Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,4 +1,5 @@
|
|
| 1 |
-
from smolagents import CodeAgent, DuckDuckGoSearchTool, HfApiModel, load_tool
|
|
|
|
| 2 |
import datetime
|
| 3 |
import requests
|
| 4 |
import pytz
|
|
@@ -14,17 +15,7 @@ from transformers import pipeline
|
|
| 14 |
API_KEY = os.getenv("Weather_Token")
|
| 15 |
|
| 16 |
# -------------------- TOOL 1: Get Weather --------------------
|
| 17 |
-
@tool
|
| 18 |
def get_current_weather(place: str) -> str:
|
| 19 |
-
"""
|
| 20 |
-
A tool that fetches the current weather of a particular place.
|
| 21 |
-
|
| 22 |
-
Args:
|
| 23 |
-
place (str): A string representing a valid place (e.g., 'London/Paris').
|
| 24 |
-
|
| 25 |
-
Returns:
|
| 26 |
-
str: Weather description including condition, temperature, humidity, and wind speed.
|
| 27 |
-
"""
|
| 28 |
api_key = API_KEY
|
| 29 |
url = "https://api.openweathermap.org/data/2.5/weather"
|
| 30 |
params = {
|
|
@@ -32,17 +23,14 @@ def get_current_weather(place: str) -> str:
|
|
| 32 |
"appid": api_key,
|
| 33 |
"units": "metric"
|
| 34 |
}
|
| 35 |
-
|
| 36 |
try:
|
| 37 |
response = requests.get(url, params=params)
|
| 38 |
data = response.json()
|
| 39 |
-
|
| 40 |
if response.status_code == 200:
|
| 41 |
weather_desc = data["weather"][0]["description"]
|
| 42 |
temperature = data["main"]["temp"]
|
| 43 |
humidity = data["main"]["humidity"]
|
| 44 |
wind_speed = data["wind"]["speed"]
|
| 45 |
-
|
| 46 |
return (
|
| 47 |
f"Weather in {place}:\n"
|
| 48 |
f"- Condition: {weather_desc}\n"
|
|
@@ -55,19 +43,10 @@ def get_current_weather(place: str) -> str:
|
|
| 55 |
except Exception as e:
|
| 56 |
return f"Error fetching weather data for '{place}': {str(e)}"
|
| 57 |
|
|
|
|
| 58 |
|
| 59 |
# -------------------- TOOL 2: Get Time --------------------
|
| 60 |
-
@tool
|
| 61 |
def get_current_time_in_timezone(timezone: str) -> str:
|
| 62 |
-
"""
|
| 63 |
-
A tool that fetches the current local time in a specified timezone.
|
| 64 |
-
|
| 65 |
-
Args:
|
| 66 |
-
timezone (str): A string representing a valid timezone (e.g., 'America/New_York').
|
| 67 |
-
|
| 68 |
-
Returns:
|
| 69 |
-
str: The current local time formatted as a string.
|
| 70 |
-
"""
|
| 71 |
try:
|
| 72 |
tz = pytz.timezone(timezone)
|
| 73 |
local_time = datetime.datetime.now(tz).strftime("%Y-%m-%d %H:%M:%S")
|
|
@@ -75,30 +54,14 @@ def get_current_time_in_timezone(timezone: str) -> str:
|
|
| 75 |
except Exception as e:
|
| 76 |
return f"Error fetching time for timezone '{timezone}': {str(e)}"
|
| 77 |
|
|
|
|
| 78 |
|
| 79 |
# -------------------- TOOL 3: Document QnA --------------------
|
| 80 |
embedding_model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
|
| 81 |
qa_pipeline = pipeline("text2text-generation", model="google/flan-t5-base")
|
| 82 |
|
| 83 |
-
from smolagents.tools import Tool
|
| 84 |
-
|
| 85 |
-
@Tool(name="document_qna_tool", description="Answer questions about a PDF document.")
|
| 86 |
def document_qna_tool(pdf_path: str, question: str) -> str:
|
| 87 |
-
|
| 88 |
-
"""
|
| 89 |
-
A tool that answers natural language questions about a given PDF document.
|
| 90 |
-
|
| 91 |
-
Args:
|
| 92 |
-
pdf_path (str): Path to the local PDF file.
|
| 93 |
-
question (str): Question about the content of the PDF.
|
| 94 |
-
|
| 95 |
-
Returns:
|
| 96 |
-
str: Answer to the question based on the content.
|
| 97 |
-
"""
|
| 98 |
import os, fitz, traceback
|
| 99 |
-
from sentence_transformers import SentenceTransformer, util
|
| 100 |
-
from transformers import pipeline
|
| 101 |
-
|
| 102 |
try:
|
| 103 |
print(f"[DEBUG] PDF Path: {pdf_path}")
|
| 104 |
print(f"[DEBUG] Question: {question}")
|
|
@@ -106,50 +69,32 @@ def document_qna_tool(pdf_path: str, question: str) -> str:
|
|
| 106 |
if not os.path.exists(pdf_path):
|
| 107 |
return f"[ERROR] File not found: {pdf_path}"
|
| 108 |
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
doc = fitz.open(pdf_path)
|
| 112 |
-
except RuntimeError as e:
|
| 113 |
-
return f"[ERROR] Could not open PDF. It may be corrupted or encrypted. Details: {str(e)}"
|
| 114 |
-
|
| 115 |
-
text_chunks = []
|
| 116 |
-
for page in doc:
|
| 117 |
-
text = page.get_text()
|
| 118 |
-
if text.strip():
|
| 119 |
-
text_chunks.append(text)
|
| 120 |
doc.close()
|
| 121 |
|
| 122 |
if not text_chunks:
|
| 123 |
return "[ERROR] No readable text in the PDF."
|
| 124 |
|
| 125 |
-
print(f"[DEBUG] Extracted {len(text_chunks)} text chunks.")
|
| 126 |
-
print(f"[DEBUG] First text chunk preview:\n{text_chunks[0][:300]}...")
|
| 127 |
-
|
| 128 |
embeddings = embedding_model.encode(text_chunks, convert_to_tensor=True)
|
| 129 |
question_embedding = embedding_model.encode(question, convert_to_tensor=True)
|
| 130 |
|
| 131 |
-
print("[DEBUG] Performing semantic search...")
|
| 132 |
scores = util.pytorch_cos_sim(question_embedding, embeddings)[0]
|
| 133 |
-
|
| 134 |
-
print(f"[DEBUG] Similarity scores: {scores}")
|
| 135 |
-
|
| 136 |
if scores.shape[0] == 0:
|
| 137 |
return "[ERROR] No semantic matches found in PDF text."
|
| 138 |
|
| 139 |
best_match_idx = scores.argmax().item()
|
| 140 |
best_context = text_chunks[best_match_idx]
|
| 141 |
|
| 142 |
-
print(f"[DEBUG] Best context preview:\n{best_context[:300]}...")
|
| 143 |
-
|
| 144 |
prompt = f"Context: {best_context}\nQuestion: {question}"
|
| 145 |
-
print("[DEBUG] Calling QA model...")
|
| 146 |
answer = qa_pipeline(prompt, max_new_tokens=100)[0]['generated_text']
|
| 147 |
-
|
| 148 |
return f"Answer: {answer.strip()}"
|
| 149 |
|
| 150 |
except Exception as e:
|
| 151 |
return f"[EXCEPTION] {type(e).__name__}: {str(e)}\n{traceback.format_exc()}"
|
| 152 |
|
|
|
|
|
|
|
| 153 |
# -------------------- Other Components --------------------
|
| 154 |
final_answer = FinalAnswerTool()
|
| 155 |
search_tool = DuckDuckGoSearchTool()
|
|
@@ -173,7 +118,7 @@ agent = CodeAgent(
|
|
| 173 |
get_current_weather,
|
| 174 |
image_generation_tool,
|
| 175 |
search_tool,
|
| 176 |
-
document_qna_tool,
|
| 177 |
final_answer
|
| 178 |
],
|
| 179 |
max_steps=6,
|
|
@@ -184,8 +129,9 @@ agent = CodeAgent(
|
|
| 184 |
description=None,
|
| 185 |
prompt_templates=prompt_templates
|
| 186 |
)
|
|
|
|
| 187 |
print("[DEBUG] Registered Tools:")
|
| 188 |
for t in agent.tools:
|
| 189 |
print(f" - {getattr(t, 'name', str(t))}")
|
| 190 |
|
| 191 |
-
GradioUI(agent).launch()
|
|
|
|
| 1 |
+
from smolagents import CodeAgent, DuckDuckGoSearchTool, HfApiModel, load_tool
|
| 2 |
+
from smolagents.tools import Tool
|
| 3 |
import datetime
|
| 4 |
import requests
|
| 5 |
import pytz
|
|
|
|
| 15 |
API_KEY = os.getenv("Weather_Token")
|
| 16 |
|
| 17 |
# -------------------- TOOL 1: Get Weather --------------------
|
|
|
|
| 18 |
def get_current_weather(place: str) -> str:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
api_key = API_KEY
|
| 20 |
url = "https://api.openweathermap.org/data/2.5/weather"
|
| 21 |
params = {
|
|
|
|
| 23 |
"appid": api_key,
|
| 24 |
"units": "metric"
|
| 25 |
}
|
|
|
|
| 26 |
try:
|
| 27 |
response = requests.get(url, params=params)
|
| 28 |
data = response.json()
|
|
|
|
| 29 |
if response.status_code == 200:
|
| 30 |
weather_desc = data["weather"][0]["description"]
|
| 31 |
temperature = data["main"]["temp"]
|
| 32 |
humidity = data["main"]["humidity"]
|
| 33 |
wind_speed = data["wind"]["speed"]
|
|
|
|
| 34 |
return (
|
| 35 |
f"Weather in {place}:\n"
|
| 36 |
f"- Condition: {weather_desc}\n"
|
|
|
|
| 43 |
except Exception as e:
|
| 44 |
return f"Error fetching weather data for '{place}': {str(e)}"
|
| 45 |
|
| 46 |
+
get_current_weather = Tool(name="get_current_weather", description="Fetch current weather for a given place")(get_current_weather)
|
| 47 |
|
| 48 |
# -------------------- TOOL 2: Get Time --------------------
|
|
|
|
| 49 |
def get_current_time_in_timezone(timezone: str) -> str:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 50 |
try:
|
| 51 |
tz = pytz.timezone(timezone)
|
| 52 |
local_time = datetime.datetime.now(tz).strftime("%Y-%m-%d %H:%M:%S")
|
|
|
|
| 54 |
except Exception as e:
|
| 55 |
return f"Error fetching time for timezone '{timezone}': {str(e)}"
|
| 56 |
|
| 57 |
+
get_current_time_in_timezone = Tool(name="get_current_time_in_timezone", description="Fetch local time for a given timezone")(get_current_time_in_timezone)
|
| 58 |
|
| 59 |
# -------------------- TOOL 3: Document QnA --------------------
|
| 60 |
embedding_model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
|
| 61 |
qa_pipeline = pipeline("text2text-generation", model="google/flan-t5-base")
|
| 62 |
|
|
|
|
|
|
|
|
|
|
| 63 |
def document_qna_tool(pdf_path: str, question: str) -> str:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 64 |
import os, fitz, traceback
|
|
|
|
|
|
|
|
|
|
| 65 |
try:
|
| 66 |
print(f"[DEBUG] PDF Path: {pdf_path}")
|
| 67 |
print(f"[DEBUG] Question: {question}")
|
|
|
|
| 69 |
if not os.path.exists(pdf_path):
|
| 70 |
return f"[ERROR] File not found: {pdf_path}"
|
| 71 |
|
| 72 |
+
doc = fitz.open(pdf_path)
|
| 73 |
+
text_chunks = [page.get_text() for page in doc if page.get_text().strip()]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 74 |
doc.close()
|
| 75 |
|
| 76 |
if not text_chunks:
|
| 77 |
return "[ERROR] No readable text in the PDF."
|
| 78 |
|
|
|
|
|
|
|
|
|
|
| 79 |
embeddings = embedding_model.encode(text_chunks, convert_to_tensor=True)
|
| 80 |
question_embedding = embedding_model.encode(question, convert_to_tensor=True)
|
| 81 |
|
|
|
|
| 82 |
scores = util.pytorch_cos_sim(question_embedding, embeddings)[0]
|
|
|
|
|
|
|
|
|
|
| 83 |
if scores.shape[0] == 0:
|
| 84 |
return "[ERROR] No semantic matches found in PDF text."
|
| 85 |
|
| 86 |
best_match_idx = scores.argmax().item()
|
| 87 |
best_context = text_chunks[best_match_idx]
|
| 88 |
|
|
|
|
|
|
|
| 89 |
prompt = f"Context: {best_context}\nQuestion: {question}"
|
|
|
|
| 90 |
answer = qa_pipeline(prompt, max_new_tokens=100)[0]['generated_text']
|
|
|
|
| 91 |
return f"Answer: {answer.strip()}"
|
| 92 |
|
| 93 |
except Exception as e:
|
| 94 |
return f"[EXCEPTION] {type(e).__name__}: {str(e)}\n{traceback.format_exc()}"
|
| 95 |
|
| 96 |
+
document_qna_tool = Tool(name="document_qna_tool", description="Answer questions about a PDF document")(document_qna_tool)
|
| 97 |
+
|
| 98 |
# -------------------- Other Components --------------------
|
| 99 |
final_answer = FinalAnswerTool()
|
| 100 |
search_tool = DuckDuckGoSearchTool()
|
|
|
|
| 118 |
get_current_weather,
|
| 119 |
image_generation_tool,
|
| 120 |
search_tool,
|
| 121 |
+
document_qna_tool,
|
| 122 |
final_answer
|
| 123 |
],
|
| 124 |
max_steps=6,
|
|
|
|
| 129 |
description=None,
|
| 130 |
prompt_templates=prompt_templates
|
| 131 |
)
|
| 132 |
+
|
| 133 |
print("[DEBUG] Registered Tools:")
|
| 134 |
for t in agent.tools:
|
| 135 |
print(f" - {getattr(t, 'name', str(t))}")
|
| 136 |
|
| 137 |
+
GradioUI(agent).launch()
|