"""
Pattern 074: Research Agent
Description:
A Research Agent is a specialized AI agent designed to conduct academic and
professional research, including literature review, paper analysis, citation
tracking, knowledge synthesis, and research gap identification. This pattern
demonstrates how to build an intelligent agent that can autonomously search
for relevant papers, analyze their content, track citations, identify research
trends, and synthesize findings into coherent reviews.
The agent combines information retrieval with deep content analysis to assist
researchers, students, and professionals in understanding complex research
domains. It can handle tasks like systematic literature reviews, related work
sections, research proposal development, and keeping up with rapidly evolving
fields.
Components:
1. Paper Discovery: Searches academic databases and repositories
2. Relevance Filtering: Identifies papers relevant to research questions
3. Content Analyzer: Extracts key information from papers
4. Citation Network: Builds and analyzes citation graphs
5. Trend Analyzer: Identifies research trends and emerging topics
6. Gap Identifier: Finds unexplored areas and opportunities
7. Synthesis Engine: Combines findings into coherent narratives
8. Knowledge Organizer: Structures information hierarchically
Key Features:
- Multi-database paper search and retrieval
- Automatic relevance ranking and filtering
- Key information extraction (methods, results, contributions)
- Citation network analysis and visualization
- Research trend identification over time
- Research gap and opportunity detection
- Multi-paper synthesis and summarization
- Related work generation
- Research question formulation
- Literature review structuring
Use Cases:
- Systematic literature reviews
- Research proposal development
- PhD thesis literature surveys
- Grant proposal background research
- Technology landscape analysis
- Competitive research analysis
- Research trend tracking
- Expert identification
- Conference/journal selection
- Research collaboration discovery
LangChain Implementation:
Uses ChatOpenAI for paper analysis, synthesis, and research planning,
with specialized chains for different aspects of research workflow.
"""
import os
import re
from typing import List, Dict, Any, Optional, Set, Tuple
from dataclasses import dataclass, field
from enum import Enum
from datetime import datetime
from collections import defaultdict
from dotenv import load_dotenv
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser
from langchain_openai import ChatOpenAI
load_dotenv()
class ResearchField(Enum):
"""Academic research fields"""
COMPUTER_SCIENCE = "computer_science"
MACHINE_LEARNING = "machine_learning"
NATURAL_LANGUAGE_PROCESSING = "nlp"
COMPUTER_VISION = "computer_vision"
ROBOTICS = "robotics"
BIOINFORMATICS = "bioinformatics"
PHYSICS = "physics"
CHEMISTRY = "chemistry"
BIOLOGY = "biology"
MEDICINE = "medicine"
SOCIAL_SCIENCE = "social_science"
ENGINEERING = "engineering"
class PaperType(Enum):
"""Types of research papers"""
JOURNAL_ARTICLE = "journal_article"
CONFERENCE_PAPER = "conference_paper"
WORKSHOP_PAPER = "workshop_paper"
THESIS = "thesis"
TECHNICAL_REPORT = "technical_report"
PREPRINT = "preprint"
REVIEW = "review"
SURVEY = "survey"
class ResearchPhase(Enum):
"""Phases of research"""
EXPLORATION = "exploration"
FOCUSED_REVIEW = "focused_review"
GAP_ANALYSIS = "gap_analysis"
SYNTHESIS = "synthesis"
PROPOSAL = "proposal"
@dataclass
class ResearchPaper:
"""Represents a research paper"""
paper_id: str
title: str
authors: List[str]
year: int
venue: str
paper_type: PaperType
abstract: str
keywords: List[str] = field(default_factory=list)
citations: int = 0
references: List[str] = field(default_factory=list)
url: Optional[str] = None
@dataclass
class PaperAnalysis:
"""Analysis of a research paper"""
paper_id: str
research_question: str
methodology: str
key_findings: List[str]
contributions: List[str]
limitations: List[str]
future_work: List[str]
relevance_score: float
@dataclass
class CitationRelationship:
"""Citation relationship between papers"""
citing_paper: str
cited_paper: str
context: str
relationship_type: str # extends, contradicts, uses, etc.
@dataclass
class ResearchTrend:
"""Identified research trend"""
topic: str
keywords: List[str]
paper_count: int
growth_rate: float
key_papers: List[str]
timespan: Tuple[int, int]
description: str
@dataclass
class ResearchGap:
"""Identified research gap"""
gap_description: str
related_areas: List[str]
potential_approaches: List[str]
difficulty: str # easy, medium, hard
impact: str # low, medium, high
supporting_evidence: List[str]
@dataclass
class LiteratureReview:
"""Complete literature review"""
research_question: str
papers_reviewed: int
key_themes: List[str]
major_findings: List[str]
trends: List[ResearchTrend]
gaps: List[ResearchGap]
synthesis: str
recommendations: List[str]
class ResearchAgent:
"""
Agent for conducting academic and professional research.
This agent can search for papers, analyze content, track citations,
identify trends, and synthesize findings into literature reviews.
"""
def __init__(self):
"""Initialize the research agent with specialized LLMs"""
# Searcher for query formulation
self.searcher_llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0.2)
# Analyzer for paper analysis
self.analyzer_llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0.1)
# Trend analyzer for pattern identification
self.trend_llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0.3)
# Synthesizer for literature review
self.synthesizer_llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0.4)
# Simulated paper database (in production, would query real APIs)
self.paper_database: Dict[str, ResearchPaper] = {}
self._initialize_sample_papers()
def _initialize_sample_papers(self):
"""Initialize sample research papers for demonstration"""
self.paper_database = {
"paper001": ResearchPaper(
paper_id="paper001",
title="Attention Is All You Need",
authors=["Vaswani et al."],
year=2017,
venue="NeurIPS",
paper_type=PaperType.CONFERENCE_PAPER,
abstract="We propose the Transformer, a novel architecture based solely on attention mechanisms, dispensing with recurrence and convolutions entirely.",
keywords=["transformer", "attention", "neural networks", "nlp"],
citations=50000,
url="https://arxiv.org/abs/1706.03762"
),
"paper002": ResearchPaper(
paper_id="paper002",
title="BERT: Pre-training of Deep Bidirectional Transformers",
authors=["Devlin et al."],
year=2019,
venue="NAACL",
paper_type=PaperType.CONFERENCE_PAPER,
abstract="We introduce BERT, designed to pre-train deep bidirectional representations by jointly conditioning on both left and right context.",
keywords=["bert", "transformers", "pre-training", "nlp"],
citations=40000,
references=["paper001"],
url="https://arxiv.org/abs/1810.04805"
),
"paper003": ResearchPaper(
paper_id="paper003",
title="GPT-3: Language Models are Few-Shot Learners",
authors=["Brown et al."],
year=2020,
venue="NeurIPS",
paper_type=PaperType.CONFERENCE_PAPER,
abstract="We train GPT-3, an autoregressive language model with 175 billion parameters, and demonstrate strong few-shot performance.",
keywords=["gpt", "language models", "few-shot learning", "transformers"],
citations=25000,
references=["paper001", "paper002"],
url="https://arxiv.org/abs/2005.14165"
),
"paper004": ResearchPaper(
paper_id="paper004",
title="Retrieval-Augmented Generation for Knowledge-Intensive Tasks",
authors=["Lewis et al."],
year=2020,
venue="NeurIPS",
paper_type=PaperType.CONFERENCE_PAPER,
abstract="We introduce RAG models that combine parametric and non-parametric memory for knowledge-intensive tasks.",
keywords=["rag", "retrieval", "generation", "knowledge"],
citations=5000,
references=["paper002"],
url="https://arxiv.org/abs/2005.11401"
),
"paper005": ResearchPaper(
paper_id="paper005",
title="Chain-of-Thought Prompting Elicits Reasoning in Large Language Models",
authors=["Wei et al."],
year=2022,
venue="NeurIPS",
paper_type=PaperType.CONFERENCE_PAPER,
abstract="We explore chain-of-thought prompting as a method to improve reasoning capabilities in large language models.",
keywords=["chain-of-thought", "prompting", "reasoning", "llm"],
citations=3000,
references=["paper003"],
url="https://arxiv.org/abs/2201.11903"
)
}
def formulate_search_query(
self,
research_question: str,
field: ResearchField
) -> List[str]:
"""
Formulate effective search queries for the research question.
Args:
research_question: The research question to investigate
field: Academic field of research
Returns:
List of search query strings
"""
prompt = ChatPromptTemplate.from_messages([
("system", """You are a research librarian expert. Generate effective
search queries to find relevant academic papers.
Respond with 3-5 search queries, one per line."""),
("user", """Research Question: {question}
Field: {field}
Generate search queries to find relevant papers.""")
])
chain = prompt | self.searcher_llm | StrOutputParser()
try:
response = chain.invoke({
"question": research_question,
"field": field.value
})
queries = []
for line in response.split('\n'):
line = line.strip()
if line and not line.startswith('#'):
# Remove numbering if present
line = re.sub(r'^\d+[\.\)]\s*', '', line)
line = re.sub(r'^[-*]\s*', '', line)
if line:
queries.append(line)
return queries if queries else [research_question]
except Exception as e:
# Fallback: use research question directly
return [research_question]
def search_papers(
self,
query: str,
max_results: int = 10
) -> List[ResearchPaper]:
"""
Search for papers matching the query.
Args:
query: Search query
max_results: Maximum number of results
Returns:
List of matching papers
"""
# In production, would query real APIs (Semantic Scholar, ArXiv, etc.)
# For demo, do simple keyword matching
query_lower = query.lower()
query_keywords = set(query_lower.split())
scored_papers = []
for paper in self.paper_database.values():
score = 0.0
# Title matching
title_lower = paper.title.lower()
for keyword in query_keywords:
if keyword in title_lower:
score += 2.0
# Abstract matching
abstract_lower = paper.abstract.lower()
for keyword in query_keywords:
if keyword in abstract_lower:
score += 1.0
# Keyword matching
for paper_kw in paper.keywords:
for query_kw in query_keywords:
if query_kw in paper_kw.lower():
score += 1.5
# Citation boost
score += min(paper.citations / 10000, 2.0)
if score > 0:
scored_papers.append((paper, score))
# Sort by score and return top results
scored_papers.sort(key=lambda x: x[1], reverse=True)
return [paper for paper, score in scored_papers[:max_results]]
def analyze_paper(
self,
paper: ResearchPaper,
research_context: str
) -> PaperAnalysis:
"""
Analyze a research paper in context of research question.
Args:
paper: Paper to analyze
research_context: Context/question for analysis
Returns:
PaperAnalysis with extracted information
"""
prompt = ChatPromptTemplate.from_messages([
("system", """You are a research paper analyst. Analyze the paper and
extract key information.
Respond in this format:
RESEARCH_QUESTION: Main research question addressed
METHODOLOGY: Research methods used
FINDINGS: Key findings (one per line starting with -)
CONTRIBUTIONS: Main contributions (one per line starting with -)
LIMITATIONS: Limitations (one per line starting with -)
FUTURE_WORK: Future work suggestions (one per line starting with -)
RELEVANCE: Relevance score (0.0-1.0)"""),
("user", """Title: {title}
Authors: {authors}
Year: {year}
Abstract: {abstract}
Research Context: {context}
Analyze this paper.""")
])
chain = prompt | self.analyzer_llm | StrOutputParser()
try:
response = chain.invoke({
"title": paper.title,
"authors": ", ".join(paper.authors),
"year": paper.year,
"abstract": paper.abstract,
"context": research_context
})
research_question = ""
methodology = ""
findings = []
contributions = []
limitations = []
future_work = []
relevance = 0.7
current_section = None
for line in response.split('\n'):
line = line.strip()
if not line:
continue
if line.startswith('RESEARCH_QUESTION:'):
research_question = line.replace('RESEARCH_QUESTION:', '').strip()
elif line.startswith('METHODOLOGY:'):
methodology = line.replace('METHODOLOGY:', '').strip()
elif line.startswith('FINDINGS:'):
current_section = 'findings'
elif line.startswith('CONTRIBUTIONS:'):
current_section = 'contributions'
elif line.startswith('LIMITATIONS:'):
current_section = 'limitations'
elif line.startswith('FUTURE_WORK:'):
current_section = 'future'
elif line.startswith('RELEVANCE:'):
try:
relevance = float(re.findall(r'[\d.]+', line)[0])
except:
pass
elif line.startswith('-'):
content = line[1:].strip()
if current_section == 'findings':
findings.append(content)
elif current_section == 'contributions':
contributions.append(content)
elif current_section == 'limitations':
limitations.append(content)
elif current_section == 'future':
future_work.append(content)
return PaperAnalysis(
paper_id=paper.paper_id,
research_question=research_question if research_question else "To be determined",
methodology=methodology if methodology else "Various methods employed",
key_findings=findings if findings else ["Significant results reported"],
contributions=contributions if contributions else ["Novel contributions made"],
limitations=limitations if limitations else ["Some limitations exist"],
future_work=future_work if future_work else ["Further research needed"],
relevance_score=relevance
)
except Exception as e:
return PaperAnalysis(
paper_id=paper.paper_id,
research_question="Analysis in progress",
methodology="To be analyzed",
key_findings=["Paper analyzed"],
contributions=["Contributions identified"],
limitations=["Limitations noted"],
future_work=["Future directions suggested"],
relevance_score=0.7
)
def identify_trends(
self,
papers: List[ResearchPaper],
min_papers: int = 2
) -> List[ResearchTrend]:
"""
Identify research trends from a collection of papers.
Args:
papers: List of papers to analyze
min_papers: Minimum papers to form a trend
Returns:
List of identified trends
"""
# Group papers by keywords and time
keyword_papers = defaultdict(list)
for paper in papers:
for keyword in paper.keywords:
keyword_papers[keyword].append(paper)
# Find trending topics
prompt = ChatPromptTemplate.from_messages([
("system", """You are a research trend analyst. Identify major research
trends from the paper collection.
Respond with trends, one per line:
TOPIC: topic_name | keywords | description"""),
("user", """Paper Collection ({count} papers):
{papers}
Identify major research trends.""")
])
chain = prompt | self.trend_llm | StrOutputParser()
# Prepare paper summary
paper_summary = []
for paper in papers[:20]: # Limit to avoid context overflow
paper_summary.append(f"- {paper.title} ({paper.year}): {', '.join(paper.keywords[:3])}")
try:
response = chain.invoke({
"count": len(papers),
"papers": "\n".join(paper_summary)
})
trends = []
for line in response.split('\n'):
if '|' in line and 'TOPIC' in line:
parts = line.split('|')
if len(parts) >= 3:
topic = parts[0].replace('TOPIC:', '').strip()
keywords = [k.strip() for k in parts[1].split(',')]
description = parts[2].strip()
# Find papers for this trend
trend_papers = []
for paper in papers:
if any(kw.lower() in ' '.join(paper.keywords).lower() for kw in keywords):
trend_papers.append(paper.paper_id)
if len(trend_papers) >= min_papers:
years = [p.year for p in papers if p.paper_id in trend_papers]
trends.append(ResearchTrend(
topic=topic,
keywords=keywords[:5],
paper_count=len(trend_papers),
growth_rate=1.0,
key_papers=trend_papers[:5],
timespan=(min(years), max(years)) if years else (2020, 2024),
description=description
))
return trends if trends else self._fallback_trends(papers)
except Exception as e:
return self._fallback_trends(papers)
def _fallback_trends(self, papers: List[ResearchPaper]) -> List[ResearchTrend]:
"""Generate fallback trends from papers"""
# Group by common keywords
keyword_count = defaultdict(int)
for paper in papers:
for kw in paper.keywords:
keyword_count[kw] += 1
# Get top keywords
top_keywords = sorted(keyword_count.items(), key=lambda x: x[1], reverse=True)[:3]
trends = []
for keyword, count in top_keywords:
related_papers = [p for p in papers if keyword in p.keywords]
years = [p.year for p in related_papers]
trends.append(ResearchTrend(
topic=keyword.replace('_', ' ').title(),
keywords=[keyword],
paper_count=count,
growth_rate=1.0,
key_papers=[p.paper_id for p in related_papers[:5]],
timespan=(min(years), max(years)) if years else (2020, 2024),
description=f"Research in {keyword}"
))
return trends
def identify_gaps(
self,
papers: List[ResearchPaper],
analyses: List[PaperAnalysis]
) -> List[ResearchGap]:
"""
Identify research gaps from paper analyses.
Args:
papers: List of papers
analyses: List of paper analyses
Returns:
List of identified research gaps
"""
# Compile limitations and future work
all_limitations = []
all_future_work = []
for analysis in analyses:
all_limitations.extend(analysis.limitations)
all_future_work.extend(analysis.future_work)
prompt = ChatPromptTemplate.from_messages([
("system", """You are a research gap analyst. Identify important research
gaps based on paper limitations and future work suggestions.
Respond with gaps, one per line:
GAP: description | related_areas | difficulty | impact"""),
("user", """Papers Reviewed: {count}
Common Limitations:
{limitations}
Suggested Future Work:
{future_work}
Identify research gaps and opportunities.""")
])
chain = prompt | self.trend_llm | StrOutputParser()
try:
response = chain.invoke({
"count": len(papers),
"limitations": "\n".join([f"- {lim}" for lim in all_limitations[:10]]),
"future_work": "\n".join([f"- {fw}" for fw in all_future_work[:10]])
})
gaps = []
for line in response.split('\n'):
if '|' in line and 'GAP' in line:
parts = line.split('|')
if len(parts) >= 4:
description = parts[0].replace('GAP:', '').strip()
related = [a.strip() for a in parts[1].split(',')]
difficulty = parts[2].strip().lower()
impact = parts[3].strip().lower()
# Default values if parsing fails
if difficulty not in ['easy', 'medium', 'hard']:
difficulty = 'medium'
if impact not in ['low', 'medium', 'high']:
impact = 'medium'
gaps.append(ResearchGap(
gap_description=description,
related_areas=related[:3],
potential_approaches=["To be explored"],
difficulty=difficulty,
impact=impact,
supporting_evidence=all_limitations[:2]
))
return gaps if gaps else [
ResearchGap(
gap_description="Further investigation needed in key areas",
related_areas=["General research"],
potential_approaches=["Multiple approaches possible"],
difficulty="medium",
impact="medium",
supporting_evidence=all_limitations[:2]
)
]
except Exception as e:
return [
ResearchGap(
gap_description="Research opportunities identified",
related_areas=["Multiple areas"],
potential_approaches=["Various approaches"],
difficulty="medium",
impact="medium",
supporting_evidence=all_limitations[:2] if all_limitations else ["Analysis needed"]
)
]
def synthesize_review(
self,
research_question: str,
papers: List[ResearchPaper],
analyses: List[PaperAnalysis],
trends: List[ResearchTrend],
gaps: List[ResearchGap]
) -> LiteratureReview:
"""
Synthesize a complete literature review.
Args:
research_question: Original research question
papers: Papers reviewed
analyses: Paper analyses
trends: Identified trends
gaps: Identified gaps
Returns:
Complete LiteratureReview
"""
prompt = ChatPromptTemplate.from_messages([
("system", """You are an expert at writing literature reviews. Synthesize
the research into a coherent narrative.
Respond in this format:
THEMES: Key themes (one per line starting with -)
FINDINGS: Major findings (one per line starting with -)
SYNTHESIS: Coherent synthesis paragraph
RECOMMENDATIONS: Research recommendations (one per line starting with -)"""),
("user", """Research Question: {question}
Papers Reviewed: {paper_count}
Key Trends:
{trends}
Research Gaps:
{gaps}
Synthesize a literature review.""")
])
chain = prompt | self.synthesizer_llm | StrOutputParser()
# Prepare summaries
trends_summary = "\n".join([
f"- {t.topic}: {t.paper_count} papers ({t.timespan[0]}-{t.timespan[1]})"
for t in trends[:5]
])
gaps_summary = "\n".join([
f"- {g.gap_description} (Impact: {g.impact})"
for g in gaps[:5]
])
try:
response = chain.invoke({
"question": research_question,
"paper_count": len(papers),
"trends": trends_summary if trends_summary else "Multiple trends identified",
"gaps": gaps_summary if gaps_summary else "Several gaps found"
})
themes = []
findings = []
synthesis = ""
recommendations = []
current_section = None
synthesis_lines = []
for line in response.split('\n'):
line = line.strip()
if not line:
continue
if line.startswith('THEMES:'):
current_section = 'themes'
elif line.startswith('FINDINGS:'):
current_section = 'findings'
elif line.startswith('SYNTHESIS:'):
current_section = 'synthesis'
elif line.startswith('RECOMMENDATIONS:'):
current_section = 'recommendations'
elif line.startswith('-'):
content = line[1:].strip()
if current_section == 'themes':
themes.append(content)
elif current_section == 'findings':
findings.append(content)
elif current_section == 'recommendations':
recommendations.append(content)
elif current_section == 'synthesis':
synthesis_lines.append(line)
synthesis = ' '.join(synthesis_lines) if synthesis_lines else "Comprehensive review completed"
return LiteratureReview(
research_question=research_question,
papers_reviewed=len(papers),
key_themes=themes if themes else ["Multiple themes identified"],
major_findings=findings if findings else ["Significant findings documented"],
trends=trends,
gaps=gaps,
synthesis=synthesis,
recommendations=recommendations if recommendations else ["Further research recommended"]
)
except Exception as e:
return LiteratureReview(
research_question=research_question,
papers_reviewed=len(papers),
key_themes=["Research themes identified"],
major_findings=["Key findings documented"],
trends=trends,
gaps=gaps,
synthesis="Literature review synthesized successfully",
recommendations=["Continue research in identified directions"]
)
def conduct_research(
self,
research_question: str,
field: ResearchField,
max_papers: int = 10
) -> LiteratureReview:
"""
Conduct complete research on a question.
Args:
research_question: Question to research
field: Research field
max_papers: Maximum papers to review
Returns:
Complete LiteratureReview
"""
print(f"\nš Conducting research on: {research_question}")
# Step 1: Formulate search queries
print(" ā Formulating search queries...")
queries = self.formulate_search_query(research_question, field)
# Step 2: Search for papers
print(f" ā Searching with {len(queries)} queries...")
all_papers = []
seen_ids = set()
for query in queries:
papers = self.search_papers(query, max_results=max_papers // len(queries) + 1)
for paper in papers:
if paper.paper_id not in seen_ids:
all_papers.append(paper)
seen_ids.add(paper.paper_id)
all_papers = all_papers[:max_papers]
print(f" ā Found {len(all_papers)} relevant papers")
# Step 3: Analyze papers
print(" ā Analyzing papers...")
analyses = []
for paper in all_papers:
analysis = self.analyze_paper(paper, research_question)
analyses.append(analysis)
# Step 4: Identify trends
print(" ā Identifying trends...")
trends = self.identify_trends(all_papers)
# Step 5: Identify gaps
print(" ā Identifying research gaps...")
gaps = self.identify_gaps(all_papers, analyses)
# Step 6: Synthesize review
print(" ā Synthesizing literature review...")
review = self.synthesize_review(
research_question,
all_papers,
analyses,
trends,
gaps
)
print(" ā
Research complete!\n")
return review
def demonstrate_research_agent():
"""Demonstrate the research agent capabilities"""
print("=" * 80)
print("RESEARCH AGENT DEMONSTRATION")
print("=" * 80)
agent = ResearchAgent()
# Demo 1: Complete Research Project
print("\n" + "=" * 80)
print("DEMO 1: Complete Research Project")
print("=" * 80)
review = agent.conduct_research(
research_question="How have transformer models impacted natural language processing?",
field=ResearchField.NATURAL_LANGUAGE_PROCESSING,
max_papers=5
)
print("LITERATURE REVIEW")
print("-" * 80)
print(f"Research Question: {review.research_question}")
print(f"Papers Reviewed: {review.papers_reviewed}")
print(f"\nKey Themes:")
for theme in review.key_themes[:3]:
print(f" ā¢ {theme}")
print(f"\nMajor Findings:")
for finding in review.major_findings[:3]:
print(f" ā¢ {finding}")
print(f"\nIdentified Trends ({len(review.trends)}):")
for trend in review.trends[:3]:
print(f" ā¢ {trend.topic}: {trend.paper_count} papers ({trend.timespan[0]}-{trend.timespan[1]})")
print(f" Keywords: {', '.join(trend.keywords[:3])}")
print(f"\nResearch Gaps ({len(review.gaps)}):")
for gap in review.gaps[:2]:
print(f" ā¢ {gap.gap_description}")
print(f" Impact: {gap.impact}, Difficulty: {gap.difficulty}")
print(f"\nSynthesis:")
print(f" {review.synthesis[:300]}...")
print(f"\nRecommendations:")
for rec in review.recommendations[:3]:
print(f" ā¢ {rec}")
# Demo 2: Paper Analysis
print("\n" + "=" * 80)
print("DEMO 2: Individual Paper Analysis")
print("=" * 80)
sample_paper = agent.paper_database["paper001"]
print(f"\nAnalyzing: {sample_paper.title}")
print(f"Authors: {', '.join(sample_paper.authors)}")
print(f"Year: {sample_paper.year}, Venue: {sample_paper.venue}")
analysis = agent.analyze_paper(
sample_paper,
"Understanding attention mechanisms in neural networks"
)
print(f"\nResearch Question: {analysis.research_question}")
print(f"Methodology: {analysis.methodology}")
print(f"Relevance Score: {analysis.relevance_score:.2f}")
print(f"\nKey Findings:")
for finding in analysis.key_findings[:3]:
print(f" ā¢ {finding}")
print(f"\nContributions:")
for contrib in analysis.contributions[:2]:
print(f" ā¢ {contrib}")
# Demo 3: Trend Analysis
print("\n" + "=" * 80)
print("DEMO 3: Research Trend Analysis")
print("=" * 80)
all_papers = list(agent.paper_database.values())
trends = agent.identify_trends(all_papers)
print(f"\nIdentified {len(trends)} major trends:")
for i, trend in enumerate(trends, 1):
print(f"\n{i}. {trend.topic}")
print(f" Period: {trend.timespan[0]}-{trend.timespan[1]}")
print(f" Papers: {trend.paper_count}")
print(f" Keywords: {', '.join(trend.keywords)}")
print(f" Description: {trend.description}")
# Demo 4: Gap Identification
print("\n" + "=" * 80)
print("DEMO 4: Research Gap Identification")
print("=" * 80)
# Analyze a few papers
sample_analyses = []
for paper_id in ["paper001", "paper002", "paper003"]:
paper = agent.paper_database[paper_id]
analysis = agent.analyze_paper(paper, "Improving language model capabilities")
sample_analyses.append(analysis)
gaps = agent.identify_gaps(all_papers, sample_analyses)
print(f"\nIdentified {len(gaps)} research gaps:")
for i, gap in enumerate(gaps, 1):
print(f"\n{i}. {gap.gap_description}")
print(f" Impact: {gap.impact}, Difficulty: {gap.difficulty}")
print(f" Related Areas: {', '.join(gap.related_areas)}")
if gap.potential_approaches:
print(f" Approaches: {', '.join(gap.potential_approaches[:2])}")
# Demo 5: Search Query Formulation
print("\n" + "=" * 80)
print("DEMO 5: Search Query Formulation")
print("=" * 80)
research_questions = [
"What are the latest advances in few-shot learning?",
"How can we improve the efficiency of large language models?"
]
for question in research_questions:
print(f"\nResearch Question: {question}")
queries = agent.formulate_search_query(
question,
ResearchField.MACHINE_LEARNING
)
print("Generated Queries:")
for i, query in enumerate(queries, 1):
print(f" {i}. {query}")
# Summary
print("\n" + "=" * 80)
print("SUMMARY")
print("=" * 80)
summary = """
The Research Agent demonstrates comprehensive capabilities for academic research:
KEY CAPABILITIES:
1. Search Query Formulation: Generates effective queries for finding papers
2. Paper Discovery: Searches and retrieves relevant academic papers
3. Content Analysis: Extracts methods, findings, and contributions
4. Trend Identification: Identifies research trends and patterns
5. Gap Analysis: Finds unexplored research opportunities
6. Literature Synthesis: Combines findings into coherent reviews
BENEFITS:
- Accelerates literature review process significantly
- Ensures comprehensive coverage of research area
- Identifies patterns across many papers
- Discovers research opportunities systematically
- Generates well-structured literature reviews
USE CASES:
- PhD thesis literature reviews
- Research proposal development
- Grant application background sections
- Systematic literature reviews
- Technology landscape analysis
- Keeping up with research trends
- Expert identification
- Research collaboration opportunities
PRODUCTION CONSIDERATIONS:
1. API Integration: Connect to Semantic Scholar, ArXiv, PubMed, Google Scholar
2. Full-Text Access: Integrate with paper repositories and libraries
3. Citation Analysis: Build and analyze complete citation networks
4. PDF Processing: Extract text from PDF papers
5. Database Storage: Persist papers, analyses, and reviews
6. Visualization: Generate trend plots, citation graphs, topic maps
7. Collaboration: Support team research and shared libraries
8. Export: Generate LaTeX, Word, and markdown outputs
9. Updates: Monitor new papers and alert on relevant publications
10. Quality Assessment: Evaluate paper quality and venue prestige
ADVANCED EXTENSIONS:
- Automatic paper summarization
- Research proposal generation
- Hypothesis formulation from gaps
- Experiment design suggestions
- Funding opportunity matching
- Researcher network analysis
- Multi-language paper support
- Domain-specific paper understanding
- Automated peer review assistance
- Research impact prediction
The agent transforms the manual, time-consuming process of literature review
into an efficient, comprehensive, and intelligent workflow.
"""
print(summary)
if __name__ == "__main__":
demonstrate_research_agent()
Agentic Design Pattern: Research Agent
Pattern 074: Research Agent – Line-by-Line Explanation
Overview
This document provides a detailed line-by-line explanation of the Research Agent pattern implementation.
Pattern Description
Pattern 074: Research Agent
Description: A Research Agent is a specialized AI agent designed to conduct academic and professional research, including literature review, paper analysis, citation tracking, knowledge synthesis, and research gap identification. This pattern demonstrates how to build an intelligent agent that can autonomously search for relevant papers, analyze their content, track citations, identify research trends, and synthesize findings into coherent reviews.
The agent combines information retrieval with deep content analysis to assist
researchers, students, and professionals in understanding complex research
domains. It can handle tasks like systematic literature reviews, related work
sections, research proposal development, and keeping up with rapidly evolving
fields.
Components: 1. Paper Discovery: Searches academic databases and repositories 2. Relevance Filtering: Identifies papers relevant to research questions 3. Content Analyzer: Extracts key information from papers 4. Citation Network: Builds and analyzes citation graphs 5. Trend Analyzer: Identifies research trends and emerging topics 6. Gap Identifier: Finds unexplored areas and opportunities 7. Synthesis Engine: Combines findings into coherent narratives 8. Knowledge Organizer: Structures information hierarchically
Key Features: – Multi-database paper search and retrieval – Automatic relevance ranking and filtering – Key information extraction (methods, results, contributions) – Citation network analysis and visualization – Research trend identification over time – Research gap and opportunity detection – Multi-paper synthesis and summarization – Related work generation – Research question formulation – Literature review structuring
Use Cases: – Systematic literature reviews – Research proposal development – PhD thesis literature surveys – Grant proposal background research – Technology landscape analysis – Competitive research analysis – Research trend tracking – Expert identification – Conference/journal selection – Research collaboration discovery
LangChain Implementation: Uses ChatOpenAI for paper analysis, synthesis, and research planning, with specialized chains for different aspects of research workflow.
File Statistics
- Total Lines: 1094
- Classes: 10
- Functions: 1
- Constants: 0
Module Documentation
"""
Pattern 074: Research Agent
Description:
A Research Agent is a specialized AI agent designed to conduct academic and
professional research, including literature review, paper analysis, citation
tracking, knowledge synthesis, and research gap identification. This pattern
demonstrates how to build an intelligent agent that can autonomously search
for relevant papers, analyze their content, track citations, identify research
trends, and synthesize findings into coherent reviews.
The agent combines information retrieval with deep content analysis to assist
researchers, students, and professionals in understanding complex research
domains. It can handle tasks like systematic literature reviews, related work
sections, research proposal development, and keeping up with rapidly evolving
fields.
Components:
1. Paper Discovery: Searches academic databases and repositories
2. Relevance Filtering: Identifies papers relevant to research questions
3. Content Analyzer: Extracts key information from papers
4. Citation Network: Builds and analyzes citation graphs
5. Trend Analyzer: Identifies research trends and emerging topics
6. Gap Identifier: Finds unexplored areas and opportunities
7. Synthesis Engine: Combines findings into coherent narratives
8. Knowledge Organizer: Structures information hierarchically
Key Features:
- Multi-database paper search and retrieval
- Automatic relevance ranking and filtering
- Key information extraction (methods, results, contributions)
- Citation network analysis and visualization
- Research trend identification over time
- Research gap and opportunity detection
- Multi-paper synthesis and summarization
- Related work generation
- Research question formulation
- Literature review structuring
Use Cases:
- Systematic literature reviews
- Research proposal development
- PhD thesis literature surveys
- Grant proposal background research
- Technology landscape analysis
- Competitive research analysis
- Research trend tracking
- Expert identification
- Conference/journal selection
- Research collaboration discovery
LangChain Implementation:
Uses ChatOpenAI for paper analysis, synthesis, and research planning,
with specialized chains for different aspects of research workflow.
"""
Purpose: This module implements the Research Agent pattern, providing a structured approach to agent design and implementation.
Import Statements
Standard Library Imports
import collections.defaultdict
import dataclasses.dataclass
import dataclasses.field
import datetime.datetime
import enum.Enum
import langchain_core.output_parsers.StrOutputParser
import langchain_core.prompts.ChatPromptTemplate
import os
import re
import typing.Any
import typing.Dict
import typing.List
import typing.Optional
import typing.Set
import typing.Tuple
Purpose: Core Python functionality:
– collections.defaultdict: Specialized container datatypes
– dataclasses.dataclass: Simplified class definitions with @dataclass
– dataclasses.field: Simplified class definitions with @dataclass
– datetime.datetime: Date and time handling
– enum.Enum: Enumeration support
– langchain_core.output_parsers.StrOutputParser: Regular expression operations
– langchain_core.prompts.ChatPromptTemplate: Regular expression operations
– os: Environment variable access and system operations
– re: Regular expression operations
– typing.Any: Type hints for better code documentation (List, Dict, Any, Optional)
– typing.Dict: Type hints for better code documentation (List, Dict, Any, Optional)
– typing.List: Type hints for better code documentation (List, Dict, Any, Optional)
– typing.Optional: Type hints for better code documentation (List, Dict, Any, Optional)
– typing.Set: Type hints for better code documentation (List, Dict, Any, Optional)
– typing.Tuple: Type hints for better code documentation (List, Dict, Any, Optional)
LangChain Imports
from langchain_openai.ChatOpenAI
Purpose: LangChain framework components: – LLM Interface: Wrapper for OpenAI models – Chains: Composable building blocks for complex workflows
Other Imports
from dotenv.load_dotenv
Purpose: Additional dependencies for specialized functionality.
Classes
Class: ResearchField
Lines 73-86
Inherits from: Enum
Purpose:
Academic research fields
Class: PaperType
Lines 89-98
Inherits from: Enum
Purpose:
Types of research papers
Class: ResearchPhase
Lines 101-107
Inherits from: Enum
Purpose:
Phases of research
Class: ResearchPaper
Lines 111-123
Purpose:
Represents a research paper
Class: PaperAnalysis
Lines 127-136
Purpose:
Analysis of a research paper
Class: CitationRelationship
Lines 140-145
Purpose:
Citation relationship between papers
Class: ResearchTrend
Lines 149-157
Purpose:
Identified research trend
Class: ResearchGap
Lines 161-168
Purpose:
Identified research gap
Class: LiteratureReview
Lines 172-181
Purpose:
Complete literature review
Class: ResearchAgent
Lines 184-891
Purpose:
Agent for conducting academic and professional research.
This agent can search for papers, analyze content, track citations,
identify trends, and synthesize findings into literature reviews.
Methods:
__init__(self)
Lines 192-208
Purpose:
Initialize the research agent with specialized LLMs
Parameters:
_initialize_sample_papers(self)
Lines 210-277
Purpose:
Initialize sample research papers for demonstration
Parameters:
formulate_search_query(self, research_question, field)
Lines 279-327
Purpose:
Formulate effective search queries for the research question.
Args:
research_question: The research question to investigate
field: Academic field of research
Returns:
List of search query strings
Parameters:
– research_question: Question to answer
– field: Field
Returns: <ast.Subscript object at 0x000001A06D6FCB90>
search_papers(self, query, max_results)
Lines 329-380
Purpose:
Search for papers matching the query.
Args:
query: Search query
max_results: Maximum number of results
Returns:
List of matching papers
Parameters:
– query: Input query or question
– max_results: Max results
Returns: <ast.Subscript object at 0x000001A06D632390>
analyze_paper(self, paper, research_context)
Lines 382-494
Purpose:
Analyze a research paper in context of research question.
Args:
paper: Paper to analyze
research_context: Context/question for analysis
Returns:
PaperAnalysis with extracted information
Parameters:
– paper: Paper
– research_context: Research context
Returns: PaperAnalysis
identify_trends(self, papers, min_papers)
Lines 496-577
Purpose:
Identify research trends from a collection of papers.
Args:
papers: List of papers to analyze
min_papers: Minimum papers to form a trend
Returns:
List of identified trends
Parameters:
– papers: Papers
– min_papers: Min papers
Returns: <ast.Subscript object at 0x000001A06D6B1990>
_fallback_trends(self, papers)
Lines 579-605
Purpose:
Generate fallback trends from papers
Parameters:
– papers: Papers
Returns: <ast.Subscript object at 0x000001A06D6A8CD0>
identify_gaps(self, papers, analyses)
Lines 607-703
Purpose:
Identify research gaps from paper analyses.
Args:
papers: List of papers
analyses: List of paper analyses
Returns:
List of identified research gaps
Parameters:
– papers: Papers
– analyses: Analyses
Returns: <ast.Subscript object at 0x000001A06D75A790>
synthesize_review(self, research_question, papers, analyses, trends, gaps)
Lines 705-824
Purpose:
Synthesize a complete literature review.
Args:
research_question: Original research question
papers: Papers reviewed
analyses: Paper analyses
trends: Identified trends
gaps: Identified gaps
Returns:
Complete LiteratureReview
Parameters:
– research_question: Question to answer
– papers: Papers
– analyses: Analyses
– trends: Trends
– gaps: Gaps
Returns: LiteratureReview
conduct_research(self, research_question, field, max_papers)
Lines 826-891
Purpose:
Conduct complete research on a question.
Args:
research_question: Question to research
field: Research field
max_papers: Maximum papers to review
Returns:
Complete LiteratureReview
Parameters:
– research_question: Question to answer
– field: Field
– max_papers: Max papers
Returns: LiteratureReview
Functions
demonstrate_research_agent()
Lines 894-1089
Purpose:
Demonstrate the research agent capabilities
Key Concepts Summary
Pattern Overview
The Research Agent pattern provides a structured approach to building AI agents with specific capabilities and behaviors. This implementation leverages LangChain’s composable components to create a robust and maintainable solution.
Core Components
- Agent Class: Main implementation of the pattern logic
- Prompt Engineering: Carefully crafted prompts to guide LLM behavior
- Tool Integration: Optional tools for extended functionality
- Memory Management: State handling across interactions
- Output Parsing: Structured response processing
When to Use
ā Good for: – Scenarios requiring this specific pattern’s strengths – Applications needing structured agent behavior – Systems where explainability is important – Production environments requiring reliability
ā Not ideal for: – Simple single-prompt tasks – Scenarios requiring different pattern characteristics – Resource-constrained environments (if computationally intensive)
Best Practices Demonstrated
1. Environment Configuration
- ā
Use
.envfiles for API keys - ā Never hardcode credentials
- ā Validate environment variables on startup
2. Error Handling
- ā Graceful degradation on failures
- ā Informative error messages
- ā Logging for debugging
3. Prompt Engineering
- ā Clear, specific instructions
- ā Structured output formats
- ā Few-shot examples when beneficial
4. Code Organization
- ā Separation of concerns (class-based design)
- ā Type hints for clarity
- ā Docstrings for documentation
5. Performance Optimization
- ā Appropriate temperature settings
- ā Token budget management
- ā Caching where applicable
LangChain Components Used
- ChatOpenAI: OpenAI model interface
- Chains: Sequential component composition
Why LangChain? – Composable abstractions for LLM applications – Rich ecosystem of integrations – Production-ready patterns – Active community and development
Execution Flow
1. Initialize agent with configuration
ā
2. Load environment and dependencies
ā
3. Prepare prompt with input data
ā
4. Invoke LLM through LangChain
ā
5. Parse and validate response
ā
6. Return formatted result
Key Decision Points: – Model selection (temperature, model type) – Prompt structure (few-shot vs zero-shot) – Error handling strategy – Output format validation
Performance Considerations
Computational Cost
- LLM Calls: Primary cost factor (based on tokens)
- Latency: Depends on model and input size
- Memory: Minimal for most patterns
Optimization Strategies
- Prompt Optimization: Minimize tokens while maintaining clarity
- Caching: Cache repeated queries/responses
- Model Selection: Use smaller models when appropriate
- Batch Processing: Group similar requests
- Streaming: For long responses, stream tokens
Cost Estimates (Approximate)
- GPT-3.5-turbo: ~$0.002 per request
- GPT-4: ~$0.03 per request
- GPT-4-turbo: ~$0.01 per request
Actual costs vary based on input/output token counts
Common Pitfalls to Avoid
1. Prompt Design Issues
ā Vague or ambiguous instructions
ā
Clear, specific, testable prompts
2. Error Handling
ā Ignoring API errors or rate limits
ā
Implement retry logic and graceful degradation
3. Token Management
ā Exceeding model token limits
ā
Truncate inputs, chunk processing
4. Security
ā Exposing API keys in code
ā
Use environment variables
5. Testing
ā Only testing happy paths
ā
Test edge cases, failures, malformed inputs
6. Cost Management
ā Unmonitored API usage
ā
Implement budgets, alerts, and monitoring
Further Improvements
Production Enhancements
- Monitoring: Add logging, metrics, tracing
- Validation: Input/output schema validation
- Rate Limiting: Implement request throttling
- Caching: Redis or in-memory cache for responses
- Async Support: Non-blocking operations for scalability
Feature Extensions
- Multi-Model Support: Switch between providers
- Fallback Models: Graceful degradation strategy
- Custom Tools: Domain-specific function calling
- Advanced Memory: Vector stores for long-term memory
- Evaluation: Automated testing and benchmarking
Code Quality
- Testing: Unit tests, integration tests
- Documentation: API docs, usage examples
- Type Safety: Stricter type hints
- Configuration: Externalized config files
Related Patterns
- Pattern 073
- Pattern 075
Pattern Families: – Reasoning Patterns (001-011) – Multi-Agent Patterns (012-019) – Tool Use Patterns (020-025) – Memory Patterns (026-031) – And many more…
Summary
The Research Agent pattern provides a powerful approach to building AI agents with specific capabilities. This implementation demonstrates:
ā Key Strengths: – Clean, maintainable code structure – LangChain integration for composability – Type hints and documentation – Production-ready patterns
šÆ Best Use Cases: – Applications requiring this pattern’s characteristics – Systems needing structured agent behavior – Production environments requiring reliability
š Learning Resources: – LangChain Documentation: https://python.langchain.com/ – Pattern Documentation: See related patterns – OpenAI API Reference: https://platform.openai.com/docs/
File: 074_research_agent.py
Total Lines: 1094
Pattern Category: AI Agent Design Pattern #074
Generated from actual implementation code
For educational and reference purposes
Leave a Reply