Spaces:
Sleeping
Sleeping
Leonardo Kamigauti
commited on
Commit
·
c4f01da
1
Parent(s):
b661c94
Try make api secrets via user prompt
Browse files
app.py
CHANGED
@@ -4,56 +4,53 @@ import requests
|
|
4 |
import pytz
|
5 |
import yaml
|
6 |
from tools.final_answer import FinalAnswerTool
|
7 |
-
# from typing import Optional
|
8 |
|
9 |
from kaggle.api.kaggle_api_extended import KaggleApi
|
10 |
import os
|
11 |
|
12 |
from Gradio_UI import GradioUI
|
13 |
|
14 |
-
|
15 |
-
|
16 |
-
def my_custom_tool(arg1:str, arg2:int)-> str: #it's import to specify the return type
|
17 |
-
#Keep this format for the description / args / args description but feel free to modify the tool
|
18 |
-
"""A tool that does nothing yet
|
19 |
-
Args:
|
20 |
-
arg1: the first argument
|
21 |
-
arg2: the second argument
|
22 |
-
"""
|
23 |
-
return "What magic will you build ?"
|
24 |
|
25 |
@tool
|
26 |
-
def
|
27 |
-
"""
|
28 |
Args:
|
29 |
-
|
30 |
-
|
31 |
-
kaggle_key: Your Kaggle API key.
|
32 |
-
max_results: Maximum number of results to return.
|
33 |
"""
|
34 |
-
|
|
|
35 |
api = KaggleApi()
|
36 |
-
|
37 |
-
# Authenticate using provided credentials
|
38 |
-
if kaggle_username and kaggle_key:
|
39 |
-
# Create a temporary kaggle.json file
|
40 |
-
kaggle_json_content = f'{{"username":"{kaggle_username}","key":"{kaggle_key}"}}'
|
41 |
-
kaggle_json_path = os.path.expanduser("~/.kaggle/kaggle.json")
|
42 |
-
os.makedirs(os.path.dirname(kaggle_json_path), exist_ok=True)
|
43 |
-
with open(kaggle_json_path, "w") as f:
|
44 |
-
f.write(kaggle_json_content)
|
45 |
-
os.chmod(kaggle_json_path, 0o600) # Set permissions to read/write for the owner only
|
46 |
-
else:
|
47 |
-
# Use the default kaggle.json file if no credentials are provided
|
48 |
-
return 'Error in searching Kaggle datasets: No username or key provided.'
|
49 |
-
|
50 |
try:
|
51 |
api.authenticate()
|
52 |
except Exception as e:
|
53 |
return f"Error authenticating with Kaggle: {str(e)}"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
|
55 |
# Search for datasets
|
56 |
-
datasets =
|
57 |
|
58 |
# Limit the number of results
|
59 |
datasets = datasets[:max_results]
|
@@ -61,7 +58,7 @@ def search_kaggle_datasets(search_term:str, kaggle_username:str = None, kaggle_k
|
|
61 |
# Extract relevant information
|
62 |
results = []
|
63 |
for dataset in datasets:
|
64 |
-
dataset_info =
|
65 |
results.append({
|
66 |
'title': dataset_info['title'],
|
67 |
'url': f"https://www.kaggle.com/{dataset_info['ref']}",
|
@@ -69,59 +66,27 @@ def search_kaggle_datasets(search_term:str, kaggle_username:str = None, kaggle_k
|
|
69 |
'files': dataset_info['files'],
|
70 |
'last_updated': dataset_info['lastUpdated']
|
71 |
})
|
72 |
-
|
73 |
-
# Clean up the temporary kaggle.json file if it was created
|
74 |
-
if kaggle_username and kaggle_key:
|
75 |
-
os.remove(kaggle_json_path)
|
76 |
-
|
77 |
return results
|
78 |
|
79 |
@tool
|
80 |
def download_kaggle_dataset(
|
|
|
81 |
dataset_ref: str,
|
82 |
download_path: str,
|
83 |
-
kaggle_username: str = None,
|
84 |
-
kaggle_key: str = None,
|
85 |
unzip: bool = True
|
86 |
) -> str:
|
87 |
"""Download a dataset from Kaggle.
|
88 |
Args:
|
|
|
89 |
dataset_ref: The reference of the dataset (e.g., "username/dataset-name").
|
90 |
download_path: The directory where the dataset will be downloaded.
|
91 |
-
kaggle_username: Your Kaggle username (from kaggle.json).
|
92 |
-
kaggle_key: Your Kaggle API key (from kaggle.json).
|
93 |
unzip: Whether to unzip the dataset after downloading. Default is True.
|
94 |
"""
|
95 |
-
# Initialize the Kaggle API
|
96 |
-
api = KaggleApi()
|
97 |
-
|
98 |
-
# Authenticate using provided credentials
|
99 |
-
if kaggle_username and kaggle_key:
|
100 |
-
# Create a temporary kaggle.json file
|
101 |
-
kaggle_json_content = f'{{"username":"{kaggle_username}","key":"{kaggle_key}"}}'
|
102 |
-
kaggle_json_path = os.path.expanduser("~/.kaggle/kaggle.json")
|
103 |
-
os.makedirs(os.path.dirname(kaggle_json_path), exist_ok=True)
|
104 |
-
with open(kaggle_json_path, "w") as f:
|
105 |
-
f.write(kaggle_json_content)
|
106 |
-
os.chmod(kaggle_json_path, 0o600) # Set permissions to read/write for the owner only
|
107 |
-
else:
|
108 |
-
# Use the default kaggle.json file if no credentials are provided
|
109 |
-
pass
|
110 |
-
|
111 |
-
try:
|
112 |
-
api.authenticate()
|
113 |
-
except Exception as e:
|
114 |
-
return f"Error authenticating with Kaggle: {str(e)}"
|
115 |
-
|
116 |
# Ensure the download path exists
|
117 |
os.makedirs(download_path, exist_ok=True)
|
118 |
|
119 |
# Download the dataset
|
120 |
-
|
121 |
-
|
122 |
-
# Clean up the temporary kaggle.json file if it was created
|
123 |
-
if kaggle_username and kaggle_key:
|
124 |
-
os.remove(kaggle_json_path)
|
125 |
|
126 |
return f"Dataset '{dataset_ref}' downloaded to '{download_path}'."
|
127 |
|
@@ -137,7 +102,6 @@ model_id='Qwen/Qwen2.5-Coder-32B-Instruct',# it is possible that this model may
|
|
137 |
custom_role_conversions=None,
|
138 |
)
|
139 |
|
140 |
-
|
141 |
# Import tool from Hub
|
142 |
image_generation_tool = load_tool("agents-course/text-to-image", trust_remote_code=True)
|
143 |
|
@@ -146,7 +110,12 @@ with open("prompts.yaml", 'r') as stream:
|
|
146 |
|
147 |
agent = CodeAgent(
|
148 |
model=model,
|
149 |
-
tools=[final_answer,
|
|
|
|
|
|
|
|
|
|
|
150 |
max_steps=6,
|
151 |
verbosity_level=1,
|
152 |
grammar=None,
|
@@ -154,7 +123,9 @@ agent = CodeAgent(
|
|
154 |
name=None,
|
155 |
description=None,
|
156 |
prompt_templates=prompt_templates,
|
157 |
-
additional_authorized_imports=['pandas',
|
|
|
|
|
158 |
)
|
159 |
|
160 |
|
|
|
4 |
import pytz
|
5 |
import yaml
|
6 |
from tools.final_answer import FinalAnswerTool
|
|
|
7 |
|
8 |
from kaggle.api.kaggle_api_extended import KaggleApi
|
9 |
import os
|
10 |
|
11 |
from Gradio_UI import GradioUI
|
12 |
|
13 |
+
os.environ['KAGGLE_USERNAME'] = ''
|
14 |
+
os.environ['KAGGLE_KEY'] = ''
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
|
16 |
@tool
|
17 |
+
def auth_kaggle(username:str, key:str) -> KaggleApi:
|
18 |
+
"""Authenticate with Kaggle using the provided username and key.
|
19 |
Args:
|
20 |
+
username (str): Username for Kaggle.
|
21 |
+
key (str): API key for Kaggle.
|
|
|
|
|
22 |
"""
|
23 |
+
os.environ['KAGGLE_USERNAME'] = username
|
24 |
+
os.environ['KAGGLE_KEY'] = key
|
25 |
api = KaggleApi()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
try:
|
27 |
api.authenticate()
|
28 |
except Exception as e:
|
29 |
return f"Error authenticating with Kaggle: {str(e)}"
|
30 |
+
return api
|
31 |
+
|
32 |
+
@tool
|
33 |
+
def clean_kaggle_auth() -> str:
|
34 |
+
"""Remove Kaggle credentials from the environment variables for security purposes.
|
35 |
+
"""
|
36 |
+
os.environ['KAGGLE_USERNAME'] = ''
|
37 |
+
os.environ['KAGGLE_KEY'] = ''
|
38 |
+
return 'Kaggle credentials removed'
|
39 |
+
|
40 |
+
@tool
|
41 |
+
def search_kaggle_datasets(kaggle_api:KaggleApi,
|
42 |
+
search_term:str,
|
43 |
+
max_results:int = 10
|
44 |
+
) -> list[dict[str]]:
|
45 |
+
"""Search for datasets on Kaggle based on a search term and return list of datasets metadata.
|
46 |
+
Args:
|
47 |
+
kaggle_api: Kaggle API object with authentication.
|
48 |
+
search_term: The term to search for.
|
49 |
+
max_results: Maximum number of results to return.
|
50 |
+
"""
|
51 |
|
52 |
# Search for datasets
|
53 |
+
datasets = kaggle_api.dataset_list(search=search_term)
|
54 |
|
55 |
# Limit the number of results
|
56 |
datasets = datasets[:max_results]
|
|
|
58 |
# Extract relevant information
|
59 |
results = []
|
60 |
for dataset in datasets:
|
61 |
+
dataset_info = kaggle_api.dataset_view(dataset)
|
62 |
results.append({
|
63 |
'title': dataset_info['title'],
|
64 |
'url': f"https://www.kaggle.com/{dataset_info['ref']}",
|
|
|
66 |
'files': dataset_info['files'],
|
67 |
'last_updated': dataset_info['lastUpdated']
|
68 |
})
|
|
|
|
|
|
|
|
|
|
|
69 |
return results
|
70 |
|
71 |
@tool
|
72 |
def download_kaggle_dataset(
|
73 |
+
kaggle_api: KaggleApi,
|
74 |
dataset_ref: str,
|
75 |
download_path: str,
|
|
|
|
|
76 |
unzip: bool = True
|
77 |
) -> str:
|
78 |
"""Download a dataset from Kaggle.
|
79 |
Args:
|
80 |
+
kaggle_api: Kaggle API object with authentication.
|
81 |
dataset_ref: The reference of the dataset (e.g., "username/dataset-name").
|
82 |
download_path: The directory where the dataset will be downloaded.
|
|
|
|
|
83 |
unzip: Whether to unzip the dataset after downloading. Default is True.
|
84 |
"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
85 |
# Ensure the download path exists
|
86 |
os.makedirs(download_path, exist_ok=True)
|
87 |
|
88 |
# Download the dataset
|
89 |
+
kaggle_api.dataset_download_files(dataset_ref, path=download_path, unzip=unzip)
|
|
|
|
|
|
|
|
|
90 |
|
91 |
return f"Dataset '{dataset_ref}' downloaded to '{download_path}'."
|
92 |
|
|
|
102 |
custom_role_conversions=None,
|
103 |
)
|
104 |
|
|
|
105 |
# Import tool from Hub
|
106 |
image_generation_tool = load_tool("agents-course/text-to-image", trust_remote_code=True)
|
107 |
|
|
|
110 |
|
111 |
agent = CodeAgent(
|
112 |
model=model,
|
113 |
+
tools=[final_answer,
|
114 |
+
auth_kaggle,
|
115 |
+
clean_kaggle_auth,
|
116 |
+
search_kaggle_datasets,
|
117 |
+
download_kaggle_dataset,
|
118 |
+
image_generation_tool],
|
119 |
max_steps=6,
|
120 |
verbosity_level=1,
|
121 |
grammar=None,
|
|
|
123 |
name=None,
|
124 |
description=None,
|
125 |
prompt_templates=prompt_templates,
|
126 |
+
additional_authorized_imports=['pandas',
|
127 |
+
'matplotlib',
|
128 |
+
'seaborn'],
|
129 |
)
|
130 |
|
131 |
|