
|
|
|
|
|
18 |
|
19 |
@tool
|
20 |
def get_youtube_transcript_from_url(video_url: str)->str:
|
|
|
3 |
from smolagents.tools import Tool, tool
|
4 |
from youtube_transcript_api import YouTubeTranscriptApi
|
5 |
from youtube_transcript_api.proxies import WebshareProxyConfig
|
6 |
+
from langchain_community.tools.playwright.utils import (
|
7 |
+
create_async_playwright_browser
|
8 |
+
)
|
9 |
+
from langchain_community.agent_toolkits import PlayWrightBrowserToolkit
|
10 |
import os
|
11 |
|
12 |
google_map_tool = Tool.from_langchain(GooglePlacesTool())
|
|
|
14 |
from langchain_community.tools import WikipediaQueryRun
|
15 |
from langchain_community.utilities import WikipediaAPIWrapper
|
16 |
|
17 |
+
wikipedia_tool = Tool.from_langchain(WikipediaQueryRun(api_wrapper=WikipediaAPIWrapper(top_k_results=2)))
|
18 |
|
19 |
arxiv_tools = [Tool.from_langchain(tool) for tool in load_tools(["arxiv"])]
|
20 |
|
21 |
+
async_browser = create_async_playwright_browser()
|
22 |
+
browser_toolkit = PlayWrightBrowserToolkit.from_browser(async_browser=async_browser)
|
23 |
+
browser_tools = browser_toolkit.get_tools()
|
24 |
+
|
25 |
+
community_tools = [google_map_tool, wikipedia_tool, *arxiv_tools, *browser_tools]
|
26 |
|
27 |
@tool
|
28 |
def get_youtube_transcript_from_url(video_url: str)->str:
|
mini_agents.py
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
from smolagents import CodeAgent, InferenceClientModel
|
2 |
from tools import sort_list, operate_two_numbers, convert_number, load_dataframe_from_csv, load_dataframe_from_excel
|
3 |
-
from tools import tavily_search_tool,
|
4 |
from tools import to_dataframe, to_json, get_dataframe_data, get_dataframe_column, get_dataframe_row, get_dataframe_groupby
|
5 |
from vlm_tools import image_processing, object_detection_tool, ocr_scan_tool, extract_images_from_video, get_image_from_file_path, get_video_from_file_path
|
6 |
-
from audio_tools import transcribe_audio_tool,
|
7 |
from community_tools import community_tools, get_youtube_transcript_from_url
|
8 |
import os
|
9 |
import logging
|
@@ -54,11 +54,11 @@ audio_model = InferenceClientModel(
|
|
54 |
|
55 |
audio_agent = CodeAgent(
|
56 |
model=audio_model,
|
57 |
-
tools=[transcribe_audio_tool,
|
58 |
max_steps=6,
|
59 |
additional_authorized_imports=AUTHORIZED_IMPORTS,
|
60 |
name="audio_agent",
|
61 |
-
description="This agent is responsible for processing audio,
|
62 |
)
|
63 |
|
64 |
vlm_model = InferenceClientModel(
|
@@ -136,7 +136,7 @@ master_model = InferenceClientModel(
|
|
136 |
master_agent = CodeAgent(
|
137 |
model=master_model,
|
138 |
managed_agents=[audio_agent, vlm_agent, arithmetic_agent, pandas_agent],
|
139 |
-
tools=[sort_list, get_youtube_transcript_from_url, read_python_file, *community_tools, tavily_search_tool
|
140 |
add_base_tools=True,
|
141 |
max_steps=20,
|
142 |
additional_authorized_imports=AUTHORIZED_IMPORTS,
|
|
|
1 |
from smolagents import CodeAgent, InferenceClientModel
|
2 |
from tools import sort_list, operate_two_numbers, convert_number, load_dataframe_from_csv, load_dataframe_from_excel
|
3 |
+
from tools import tavily_search_tool, read_python_file
|
4 |
from tools import to_dataframe, to_json, get_dataframe_data, get_dataframe_column, get_dataframe_row, get_dataframe_groupby
|
5 |
from vlm_tools import image_processing, object_detection_tool, ocr_scan_tool, extract_images_from_video, get_image_from_file_path, get_video_from_file_path
|
6 |
+
from audio_tools import transcribe_audio_tool, audio_to_base64_from_file_path, noise_reduction, audio_segmentation, speaker_diarization
|
7 |
from community_tools import community_tools, get_youtube_transcript_from_url
|
8 |
import os
|
9 |
import logging
|
|
|
54 |
|
55 |
audio_agent = CodeAgent(
|
56 |
model=audio_model,
|
57 |
+
tools=[transcribe_audio_tool, audio_to_base64_from_file_path, noise_reduction, audio_segmentation, speaker_diarization],
|
58 |
max_steps=6,
|
59 |
additional_authorized_imports=AUTHORIZED_IMPORTS,
|
60 |
name="audio_agent",
|
61 |
+
description="This agent is responsible for processing audio, loading mp3 audio and converting it to base64, reducing noise, segmenting audio and transcribing audio (in base64 format). It cannot process videos."
|
62 |
)
|
63 |
|
64 |
vlm_model = InferenceClientModel(
|
|
|
136 |
master_agent = CodeAgent(
|
137 |
model=master_model,
|
138 |
managed_agents=[audio_agent, vlm_agent, arithmetic_agent, pandas_agent],
|
139 |
+
tools=[sort_list, get_youtube_transcript_from_url, read_python_file, *community_tools, tavily_search_tool],
|
140 |
add_base_tools=True,
|
141 |
max_steps=20,
|
142 |
additional_authorized_imports=AUTHORIZED_IMPORTS,
|
tools.py
CHANGED
@@ -60,7 +60,7 @@ def sort_list(my_list: List[int], order: Literal["asc", "desc", "alphabetize", "
|
|
60 |
return sorted(my_list, reverse=how[order] == "desc")
|
61 |
|
62 |
#smolagents tools
|
63 |
-
visit_webpage_tool = VisitWebpageTool()
|
64 |
tavily_search_tool = Tool.from_langchain(TavilySearchResults(k=3))
|
65 |
|
66 |
|
|
|
60 |
return sorted(my_list, reverse=how[order] == "desc")
|
61 |
|
62 |
#smolagents tools
|
63 |
+
# visit_webpage_tool = VisitWebpageTool()
|
64 |
tavily_search_tool = Tool.from_langchain(TavilySearchResults(k=3))
|
65 |
|
66 |
|