File size: 15,309 Bytes
ac301bc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ab1ba02
 
 
ac301bc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2dc923b
 
 
 
 
ac301bc
 
 
 
 
 
 
 
 
2dc923b
ac301bc
 
 
 
 
 
 
2dc923b
ac301bc
2dc923b
 
 
 
 
ac301bc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2dc923b
 
 
 
 
 
 
 
ac301bc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2dc923b
ac301bc
 
 
 
2dc923b
 
 
 
 
ac301bc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a330e89
ac301bc
 
 
 
 
 
ab1ba02
ac301bc
 
 
 
 
 
ab1ba02
d8debf8
 
 
 
 
 
 
 
ab1ba02
 
f487d08
7154062
 
 
b2d35b9
ab1ba02
 
7154062
fdc3c42
7154062
 
 
 
 
 
fdc3c42
ab1ba02
7154062
b2d35b9
7154062
 
 
 
 
fdc3c42
ab1ba02
7154062
 
 
f487d08
c266c49
fdc3c42
b2d35b9
 
d8debf8
ac301bc
 
ab1ba02
ac301bc
 
 
 
 
ab1ba02
ac301bc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2dc923b
ac301bc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
# app.py
import os
import logging
import asyncio
import nest_asyncio
from datetime import datetime
import uuid
import aiohttp
import gradio as gr
import requests
import xml.etree.ElementTree as ET
import json

from langfuse.llama_index import LlamaIndexInstrumentor
from llama_index.tools.duckduckgo import DuckDuckGoSearchToolSpec
from llama_index.tools.weather import OpenWeatherMapToolSpec
from llama_index.tools.playwright import PlaywrightToolSpec
from llama_index.core.tools import FunctionTool
from llama_index.core.agent.workflow import AgentWorkflow
from llama_index.core.workflow import Context
from llama_index.llms.huggingface_api import HuggingFaceInferenceAPI
from llama_index.core.memory import ChatMemoryBuffer
from llama_index.readers.web import RssReader, SimpleWebPageReader
from llama_index.core import SummaryIndex

# Import the event types for streaming
from llama_index.core.agent.workflow import AgentStream, ToolCall, ToolCallResult

import subprocess
subprocess.run(["playwright", "install"])

# allow nested loops in Spaces
nest_asyncio.apply()

# --- Llangfuse ---
instrumentor = LlamaIndexInstrumentor(
    public_key=os.environ.get("LANGFUSE_PUBLIC_KEY"),
    secret_key=os.environ.get("LANGFUSE_SECRET_KEY"),
    host=os.environ.get("LANGFUSE_HOST"),
)
instrumentor.start()

# --- Secrets via env vars ---
HF_TOKEN            = os.getenv("HF_TOKEN")
# OPENAI_API_KEY      = os.getenv("OPENAI_API_KEY")
OPENWEATHERMAP_KEY  = os.getenv("OPENWEATHERMAP_API_KEY")
SERPER_API_KEY      = os.getenv("SERPER_API_KEY")

# --- LLMs ---
llm = HuggingFaceInferenceAPI(
    model_name="Qwen/Qwen2.5-Coder-32B-Instruct",
    token=HF_TOKEN, 
    task="conversational",
    streaming=True
)

memory = ChatMemoryBuffer.from_defaults(token_limit=8192)
today_str = datetime.now().strftime("%B %d, %Y")
ANON_USER_ID = os.environ.get("ANON_USER_ID", uuid.uuid4().hex)

# # OpenAI for pure function-calling
# openai_llm = OpenAI(
#     model="gpt-4o",
#     api_key=OPENAI_API_KEY,
#     temperature=0.0,
#     streaming=False,
# )

# --- Tools Setup ---
# DuckDuckGo
# duck_spec = DuckDuckGoSearchToolSpec()
# search_tool = FunctionTool.from_defaults(duck_spec.duckduckgo_full_search)

# Weather
openweather_api_key=OPENWEATHERMAP_KEY
weather_tool_spec = OpenWeatherMapToolSpec(key=openweather_api_key)
weather_tool = FunctionTool.from_defaults(
    weather_tool_spec.weather_at_location,
    name="current_weather",
    description="Get the current weather at a specific location (city, country)."
)
forecast_tool = FunctionTool.from_defaults(
    weather_tool_spec.forecast_tommorrow_at_location,
    name="weather_forecast",
    description="Get tomorrow's weather forecast for a specific location (city, country)."
)

# Playwright (synchronous start)
# async def _start_browser():
#     return await PlaywrightToolSpec.create_async_playwright_browser(headless=True)
# browser = asyncio.get_event_loop().run_until_complete(_start_browser())
# playwright_tool_spec = PlaywrightToolSpec.from_async_browser(browser)

# navigate_tool = FunctionTool.from_defaults(
#     playwright_tool_spec.navigate_to,
#     name="web_navigate",
#     description="Navigate to a specific URL."
# )
# extract_text_tool = FunctionTool.from_defaults(
#     playwright_tool_spec.extract_text,
#     name="web_extract_text",
#     description="Extract all text from the current page."
# )
# extract_links_tool = FunctionTool.from_defaults(
#     playwright_tool_spec.extract_hyperlinks,
#     name="web_extract_links",
#     description="Extract all hyperlinks from the current page."
# )

# Google News RSS
# def fetch_google_news_rss():
#     docs = RssReader(html_to_text=True).load_data(["https://news.google.com/rss"])
#     return [{"title":d.metadata.get("title",""), "url":d.metadata.get("link","")} for d in docs]

# -----------------------------
# Google News RSS
# -----------------------------

def fetch_news_headlines() -> str:
    """Fetches the latest news from Google News RSS feed.
    
    Returns:
        A string containing the latest news articles from Google News, or an error message if the request fails.
    """
    url = "https://news.google.com/rss"
    
    try:
        response = requests.get(url)
        response.raise_for_status()
        
        # Parse the XML content
        root = ET.fromstring(response.content)
        
        # Format the news articles into a readable string
        formatted_news = []
        for i, item in enumerate(root.findall('.//item')):
            if i >= 5:
                break
            title = item.find('title').text if item.find('title') is not None else 'N/A'
            link = item.find('link').text if item.find('link') is not None else 'N/A'
            pub_date = item.find('pubDate').text if item.find('pubDate') is not None else 'N/A'
            description = item.find('description').text if item.find('description') is not None else 'N/A'
            
            formatted_news.append(f"Title: {title}")
            formatted_news.append(f"Published: {pub_date}")
            formatted_news.append(f"Link: {link}")
            formatted_news.append(f"Description: {description}")
            formatted_news.append("---")
        
        return "\n".join(formatted_news) if formatted_news else "No news articles found."
        
    except requests.exceptions.RequestException as e:
        return f"Error fetching news: {str(e)}"
    except Exception as e:
        return f"An unexpected error occurred: {str(e)}"

google_rss_tool = FunctionTool.from_defaults(
    fn=fetch_news_headlines,
    name="fetch_google_news_rss",
    description="Fetch latest headlines."
)
# -----------------------------
# SERPER API
# -----------------------------
def fetch_news_topics(query: str) -> str:
    """Fetches news articles about a specific topic using the Serper API.
    
    Args:
        query: The topic to search for news about.
        
    Returns:
        A string containing the news articles found, or an error message if the request fails.
    """
    url = "https://google.serper.dev/news"
    
    payload = json.dumps({
        "q": query
    })
    
    headers = {
        'X-API-KEY': os.getenv('SERPER_API_KEY'),
        'Content-Type': 'application/json'
    }
    
    try:
        response = requests.post(url, headers=headers, data=payload)
        response.raise_for_status()
        
        news_data = response.json()
        
        # Format the news articles into a readable string
        formatted_news = []
        for i, article in enumerate(news_data.get('news', [])):
            if i >= 5:
                break
            formatted_news.append(f"Title: {article.get('title', 'N/A')}")
            formatted_news.append(f"Source: {article.get('source', 'N/A')}")
            formatted_news.append(f"Link: {article.get('link', 'N/A')}")
            formatted_news.append(f"Snippet: {article.get('snippet', 'N/A')}")
            formatted_news.append("---")
        
        return "\n".join(formatted_news) if formatted_news else "No news articles found."
        
    except requests.exceptions.RequestException as e:
        return f"Error fetching news: {str(e)}"
    except Exception as e:
        return f"An unexpected error occurred: {str(e)}"

serper_news_tool = FunctionTool.from_defaults(
    fetch_news_topics,
    name="fetch_news_from_serper",
    description="Fetch news articles on a specific topic."
)

# -----------------------------
# WEB PAGE READER
# -----------------------------
def summarize_webpage(url: str) -> str:
    """Fetches and summarizes the content of a web page."""
    try:
        # NOTE: the html_to_text=True option requires html2text to be installed
        documents = SimpleWebPageReader(html_to_text=True).load_data([url])
        if not documents:
            return "No content could be loaded from the provided URL."
        index = SummaryIndex.from_documents(documents)
        query_engine = index.as_query_engine()
        response = query_engine.query("Summarize the main points of this page.")
        return str(response)
    except Exception as e:
        return f"An error occurred while summarizing the web page: {str(e)}"

webpage_reader_tool = FunctionTool.from_defaults(
    summarize_webpage,
    name="summarize_webpage",
    description="Read and summarize the main points of a web page given its URL."
)

# Create the agent workflow
tools = [
    #search_tool,
    #navigate_tool,
    #extract_text_tool,
    #extract_links_tool,
    weather_tool,
    forecast_tool,
    google_rss_tool,
    serper_news_tool,
    webpage_reader_tool,
]
web_agent = AgentWorkflow.from_tools_or_functions(
    tools, 
    llm=llm,
    system_prompt="""You are a helpful assistant with access to specialized tools for retrieving information about weather, and news.
    AVAILABLE TOOLS:
    1. current_weather - Get current weather conditions for a location
    2. weather_forecast - Get tomorrow's weather forecast for a location
    3. fetch_google_news_rss - Fetch the latest general news headlines
    4. fetch_news_from_serper - Fetch news articles on a specific topic
    5. summarize_webpage - Read and summarize the content of a web page

    WHEN AND HOW TO USE EACH TOOL:

    For weather information:
    - Use current_weather when asked about present conditions
    EXAMPLE: User asks "What's the weather in Tokyo?"
    TOOL: current_weather
    PARAMETERS: {"location": "Tokyo, JP"}

    - Use weather_forecast when asked about future weather
    EXAMPLE: User asks "What will the weather be like in Paris tomorrow?"
    TOOL: weather_forecast
    PARAMETERS: {"location": "Paris, FR"}

    For news retrieval:
    - Use fetch_google_news_rss for general headlines (requires NO parameters)
    EXAMPLE: User asks "What's happening in the news today?"
    TOOL: fetch_google_news_rss
    PARAMETERS: {}

    - Use fetch_news_from_serper for specific news topics
    EXAMPLE: User asks "Any news about AI advancements?"
    TOOL: fetch_news_from_serper
    PARAMETERS: {"query": "artificial intelligence advancements"}

    For web content:
    - Use summarize_webpage to extract information from websites
    EXAMPLE: User asks "Can you summarize the content on hf.co/learn?"
    TOOL: summarize_webpage
    PARAMETERS: {"url": "https://hf.co/learn"}

    IMPORTANT GUIDELINES:
    - Always verify the format of parameters before submitting
    - For locations, use the format "City, Country Code" (e.g., "Montreal, CA")
    - For URLs, include the full address with http:// or https://
    - When multiple tools are needed to answer a complex question, use them in sequence
    - If possible, provide clickable links for your sources in your final answer.

    When you use a tool, explain to the user that you're retrieving information. After receiving the tool's output, provide a helpful summary of the information.
    """
)
ctx = Context(web_agent)

# Async helper to run agent queries (kept for compatibility)
def run_query_sync(query: str):
    """Helper to run async agent.run in sync context."""
    return asyncio.get_event_loop().run_until_complete(
        web_agent.run(query, ctx=ctx)
    )

# Updated run_query function to use stream_events
async def run_query(query: str):
    trace_id = f"agent-run-{uuid.uuid4().hex}"
    try:
        with instrumentor.observe(
            trace_id=trace_id,
            session_id="web-agent-session",
            user_id=ANON_USER_ID,
        ):
            # Start the handler
            handler = web_agent.run(query, ctx=ctx)
            
            # Keep track of what we're showing to avoid duplicates
            tool_calls_shown = set()
            
            # Stream content
            async for event in handler.stream_events():
                if isinstance(event, AgentStream):
                    # Filter out any lines starting with "Thought:" or "Action:"
                    if hasattr(event, 'delta') and event.delta:
                        delta = event.delta
                        # Filter out thought processes and internal reasoning
                        if not (delta.strip().startswith("Thought:") or 
                                delta.strip().startswith("Action:") or
                                delta.strip().startswith("Answer:")):
                            yield delta
                
                elif isinstance(event, ToolCall):
                    tool_name = getattr(event, 'name', getattr(event, 'function_name', getattr(event, 'tool_name', "unknown tool")))
                    
                    # Only show tool call message once per tool+call combo
                    tool_call_id = f"{tool_name}_{hash(str(getattr(event, 'args', '')))}"
                    if tool_call_id not in tool_calls_shown:
                        tool_calls_shown.add(tool_call_id)
                        yield f"\n\n🔧 Using tool: {tool_name}...\n"
                
                elif isinstance(event, ToolCallResult):
                    # We don't need to show the raw tool result to the user
                    # The agent will incorporate the results in its response
                    pass
            
    except Exception as e:
        yield f"\n\n❌ Error: {str(e)}\n"
        import traceback
        yield f"Traceback: {traceback.format_exc()}"
    finally:
        instrumentor.flush()

# Updated gradio_query function
async def gradio_query(user_input, chat_history=None):
    history = chat_history or []
    history.append({"role": "user", "content": user_input})
    
    # Add initial assistant message
    history.append({"role": "assistant", "content": "Processing..."})
    yield history, history
    
    # Get streaming response
    full_response = ""
    async for chunk in run_query(user_input):
        if chunk:
            full_response += chunk
            history[-1]["content"] = full_response
            yield history, history

# Build and launch Gradio app
grb = gr.Blocks()
with grb:
    gr.Markdown("## Perspicacity")
    gr.Markdown(
        """
        This bot can check the news, tell you the weather, and even browse websites to answer follow-up questions — all powered by a team of tiny AI tools working behind the scenes.\n\n
        🧪 Built for fun during the [AI Agents course](https://huggingface.co/learn/agents-course/unit0/introduction) — it's just a demo to show what agents can do.\n
        🙌 Got ideas or improvements? PRs welcome!\n\n
        👉 Try asking 'What's the weather in Montreal?' or 'What's in the news today?'
        """
    )
    chatbot = gr.Chatbot(type="messages")
    txt = gr.Textbox(placeholder="Ask me anything...", show_label=False)
    
    # Set up event handlers for streaming
    txt.submit(
        gradio_query, 
        inputs=[txt, chatbot], 
        outputs=[chatbot, chatbot]
    ).then(
        lambda: gr.update(value=""),  # Clear the textbox after submission
        None,
        [txt]
    )
    
    # Also update the button click handler
    send_btn = gr.Button("Send")
    send_btn.click(
        gradio_query, 
        [txt, chatbot], 
        [chatbot, chatbot]
    ).then(
        lambda: gr.update(value=""),  # Clear the textbox after submission
        None,
        [txt]
    )

if __name__ == "__main__":
    grb.launch()