Update app.py
Browse files
app.py
CHANGED
@@ -1,12 +1,13 @@
|
|
1 |
# File: main/app.py
|
2 |
-
# Purpose: One Space that offers
|
3 |
-
# 1) Fetch
|
4 |
-
# 2) Websearch
|
|
|
5 |
#
|
6 |
# Notes:
|
7 |
-
# - Launched with mcp_server=True so
|
8 |
-
# - UI uses TabbedInterface
|
9 |
-
# - Inline comments
|
10 |
|
11 |
from __future__ import annotations
|
12 |
|
@@ -19,9 +20,12 @@ from bs4 import BeautifulSoup # (layman) for parsing HTML
|
|
19 |
from readability import Document # (layman) to isolate main readable content
|
20 |
from urllib.parse import urljoin, urldefrag, urlparse # (layman) to fix/clean URLs
|
21 |
|
22 |
-
#
|
23 |
from langchain_community.tools import DuckDuckGoSearchResults
|
24 |
|
|
|
|
|
|
|
25 |
|
26 |
# ==============================
|
27 |
# Fetch: HTTP + extraction utils
|
@@ -299,9 +303,9 @@ def extract_relevant( # <-- MCP tool #1
|
|
299 |
return md or "No content could be extracted."
|
300 |
|
301 |
|
302 |
-
#
|
303 |
-
# Websearch: DuckDuckGo
|
304 |
-
#
|
305 |
|
306 |
def web_search( # <-- MCP tool #2
|
307 |
input_query: str,
|
@@ -321,8 +325,73 @@ def web_search( # <-- MCP tool #2
|
|
321 |
return results
|
322 |
|
323 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
324 |
# =====================
|
325 |
-
# UI:
|
326 |
# =====================
|
327 |
|
328 |
# --- Fetch tab (compact controllable extraction) ---
|
@@ -344,7 +413,7 @@ fetch_interface = gr.Interface(
|
|
344 |
theme="Nymbo/Nymbo_Theme",
|
345 |
)
|
346 |
|
347 |
-
# --- Websearch tab (
|
348 |
websearch_interface = gr.Interface(
|
349 |
fn=web_search, # (layman) connect the function to the UI
|
350 |
inputs=[
|
@@ -352,20 +421,43 @@ websearch_interface = gr.Interface(
|
|
352 |
gr.Slider(minimum=1, maximum=20, value=5, step=1, label="Max results"),
|
353 |
],
|
354 |
outputs=gr.JSON(label="Search results"),
|
355 |
-
title="Websearch β DuckDuckGo",
|
356 |
-
description="Search the web using DuckDuckGo; returns snippet, title, and link.",
|
357 |
allow_flagging="never",
|
358 |
theme="Nymbo/Nymbo_Theme",
|
359 |
)
|
360 |
|
361 |
-
# ---
|
362 |
-
|
363 |
-
|
364 |
-
|
365 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
366 |
theme="Nymbo/Nymbo_Theme",
|
367 |
)
|
368 |
|
369 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
370 |
if __name__ == "__main__":
|
371 |
demo.launch(mcp_server=True)
|
|
|
1 |
# File: main/app.py
|
2 |
+
# Purpose: One Space that offers three tools in one MCP server:
|
3 |
+
# 1) Fetch β extract relevant page content (title, metadata, clean text, hyperlinks)
|
4 |
+
# 2) Websearch β structured DuckDuckGo results (JSON via LangChain wrapper)
|
5 |
+
# 3) DDG (Unstructured) β compact plain-text DuckDuckGo results for low token usage
|
6 |
#
|
7 |
# Notes:
|
8 |
+
# - Launched with mcp_server=True so all functions are exposed as MCP tools.
|
9 |
+
# - UI uses TabbedInterface: each tool has its own tab.
|
10 |
+
# - Inline comments describe each section in plain language.
|
11 |
|
12 |
from __future__ import annotations
|
13 |
|
|
|
20 |
from readability import Document # (layman) to isolate main readable content
|
21 |
from urllib.parse import urljoin, urldefrag, urlparse # (layman) to fix/clean URLs
|
22 |
|
23 |
+
# Structured DDG search (LangChain wrapper)
|
24 |
from langchain_community.tools import DuckDuckGoSearchResults
|
25 |
|
26 |
+
# Unstructured DDG search (lightweight direct client)
|
27 |
+
from duckduckgo_search import DDGS
|
28 |
+
|
29 |
|
30 |
# ==============================
|
31 |
# Fetch: HTTP + extraction utils
|
|
|
303 |
return md or "No content could be extracted."
|
304 |
|
305 |
|
306 |
+
# ========================================
|
307 |
+
# Websearch (Structured): DuckDuckGo (JSON)
|
308 |
+
# ========================================
|
309 |
|
310 |
def web_search( # <-- MCP tool #2
|
311 |
input_query: str,
|
|
|
325 |
return results
|
326 |
|
327 |
|
328 |
+
# ===================================================
|
329 |
+
# DDG (Unstructured): compact plain-text, low tokens
|
330 |
+
# ===================================================
|
331 |
+
|
332 |
+
def web_search_unstructured( # <-- MCP tool #3
|
333 |
+
input_query: str,
|
334 |
+
max_results: int = 5,
|
335 |
+
style: Literal["urls", "titles+urls", "titles+urls+snippets"] = "titles+urls",
|
336 |
+
snippet_max_chars: int = 160,
|
337 |
+
) -> str:
|
338 |
+
"""
|
339 |
+
(layman) A lightweight DDG search that returns a plain-text list.
|
340 |
+
- Fewer tokens than JSON; great for quick scanning or piping into LLM prompts.
|
341 |
+
- 'style' controls how much text we include per line.
|
342 |
+
"""
|
343 |
+
if not input_query or not input_query.strip():
|
344 |
+
return ""
|
345 |
+
|
346 |
+
# (layman) Run the search using the lightweight DDG client
|
347 |
+
with DDGS() as ddgs:
|
348 |
+
results = list(ddgs.text(input_query, max_results=max_results))
|
349 |
+
|
350 |
+
# (layman) Normalize fields because DDG library keys can vary by version
|
351 |
+
lines: List[str] = []
|
352 |
+
for r in results:
|
353 |
+
title = (r.get("title") or "").strip()
|
354 |
+
url = (r.get("href") or r.get("link") or r.get("url") or "").strip()
|
355 |
+
snippet = (r.get("body") or r.get("snippet") or "").strip()
|
356 |
+
|
357 |
+
# (layman) Truncate snippet to keep output tight
|
358 |
+
if snippet_max_chars and len(snippet) > snippet_max_chars:
|
359 |
+
snippet = snippet[:snippet_max_chars - 1].rstrip() + "β¦"
|
360 |
+
|
361 |
+
# (layman) Build each line according to the chosen style
|
362 |
+
if style == "urls":
|
363 |
+
if url:
|
364 |
+
lines.append(url)
|
365 |
+
elif style == "titles+urls":
|
366 |
+
if title and url:
|
367 |
+
lines.append(f"{title} β {url}")
|
368 |
+
elif url:
|
369 |
+
lines.append(url)
|
370 |
+
elif title:
|
371 |
+
lines.append(title)
|
372 |
+
else: # titles+urls+snippets
|
373 |
+
if title and url and snippet:
|
374 |
+
lines.append(f"{title} β {url}\n {snippet}")
|
375 |
+
elif title and url:
|
376 |
+
lines.append(f"{title} β {url}")
|
377 |
+
elif url:
|
378 |
+
# (layman) If only URL is available, still show it
|
379 |
+
if snippet:
|
380 |
+
lines.append(f"{url}\n {snippet}")
|
381 |
+
else:
|
382 |
+
lines.append(url)
|
383 |
+
elif title:
|
384 |
+
if snippet:
|
385 |
+
lines.append(f"{title}\n {snippet}")
|
386 |
+
else:
|
387 |
+
lines.append(title)
|
388 |
+
|
389 |
+
# (layman) Join lines with newlines to form a compact text block
|
390 |
+
return "\n".join(lines).strip()
|
391 |
+
|
392 |
+
|
393 |
# =====================
|
394 |
+
# UI: three-tab interface
|
395 |
# =====================
|
396 |
|
397 |
# --- Fetch tab (compact controllable extraction) ---
|
|
|
413 |
theme="Nymbo/Nymbo_Theme",
|
414 |
)
|
415 |
|
416 |
+
# --- Websearch tab (structured JSON) ---
|
417 |
websearch_interface = gr.Interface(
|
418 |
fn=web_search, # (layman) connect the function to the UI
|
419 |
inputs=[
|
|
|
421 |
gr.Slider(minimum=1, maximum=20, value=5, step=1, label="Max results"),
|
422 |
],
|
423 |
outputs=gr.JSON(label="Search results"),
|
424 |
+
title="Websearch β DuckDuckGo (JSON)",
|
425 |
+
description="Search the web using DuckDuckGo; returns snippet, title, and link as JSON.",
|
426 |
allow_flagging="never",
|
427 |
theme="Nymbo/Nymbo_Theme",
|
428 |
)
|
429 |
|
430 |
+
# --- DDG (Unstructured) tab (plain text, low tokens) ---
|
431 |
+
unstructured_interface = gr.Interface(
|
432 |
+
fn=web_search_unstructured, # (layman) connect the function to the UI
|
433 |
+
inputs=[
|
434 |
+
gr.Textbox(value="", label="Search query", placeholder="concise keywords"),
|
435 |
+
gr.Slider(minimum=1, maximum=20, value=5, step=1, label="Max results"),
|
436 |
+
gr.Dropdown(
|
437 |
+
label="Output style",
|
438 |
+
choices=["urls", "titles+urls", "titles+urls+snippets"],
|
439 |
+
value="titles+urls",
|
440 |
+
info="Plain text list; choose how much detail to include."
|
441 |
+
),
|
442 |
+
gr.Slider(
|
443 |
+
minimum=40, maximum=400, value=160, step=10,
|
444 |
+
label="Snippet max chars",
|
445 |
+
info="Truncate snippet length to keep token usage low."
|
446 |
+
),
|
447 |
+
],
|
448 |
+
outputs=gr.Textbox(label="Results (plain text)", interactive=False),
|
449 |
+
title="DDG β Unstructured (Compact)",
|
450 |
+
description="Outputs a plain-text list (great for low-token prompts).",
|
451 |
+
allow_flagging="never",
|
452 |
theme="Nymbo/Nymbo_Theme",
|
453 |
)
|
454 |
|
455 |
+
# --- Combine all three into a single app with tabs ---
|
456 |
+
demo = gr.TabbedInterface(
|
457 |
+
interface_list=[fetch_interface, websearch_interface, unstructured_interface],
|
458 |
+
tab_names=["Fetch", "Websearch", "DDG (Unstructured)"],
|
459 |
+
)
|
460 |
+
|
461 |
+
# Launch the UI and expose all functions as MCP tools in one server
|
462 |
if __name__ == "__main__":
|
463 |
demo.launch(mcp_server=True)
|