Create app.py

app.py

@@ -0,0 +1,164 @@
+import streamlit as st
+from crewai import Agent, Task, Crew
+from crewai_tools import SerperDevTool, BaseTool
+import arxiv
+import os
+
+# Custom ArxivSearchTool
+class ArxivSearchTool(BaseTool):
+    name: str = "ArxivSearch"
+    description: str = "Tool to search scientific papers from arXiv"
+
+    def _run(self, query: str) -> str:
+        results = list(arxiv.Search(query=query, max_results=3).results())
+        return "\n".join(f"{r.title} - {r.entry_id}" for r in results)
+
+# Custom FileIOTool
+class FileIOTool(BaseTool):
+    name: str = "FileIOTool"
+    description: str = "Tool to read from and write to files"
+
+    def _run(self, action: str, filename: str, content: str = None) -> str:
+        if action == "read":
+            try:
+                with open(filename, 'r') as f:
+                    return f.read()
+            except FileNotFoundError:
+                return f"Error: File {filename} not found."
+        elif action == "write":
+            with open(filename, 'w') as f:
+                f.write(content)
+            return f"Content written to {filename}"
+        else:
+            return "Error: Invalid action. Use 'read' or 'write'."
+
+# Streamlit App
+st.title("ResearchSpark: Generate Novel Problem Statements")
+
+# Load API keys from Streamlit secrets
+if "SERPER_API_KEY" in st.secrets:
+    os.environ["SERPER_API_KEY"] = st.secrets["SERPER_API_KEY"]
+else:
+    st.error("SERPER_API_KEY not found in secrets.")
+
+if "OPENAI_API_KEY" in st.secrets:
+    os.environ["OPENAI_API_KEY"] = st.secrets["OPENAI_API_KEY"]
+else:
+    st.error("OPENAI_API_KEY not found in secrets.")
+
+# User Inputs Form
+with st.form(key='user_inputs_form'):
+    field = st.text_input("Field of Study", value="Biology")
+    interest = st.text_input("Specific Interest", value="Genetics, CRISPR")
+    academic_level = st.text_input("Academic Level", value="Undergraduate")
+    resources = st.text_input("Available Resources", value="Python, bioinformatics tools, open-source datasets")
+    scope = st.text_input("Project Scope", value="3-month project")
+    preference = st.text_input("Preference", value="Climate change solutions")
+    submit_button = st.form_submit_button(label="Generate Problem Statement")
+
+if submit_button:
+    # User inputs dictionary
+    user_inputs = {
+        "field": field,
+        "interest": interest,
+        "academic_level": academic_level,
+        "resources": resources,
+        "scope": scope,
+        "preference": preference
+    }
+
+    # Instantiate tools
+    search_tool = SerperDevTool()
+    file_io_tool = FileIOTool()
+    arxiv_tool = ArxivSearchTool()
+
+    # Define Agents
+    researcher = Agent(
+        role='Researcher',
+        goal='Find open-access genetics papers from arXiv and Semantic Scholar',
+        backstory='Expert in sourcing academic literature from archives.',
+        tools=[search_tool, arxiv_tool, file_io_tool],
+        llm="openai/gpt-4o-mini",
+        verbose=True
+    )
+
+    analyst = Agent(
+        role='Analyst',
+        goal='Identify novel research gaps for undergraduate projects',
+        backstory='Skilled at spotting underexplored areas in research.',
+        tools=[file_io_tool],
+        llm="openai/gpt-4o-mini",
+        verbose=True
+    )
+
+    writer = Agent(
+        role='Writer',
+        goal='Craft clear, novel problem statements for students',
+        backstory='Expert in translating research gaps into actionable project aims.',
+        tools=[file_io_tool],
+        llm="openai/gpt-4o-mini",
+        verbose=True
+    )
+
+    validator = Agent(
+        role='Validator',
+        goal='Ensure the novelty of the problem statement',
+        backstory='Expert in verifying originality by cross-checking with existing research.',
+        tools=[search_tool, arxiv_tool],
+        llm="openai/gpt-4o-mini",
+        verbose=True
+    )
+
+    # Define Tasks
+    research_task = Task(
+        description=f'Search arXiv and Semantic Scholar for open-access papers on {user_inputs["interest"]} from 2024–2025. Save abstracts to a file.',
+        expected_output='A text file with 3–5 paper summaries.',
+        agent=researcher,
+        output_file='summaries.txt'
+    )
+
+    analysis_task = Task(
+        description=f'Analyze summaries.txt to identify a novel research gap suitable for an {user_inputs["academic_level"]} in {user_inputs["field"]}.',
+        expected_output='A clear description of a research gap.',
+        agent=analyst
+    )
+
+    writing_task = Task(
+        description=f'Generate a problem statement for an {user_inputs["academic_level"]} in {user_inputs["field"]} interested in {user_inputs["interest"]}, using the identified gap. Include feasibility for {user_inputs["resources"]} and {user_inputs["scope"]}.',
+        expected_output='A problem statement saved to a file in the format: "This project aims to [goal] by [approach], addressing [gap] in [context]."',
+        agent=writer,
+        output_file='problem_statement.txt'
+    )
+
+    validation_task = Task(
+        description='Search arXiv and Semantic Scholar to ensure the problem statement in problem_statement.txt is novel and not duplicated in existing research.',
+        expected_output='A confirmation that the problem statement is novel, or suggestions for refinement if duplicates are found.',
+        agent=validator,
+        output_file='validation_result.txt'
+    )
+
+    # Assemble Crew
+    crew = Crew(
+        agents=[researcher, analyst, writer, validator],
+        tasks=[research_task, analysis_task, writing_task, validation_task],
+        verbose=True
+    )
+
+    # Run Crew and display results
+    with st.spinner("Generating Problem Statement..."):
+        result = crew.kickoff()
+
+        # Display results
+        st.subheader("Problem Statement")
+        with open('problem_statement.txt', 'r') as f:
+            st.write(f.read())
+
+        st.subheader("Validation Result")
+        with open('validation_result.txt', 'r') as f:
+            st.write(f.read())
+
+        st.subheader("Summaries (References)")
+        with open('summaries.txt', 'r') as f:
+            st.write(f.read())
+
+        st.success("Generation complete!")