Thanh Vinh Vo commited on
Commit ·
9440a2b
1
Parent(s): 48dcf71
update
Browse files- app.py +31 -3
- requirements.txt +1 -0
app.py
CHANGED
|
@@ -64,9 +64,11 @@ class BasicAgent:
|
|
| 64 |
model_id="meta-llama/Llama-3.2-11B-Vision-Instruct",
|
| 65 |
)
|
| 66 |
|
| 67 |
-
self.
|
| 68 |
tools=[VisitWebpageTool(), DuckDuckGoSearchTool(), get_image_content],
|
| 69 |
-
model=
|
|
|
|
|
|
|
| 70 |
additional_authorized_imports=[
|
| 71 |
"requests",
|
| 72 |
"bs4",
|
|
@@ -75,7 +77,33 @@ class BasicAgent:
|
|
| 75 |
"pandas",
|
| 76 |
"io",
|
| 77 |
"PIL",
|
|
|
|
| 78 |
],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 79 |
)
|
| 80 |
|
| 81 |
def __call__(self, question: str, file_name: str) -> str:
|
|
@@ -89,7 +117,7 @@ class BasicAgent:
|
|
| 89 |
2. `pandas` Python package is provided. Please use it when there is need to extract structured data (such as tables) from HTML content.
|
| 90 |
3. Please take the question literally! Do not add any additional information or assumptions.
|
| 91 |
"""
|
| 92 |
-
result = self.
|
| 93 |
print(f"Agent responded with: {result}")
|
| 94 |
return result
|
| 95 |
|
|
|
|
| 64 |
model_id="meta-llama/Llama-3.2-11B-Vision-Instruct",
|
| 65 |
)
|
| 66 |
|
| 67 |
+
self.multimodal_agent = CodeAgent(
|
| 68 |
tools=[VisitWebpageTool(), DuckDuckGoSearchTool(), get_image_content],
|
| 69 |
+
model=InferenceClientModel(
|
| 70 |
+
model_id="meta-llama/Llama-3.2-11B-Vision-Instruct",
|
| 71 |
+
),
|
| 72 |
additional_authorized_imports=[
|
| 73 |
"requests",
|
| 74 |
"bs4",
|
|
|
|
| 77 |
"pandas",
|
| 78 |
"io",
|
| 79 |
"PIL",
|
| 80 |
+
"chess",
|
| 81 |
],
|
| 82 |
+
name="multimodal_agent",
|
| 83 |
+
description="I can understand natural language and answer questions about images.",
|
| 84 |
+
verbosity_level=0,
|
| 85 |
+
max_steps=10,
|
| 86 |
+
)
|
| 87 |
+
|
| 88 |
+
self.manager_agent = CodeAgent(
|
| 89 |
+
model=InferenceClientModel(
|
| 90 |
+
"deepseek-ai/DeepSeek-R1", provider="together", max_tokens=8096
|
| 91 |
+
),
|
| 92 |
+
tools=[get_image_content],
|
| 93 |
+
managed_agents=[self.multimodal_agent],
|
| 94 |
+
additional_authorized_imports=[
|
| 95 |
+
"requests",
|
| 96 |
+
"bs4",
|
| 97 |
+
"markdownify",
|
| 98 |
+
"wikipedia",
|
| 99 |
+
"pandas",
|
| 100 |
+
"io",
|
| 101 |
+
"PIL",
|
| 102 |
+
"chess",
|
| 103 |
+
],
|
| 104 |
+
planning_interval=5,
|
| 105 |
+
verbosity_level=2,
|
| 106 |
+
max_steps=15,
|
| 107 |
)
|
| 108 |
|
| 109 |
def __call__(self, question: str, file_name: str) -> str:
|
|
|
|
| 117 |
2. `pandas` Python package is provided. Please use it when there is need to extract structured data (such as tables) from HTML content.
|
| 118 |
3. Please take the question literally! Do not add any additional information or assumptions.
|
| 119 |
"""
|
| 120 |
+
result = self.manager_agent.run(prompt)
|
| 121 |
print(f"Agent responded with: {result}")
|
| 122 |
return result
|
| 123 |
|
requirements.txt
CHANGED
|
@@ -7,3 +7,4 @@ markdownify
|
|
| 7 |
wikipedia
|
| 8 |
pandas
|
| 9 |
openai
|
|
|
|
|
|
| 7 |
wikipedia
|
| 8 |
pandas
|
| 9 |
openai
|
| 10 |
+
chess
|