Create agent.py
Browse files
agent.py
ADDED
|
@@ -0,0 +1,134 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
|
| 2 |
+
from linkup_utils import LinkupSearchTool
|
| 3 |
+
from text_inspector_tool import TextInspectorTool
|
| 4 |
+
from text_web_browser import (
|
| 5 |
+
ArchiveSearchTool,
|
| 6 |
+
FinderTool,
|
| 7 |
+
FindNextTool,
|
| 8 |
+
PageDownTool,
|
| 9 |
+
PageUpTool,
|
| 10 |
+
SimpleTextBrowser,
|
| 11 |
+
VisitTool,
|
| 12 |
+
)
|
| 13 |
+
from visual_qa import visualizer
|
| 14 |
+
|
| 15 |
+
from smolagents import (
|
| 16 |
+
CodeAgent,
|
| 17 |
+
LiteLLMModel,
|
| 18 |
+
ToolCallingAgent,
|
| 19 |
+
)
|
| 20 |
+
|
| 21 |
+
|
| 22 |
+
AUTHORIZED_IMPORTS = [
|
| 23 |
+
"requests",
|
| 24 |
+
"zipfile",
|
| 25 |
+
"os",
|
| 26 |
+
"pandas",
|
| 27 |
+
"numpy",
|
| 28 |
+
"sympy",
|
| 29 |
+
"json",
|
| 30 |
+
"bs4",
|
| 31 |
+
"pubchempy",
|
| 32 |
+
"xml",
|
| 33 |
+
"yahoo_finance",
|
| 34 |
+
"Bio",
|
| 35 |
+
"sklearn",
|
| 36 |
+
"scipy",
|
| 37 |
+
"pydub",
|
| 38 |
+
"io",
|
| 39 |
+
"PIL",
|
| 40 |
+
"chess",
|
| 41 |
+
"PyPDF2",
|
| 42 |
+
"pptx",
|
| 43 |
+
"torch",
|
| 44 |
+
"datetime",
|
| 45 |
+
"fractions",
|
| 46 |
+
"csv",
|
| 47 |
+
]
|
| 48 |
+
|
| 49 |
+
# load_dotenv(override=True)
|
| 50 |
+
# login(os.getenv("HF_TOKEN"))
|
| 51 |
+
|
| 52 |
+
append_answer_lock = threading.Lock()
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
def parse_args():
|
| 56 |
+
parser = argparse.ArgumentParser()
|
| 57 |
+
parser.add_argument(
|
| 58 |
+
"question", type=str, help="for example: 'How many studio albums did Mercedes Sosa release before 2007?'"
|
| 59 |
+
)
|
| 60 |
+
parser.add_argument("--model-id", type=str, default="o1")
|
| 61 |
+
return parser.parse_args()
|
| 62 |
+
|
| 63 |
+
|
| 64 |
+
custom_role_conversions = {"tool-call": "assistant", "tool-response": "user"}
|
| 65 |
+
|
| 66 |
+
user_agent = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/119.0.0.0 Safari/537.36 Edg/119.0.0.0"
|
| 67 |
+
|
| 68 |
+
BROWSER_CONFIG = {
|
| 69 |
+
"viewport_size": 1024 * 5,
|
| 70 |
+
"downloads_folder": "downloads_folder",
|
| 71 |
+
"request_kwargs": {
|
| 72 |
+
"headers": {"User-Agent": user_agent},
|
| 73 |
+
"timeout": 300,
|
| 74 |
+
},
|
| 75 |
+
"serpapi_key": os.getenv("SERPAPI_API_KEY"),
|
| 76 |
+
}
|
| 77 |
+
|
| 78 |
+
os.makedirs(f"./{BROWSER_CONFIG['downloads_folder']}", exist_ok=True)
|
| 79 |
+
|
| 80 |
+
|
| 81 |
+
def create_agent(model_id="gpt-4o-mini"):
|
| 82 |
+
model_params = {
|
| 83 |
+
"model_id": model_id,
|
| 84 |
+
"custom_role_conversions": custom_role_conversions,
|
| 85 |
+
"max_completion_tokens": 8192,
|
| 86 |
+
}
|
| 87 |
+
if model_id == "o1":
|
| 88 |
+
model_params["reasoning_effort"] = "high"
|
| 89 |
+
model = LiteLLMModel(**model_params)
|
| 90 |
+
|
| 91 |
+
text_limit = 100000
|
| 92 |
+
browser = SimpleTextBrowser(**BROWSER_CONFIG)
|
| 93 |
+
WEB_TOOLS = [
|
| 94 |
+
LinkupSearchTool(),
|
| 95 |
+
VisitTool(browser),
|
| 96 |
+
PageUpTool(browser),
|
| 97 |
+
PageDownTool(browser),
|
| 98 |
+
FinderTool(browser),
|
| 99 |
+
FindNextTool(browser),
|
| 100 |
+
ArchiveSearchTool(browser),
|
| 101 |
+
TextInspectorTool(model, text_limit),
|
| 102 |
+
]
|
| 103 |
+
text_webbrowser_agent = ToolCallingAgent(
|
| 104 |
+
model=model,
|
| 105 |
+
tools=WEB_TOOLS,
|
| 106 |
+
max_steps=20,
|
| 107 |
+
verbosity_level=2,
|
| 108 |
+
planning_interval=4,
|
| 109 |
+
name="search_agent",
|
| 110 |
+
description="""A team member that will search the internet to answer your question.
|
| 111 |
+
Ask him for all your questions that require browsing the web.
|
| 112 |
+
Note that this agent is using a powerful language model and it can do the search and analyse the results.
|
| 113 |
+
You should ask question in a way to let the language model to perform the best, i.e. provide as much context as possible and ask in a clear way.
|
| 114 |
+
Provide him as much context as possible, in particular if you need to search on a specific timeframe!
|
| 115 |
+
And don't hesitate to provide him with a complex search task, like finding a difference between two webpages.
|
| 116 |
+
Your request must be a real sentence, not a google search! Like "Find me this information (...)" rather than a few keywords.
|
| 117 |
+
""",
|
| 118 |
+
provide_run_summary=True,
|
| 119 |
+
)
|
| 120 |
+
text_webbrowser_agent.prompt_templates["managed_agent"]["task"] += """You can navigate to .txt online files.
|
| 121 |
+
If a non-html page is in another format, especially .pdf or a Youtube video, use tool 'inspect_file_as_text' to inspect it.
|
| 122 |
+
Additionally, if after some searching you find out that you need more information to answer the question, you can use `final_answer` with your request for clarification as argument to request for more information."""
|
| 123 |
+
|
| 124 |
+
manager_agent = CodeAgent(
|
| 125 |
+
model=model,
|
| 126 |
+
tools=[visualizer, TextInspectorTool(model, text_limit)],
|
| 127 |
+
max_steps=12,
|
| 128 |
+
verbosity_level=2,
|
| 129 |
+
additional_authorized_imports=AUTHORIZED_IMPORTS,
|
| 130 |
+
planning_interval=4,
|
| 131 |
+
managed_agents=[text_webbrowser_agent],
|
| 132 |
+
)
|
| 133 |
+
|
| 134 |
+
return manager_agent
|