diff --git a/open_deep_researcher/researcher/researcher.py b/open_deep_researcher/researcher/researcher.py
index 4d4dac5..9b630c9 100644
--- a/open_deep_researcher/researcher/researcher.py
+++ b/open_deep_researcher/researcher/researcher.py
@@ -6,40 +6,53 @@
from typing import List, Dict, Optional, Tuple
import google.generativeai as genai
import time
+import os
# ---------------------------
# Configuration Constants
# ---------------------------
-GEMINI_API_KEY = "ADD_YOUR_GEMINI_API_KEY"
-SERPAPI_API_KEY = "ADD_YOUR_SERPAPI_API_KEY"
-JINA_API_KEY = "ADD_YOUR_JINA_API_KEY"
+
+
+GEMINI_API_KEY = os.environ.get("GEMINI_API_KEY")
+SERPAPI_API_KEY = os.environ.get("SERPAPI_API_KEY")
+JINA_API_KEY = os.environ.get("JINA_API_KEY")
SERPAPI_URL = "https://serpapi.com/search"
JINA_BASE_URL = "https://r.jina.ai/"
genai.configure(api_key=GEMINI_API_KEY)
+
async def call_google_gemini(messages: List[Dict]) -> Optional[str]:
"""Call Google Gemini asynchronously."""
try:
prompt = "\n".join([msg["content"] for msg in messages])
- model = genai.GenerativeModel('gemini-1.5-flash')
+ model = genai.GenerativeModel("gemini-1.5-flash")
loop = asyncio.get_event_loop()
- response = await loop.run_in_executor(None, lambda: model.generate_content(prompt))
+ response = await loop.run_in_executor(
+ None, lambda: model.generate_content(prompt)
+ )
return response.text
except Exception as e:
return None
-async def generate_search_queries_async(session: aiohttp.ClientSession, user_query: str) -> List[str]:
+
+async def generate_search_queries_async(
+ session: aiohttp.ClientSession, user_query: str
+) -> List[str]:
"""Generate search queries based on user query."""
prompt = (
"You are an expert research assistant. Given the user's query, generate up to four distinct, "
"precise search queries that would help gather complete information on the topic. "
"Return only a valid list of plain strings. Do not include markdown, code blocks, backticks, or explanations. "
- "Just return the list itself, for example: ['query1', 'query2', 'query3'].")
+ "Just return the list itself, for example: ['query1', 'query2', 'query3']."
+ )
messages = [
- {"role": "system", "content": "You are a helpful and precise research assistant."},
- {"role": "user", "content": f"User Query: {user_query}\n\n{prompt}"}
+ {
+ "role": "system",
+ "content": "You are a helpful and precise research assistant.",
+ },
+ {"role": "user", "content": f"User Query: {user_query}\n\n{prompt}"},
]
response = await call_google_gemini(messages)
if response:
@@ -50,22 +63,24 @@ async def generate_search_queries_async(session: aiohttp.ClientSession, user_que
return []
return []
+
async def perform_search_async(session: aiohttp.ClientSession, query: str) -> List[str]:
"""Perform search using SERPAPI."""
- params = {
- "q": query,
- "api_key": SERPAPI_API_KEY,
- "engine": "google"
- }
+ params = {"q": query, "api_key": SERPAPI_API_KEY, "engine": "google"}
try:
async with session.get(SERPAPI_URL, params=params) as resp:
if resp.status == 200:
results = await resp.json()
- return [item.get("link") for item in results.get("organic_results", []) if "link" in item]
+ return [
+ item.get("link")
+ for item in results.get("organic_results", [])
+ if "link" in item
+ ]
return []
except Exception:
return []
+
async def fetch_webpage_text_async(session: aiohttp.ClientSession, url: str) -> str:
"""Fetch webpage text using Jina API."""
full_url = f"{JINA_BASE_URL}{url}"
@@ -76,7 +91,10 @@ async def fetch_webpage_text_async(session: aiohttp.ClientSession, url: str) ->
except Exception:
return ""
-async def is_page_useful_async(session: aiohttp.ClientSession, user_query: str, page_text: str) -> bool:
+
+async def is_page_useful_async(
+ session: aiohttp.ClientSession, user_query: str, page_text: str
+) -> bool:
"""Determine if the page content is useful for the query."""
prompt = (
"You are a critical research evaluator. Given the user's query and the content of a webpage, "
@@ -84,17 +102,21 @@ async def is_page_useful_async(session: aiohttp.ClientSession, user_query: str,
"Respond with exactly one word: 'Yes' if the page is useful, or 'No' if it is not."
)
messages = [
- {"role": "system", "content": "You are a strict and concise evaluator of research relevance."},
- {"role": "user", "content": f"User Query: {user_query}\n\nWebpage Content:\n{page_text[:20000]}\n\n{prompt}"}
+ {
+ "role": "system",
+ "content": "You are a strict and concise evaluator of research relevance.",
+ },
+ {
+ "role": "user",
+ "content": f"User Query: {user_query}\n\nWebpage Content:\n{page_text[:20000]}\n\n{prompt}",
+ },
]
response = await call_google_gemini(messages)
return response and response.strip().lower() == "yes"
+
async def extract_relevant_context_async(
- session: aiohttp.ClientSession,
- user_query: str,
- search_query: str,
- page_text: str
+ session: aiohttp.ClientSession, user_query: str, search_query: str, page_text: str
) -> str:
"""Extract relevant information from page content."""
prompt = (
@@ -102,17 +124,24 @@ async def extract_relevant_context_async(
"Return only the relevant context as plain text."
)
messages = [
- {"role": "system", "content": "You are an expert in extracting relevant information."},
- {"role": "user", "content": f"Query: {user_query}\nSearch Query: {search_query}\n\nContent:\n{page_text[:20000]}\n\n{prompt}"}
+ {
+ "role": "system",
+ "content": "You are an expert in extracting relevant information.",
+ },
+ {
+ "role": "user",
+ "content": f"Query: {user_query}\nSearch Query: {search_query}\n\nContent:\n{page_text[:20000]}\n\n{prompt}",
+ },
]
response = await call_google_gemini(messages)
return response.strip() if response else ""
+
async def get_new_search_queries_async(
session: aiohttp.ClientSession,
user_query: str,
previous_queries: List[str],
- contexts: List[str]
+ contexts: List[str],
) -> List[str]:
"""Generate new search queries based on current findings."""
prompt = (
@@ -121,7 +150,10 @@ async def get_new_search_queries_async(
)
messages = [
{"role": "system", "content": "You are a systematic research planner."},
- {"role": "user", "content": f"Query: {user_query}\nPrevious: {previous_queries}\nContexts:\n{''.join(contexts)}\n\n{prompt}"}
+ {
+ "role": "user",
+ "content": f"Query: {user_query}\nPrevious: {previous_queries}\nContexts:\n{''.join(contexts)}\n\n{prompt}",
+ },
]
response = await call_google_gemini(messages)
if response:
@@ -132,10 +164,9 @@ async def get_new_search_queries_async(
return []
return []
+
async def generate_final_report_async(
- session: aiohttp.ClientSession,
- user_query: str,
- contexts: List[str]
+ session: aiohttp.ClientSession, user_query: str, contexts: List[str]
) -> str:
"""Generate final research report."""
prompt = (
@@ -144,18 +175,36 @@ async def generate_final_report_async(
)
messages = [
{"role": "system", "content": "You are a skilled report writer."},
- {"role": "user", "content": f"Query: {user_query}\nContexts:\n{''.join(contexts)}\n\n{prompt}"}
+ {
+ "role": "user",
+ "content": f"Query: {user_query}\nContexts:\n{''.join(contexts)}\n\n{prompt}",
+ },
]
response = await call_google_gemini(messages)
return response if response else "Unable to generate report."
+
class ResearchState(rx.State):
"""State management for the research assistant."""
+
user_query: str = ""
iteration_limit: int = 2
final_report: str = ""
process_logs: list
is_processing: bool = False
+ copy_success: bool = False
+
+ def clear_form(self):
+ """Clear the form and reset state."""
+ self.user_query = ""
+ self.final_report = ""
+ self.process_logs = ""
+ self.is_processing = False
+ self.copy_success = False
+
+ def new_research(self):
+ """Start a new research session."""
+ self.clear_form()
def update_logs(self, message: str):
"""Update process logs with timestamp."""
@@ -165,15 +214,18 @@ def update_logs(self, message: str):
else:
self.process_logs = f"[{timestamp}] {message}"
-
- async def process_link(self, session: aiohttp.ClientSession, link: str, search_query: str) -> Optional[str]:
+ async def process_link(
+ self, session: aiohttp.ClientSession, link: str, search_query: str
+ ) -> Optional[str]:
"""Process a single link and extract relevant information."""
page_text = await fetch_webpage_text_async(session, link)
if not page_text:
return None
if await is_page_useful_async(session, self.user_query, page_text):
- context = await extract_relevant_context_async(session, self.user_query, search_query, page_text)
+ context = await extract_relevant_context_async(
+ session, self.user_query, search_query, page_text
+ )
return context
return None
@@ -190,19 +242,21 @@ async def handle_submit(self):
async with aiohttp.ClientSession() as session:
self.update_logs("Generating initial search queries...")
yield
-
+
queries = await generate_search_queries_async(session, self.user_query)
if not queries:
self.update_logs("No initial queries could be generated")
yield
return
- self.update_logs(f"Generated {len(queries)} initial queries: {', '.join(queries)}")
+ self.update_logs(
+ f"Generated {len(queries)} initial queries: {', '.join(queries)}"
+ )
yield
contexts = []
iteration = 0
-
+
while iteration < self.iteration_limit:
self.update_logs(f"Starting research iteration {iteration + 1}")
yield
@@ -220,7 +274,7 @@ async def handle_submit(self):
if len(all_links) >= 10:
break
all_links.extend(links)
-
+
self.update_logs(f"Found {len(all_links)} links to process")
yield
@@ -228,34 +282,44 @@ async def handle_submit(self):
for link in all_links:
self.update_logs(f"Processing link: {link}")
yield # Update UI after log entry
-
+
context = await self.process_link(session, link, query)
if context:
- self.update_logs("Successfully extracted relevant information")
+ self.update_logs(
+ "Successfully extracted relevant information"
+ )
iteration_contexts.append(context)
yield # Update UI after successful extraction
else:
self.update_logs("No useful information found in link")
yield
-
- self.update_logs(f"Extracted information from {len(iteration_contexts)} sources")
+
+ self.update_logs(
+ f"Extracted information from {len(iteration_contexts)} sources"
+ )
yield
-
+
contexts.extend(iteration_contexts)
- queries = await get_new_search_queries_async(session, self.user_query, queries, contexts)
-
+ queries = await get_new_search_queries_async(
+ session, self.user_query, queries, contexts
+ )
+
if not queries:
self.update_logs("No more queries needed, research complete")
yield
break
-
- self.update_logs(f"Generated {len(queries)} new queries for next iteration")
+
+ self.update_logs(
+ f"Generated {len(queries)} new queries for next iteration"
+ )
yield
iteration += 1
self.update_logs("Generating final research report...")
yield
- self.final_report = await generate_final_report_async(session, self.user_query, contexts)
+ self.final_report = await generate_final_report_async(
+ session, self.user_query, contexts
+ )
self.update_logs("Research process completed successfully")
except Exception as e:
@@ -263,80 +327,273 @@ async def handle_submit(self):
finally:
self.is_processing = False
yield
-
+
+ async def copy_to_clipboard(self):
+ """Copy report to clipboard and show success state."""
+ # Set clipboard content using JavaScript
+ yield rx.set_clipboard(self.final_report)
+
+ # Show success state
+ self.copy_success = True
+ yield
+ await asyncio.sleep(2)
+ self.copy_success = False
+ yield
+
+ def export_to_pdf(self):
+ """Trigger PDF export using browser's print functionality."""
+ return rx.call_script(
+ """
+ const content = document.querySelector('#research-report').innerHTML;
+ const printWindow = window.open('', '_blank');
+ printWindow.document.write(`
+
+
+ Research Report
+
+
+ ${content}
+
+ `);
+ printWindow.document.close();
+ printWindow.print();
+ """
+ )
+
+
def index() -> rx.Component:
return rx.container(
rx.vstack(
- rx.heading("Open Deep Researcher 🔬", size="8", margin_bottom="1rem", margin_left="16rem"),
- rx.text("Enter your research query to generate a report."),
-
+ # Header Section
+ rx.hstack(
+ rx.icon("microscope", color="blue.500", size=24),
+ rx.heading("Open Deep Researcher", size="3"),
+ spacing="3",
+ align="center",
+ margin_bottom="2rem",
+ ),
+ rx.text(
+ "Enter your research query to generate a comprehensive report.",
+ color="gray.600",
+ font_size="lg",
+ margin_bottom="2rem",
+ ),
# Input Section
- rx.box(
+ rx.card(
rx.vstack(
- rx.input(
- placeholder="Research Query/Topic",
+ rx.text_area(
+ placeholder="Enter your research query here...",
value=ResearchState.user_query,
on_change=ResearchState.set_user_query,
- width="100%",
+ min_height="200px",
+ size="2",
+ border="1px solid",
+ border_color="gray.200",
+ _focus={
+ "border_color": "blue.500",
+ "box_shadow": "0 0 0 1px rgb(49, 130, 206)",
+ },
),
- rx.button(
- "Start Research",
- on_click=ResearchState.handle_submit,
- loading=ResearchState.is_processing,
- color_scheme="blue",
- width="40%",
- align_self="center",
+ rx.hstack(
+ rx.spacer(), # This pushes the buttons to the right
+ rx.button(
+ rx.hstack(
+ rx.icon("search", size=16),
+ rx.text("Start Research"),
+ spacing="2",
+ ),
+ on_click=ResearchState.handle_submit,
+ loading=ResearchState.is_processing,
+ color_scheme="blue",
+ size="2",
+ min_width="150px",
+ padding_x="2rem",
+ ),
+ rx.button(
+ rx.hstack(
+ rx.icon("rotate-ccw", size=16),
+ rx.text("New Research"),
+ spacing="2",
+ ),
+ on_click=ResearchState.new_research,
+ color_scheme="green",
+ variant="outline",
+ size="2",
+ min_width="150px",
+ padding_x="2rem",
+ ),
+ rx.button(
+ rx.hstack(
+ rx.icon("trash-2", size=16),
+ rx.text("Clear"),
+ spacing="2",
+ ),
+ on_click=ResearchState.clear_form,
+ color_scheme="red",
+ variant="ghost",
+ size="2",
+ min_width="100px",
+ padding_x="1.5rem",
+ margin_top="10px",
+ ),
+ spacing="4",
+ width="100%",
+ padding_top="1rem",
),
- spacing="3",
+ spacing="4",
+ padding="1.5rem",
+ align_items="stretch",
),
width="100%",
- padding="1rem",
- border="1px solid #e0e0e0",
- border_radius="lg",
+ border="1px solid",
+ border_color="gray.100",
),
-
# Results Section
rx.cond(
ResearchState.final_report,
- rx.box(
+ rx.card(
rx.vstack(
- rx.heading("Final Report", size="4"),
- rx.markdown(ResearchState.final_report),
- spacing="2",
+ # Report Header with Actions
+ rx.hstack(
+ rx.hstack(
+ rx.icon("file-text", color="blue.500", size=20),
+ rx.heading("Research Report", size="4"),
+ spacing="2",
+ ),
+ rx.spacer(),
+ rx.hstack(
+ # Copy Button
+ rx.button(
+ rx.cond(
+ ResearchState.copy_success,
+ rx.hstack(
+ rx.icon("check", size=14),
+ rx.text("Copied!"),
+ spacing="2",
+ ),
+ rx.hstack(
+ rx.icon("copy", size=14),
+ rx.text("Copy"),
+ spacing="2",
+ ),
+ ),
+ on_click=ResearchState.copy_to_clipboard,
+ color_scheme="gray",
+ variant="ghost",
+ size="2",
+ ),
+ # Export PDF Button
+ rx.button(
+ rx.hstack(
+ rx.icon("file-down", size=14),
+ rx.text("Export PDF"),
+ spacing="2",
+ ),
+ on_click=ResearchState.export_to_pdf,
+ color_scheme="gray",
+ variant="ghost",
+ size="2",
+ ),
+ spacing="2",
+ ),
+ width="100%",
+ padding_bottom="4",
+ border_bottom="1px solid",
+ border_color="gray.200",
+ margin_bottom="4",
+ ),
+ # Report Content
+ rx.box(
+ rx.markdown(
+ ResearchState.final_report,
+ component_map={
+ "p": lambda text: rx.text(text, color="gray.800"),
+ "h1": lambda text: rx.heading(
+ text, size="3", margin_y="1rem"
+ ),
+ "h2": lambda text: rx.heading(
+ text, size="4", margin_y="0.75rem"
+ ),
+ "h3": lambda text: rx.heading(
+ text, size="5", margin_y="0.5rem"
+ ),
+ },
+ ),
+ id="research-report", # Added ID for PDF export
+ width="100%",
+ ),
+ width="100%",
+ align_items="stretch",
),
width="100%",
- padding="1rem",
- border="1px solid #e0e0e0",
- border_radius="lg",
- margin_top="1rem",
+ padding="1.5rem",
),
),
-
# Logs Section
rx.cond(
ResearchState.process_logs,
- rx.box(
+ rx.card(
rx.vstack(
- rx.heading("Process Logs", size="4"),
- rx.markdown(ResearchState.process_logs),
- spacing="2",
+ rx.hstack(
+ rx.icon("activity", color="blue.500", size=20),
+ rx.heading("Research Progress", size="4"),
+ margin_bottom="1rem",
+ ),
+ rx.box(
+ rx.markdown(
+ ResearchState.process_logs,
+ component_map={
+ "p": lambda text: rx.text(
+ text, color="gray.600", font_family="monospace"
+ ),
+ },
+ ),
+ width="100%",
+ height="400px",
+ overflow_y="auto",
+ padding="1rem",
+ border="1px solid",
+ border_color="gray.200",
+ border_radius="md",
+ background="gray.50",
+ ),
+ width="100%",
+ align_items="stretch",
),
width="100%",
- height="400px",
- padding="1rem",
- border="1px solid #e0e0e0",
- border_radius="lg",
- margin_top="1rem",
- overflow_y="auto",
+ padding="1.5rem",
),
),
- spacing="4",
+ spacing="6",
width="100%",
max_width="1200px",
+ padding_y="2rem",
),
- padding="2rem",
+ padding_x="2rem",
+ background="gray.50",
+ min_height="100vh",
)
+
# Create app
-app = rx.App()
-app.add_page(index, title="Research Assistant")
\ No newline at end of file
+app = rx.App(
+ style={
+ "font_family": "Inter, system-ui, sans-serif",
+ "button": {
+ "shadow": "md",
+ "_hover": {
+ "transform": "translateY(-2px)",
+ "transition": "0.2s",
+ },
+ },
+ "card": {
+ "shadow": "lg",
+ "border_radius": "lg",
+ "background": "white",
+ },
+ }
+)
+app.add_page(index, title="Research Assistant")