langchain-builtsimple

LangChain integration for Built-Simple research APIs, providing easy access to PubMed and ArXiv scientific literature.

Features

• PubMed Retriever & Tool - Search 4.5M+ peer-reviewed biomedical articles
• ArXiv Retriever & Tool - Search 2.7M+ preprints in physics, math, CS, and ML
• Combined Retriever - Search both sources simultaneously
• RAG-ready - Documents include full metadata for citations
• Agent-compatible - Tools work with LangChain agents out of the box

What Data is Included

PubMed Documents

• page_content: Title + abstract (default) OR full article text (with include_full_text=True)
• metadata:
  • pmid - PubMed ID
  • title - Article title
  • journal - Journal name
  • pub_year - Publication year
  • doi - DOI identifier
  • url - Link to PubMed
  • has_full_text - Boolean indicating if full text was fetched
  • full_text_length - Character count when available

🔥 FULL TEXT AVAILABLE! Unlike most research APIs, Built-Simple provides complete article text:

# Get full articles (15K-70K chars each!)
retriever = BuiltSimplePubMedRetriever(limit=5, include_full_text=True)
docs = retriever.invoke("cancer immunotherapy")

for doc in docs:
    print(f"Full text: {len(doc.page_content)} chars")  # ~15,000-70,000!

ArXiv Documents

• page_content: Title + authors + abstract
• metadata:
  • arxiv_id - ArXiv ID (e.g., "2301.12345")
  • title - Paper title
  • authors - Author list
  • year - Publication year
  • url - ArXiv page link
  • pdf_url - Direct PDF link

⚠️ Abstracts only - Full PDFs are not downloaded. Use pdf_url to fetch if needed.

Installation

pip install langchain-builtsimple

For development with examples:

pip install langchain-builtsimple[dev]

Quick Start

Basic Retrieval

from langchain_builtsimple import BuiltSimplePubMedRetriever, BuiltSimpleArxivRetriever

# Search PubMed
pubmed = BuiltSimplePubMedRetriever(limit=5)
docs = pubmed.invoke("CRISPR gene therapy")

for doc in docs:
    print(f"Title: {doc.metadata['title']}")
    print(f"Journal: {doc.metadata['journal']}")
    print(f"URL: {doc.metadata['url']}\n")

# Search ArXiv
arxiv = BuiltSimpleArxivRetriever(limit=5)
docs = arxiv.invoke("transformer neural networks")

for doc in docs:
    print(f"Title: {doc.metadata['title']}")
    print(f"Authors: {doc.metadata['authors']}")
    print(f"ArXiv ID: {doc.metadata['arxiv_id']}\n")

RAG Chain with ChatOpenAI

from langchain_builtsimple import BuiltSimplePubMedRetriever
from langchain_openai import ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough

# Create retriever
retriever = BuiltSimplePubMedRetriever(limit=5)

# Format documents for context
def format_docs(docs):
    return "\n\n".join(
        f"[{i+1}] {doc.metadata['title']} ({doc.metadata.get('pub_year', 'N/A')})\n{doc.page_content}"
        for i, doc in enumerate(docs)
    )

# Create RAG prompt
prompt = ChatPromptTemplate.from_template("""
Answer the question based on the following research papers. 
Cite papers by number [1], [2], etc.

Papers:
{context}

Question: {question}

Answer:""")

# Build chain
llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)

chain = (
    {"context": retriever | format_docs, "question": RunnablePassthrough()}
    | prompt
    | llm
    | StrOutputParser()
)

# Ask a question
answer = chain.invoke("What are the latest developments in CAR-T cell therapy?")
print(answer)

Agent with Research Tools

from langchain_builtsimple import BuiltSimplePubMedTool, BuiltSimpleArxivTool
from langchain_openai import ChatOpenAI
from langchain.agents import AgentExecutor, create_tool_calling_agent
from langchain_core.prompts import ChatPromptTemplate

# Create tools
tools = [
    BuiltSimplePubMedTool(),  # For biomedical research
    BuiltSimpleArxivTool(),   # For CS/ML/physics papers
]

# Create agent
llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)

prompt = ChatPromptTemplate.from_messages([
    ("system", """You are a research assistant with access to scientific databases.
    Use pubmed_search for medical/biological topics.
    Use arxiv_search for AI/ML/physics/math topics.
    Always cite your sources."""),
    ("human", "{input}"),
    ("placeholder", "{agent_scratchpad}"),
])

agent = create_tool_calling_agent(llm, tools, prompt)
executor = AgentExecutor(agent=agent, tools=tools, verbose=True)

# Run agent
response = executor.invoke({
    "input": "Find recent papers on using transformers for drug discovery"
})
print(response["output"])

API Reference

Retrievers

All retrievers inherit from langchain_core.retrievers.BaseRetriever and return List[Document].

BuiltSimplePubMedRetriever

BuiltSimplePubMedRetriever(
    base_url: str = "https://pubmed.built-simple.ai",
    limit: int = 10,
    timeout: float = 30.0
)

Document Metadata:

• source: "pubmed"
• pmid: PubMed ID
• title: Paper title
• journal: Journal name
• pub_year: Publication year
• doi: DOI (if available)
• url: Link to PubMed page

BuiltSimpleArxivRetriever

BuiltSimpleArxivRetriever(
    base_url: str = "https://arxiv.built-simple.ai",
    limit: int = 10,
    timeout: float = 30.0
)

Document Metadata:

• source: "arxiv"
• arxiv_id: ArXiv identifier
• title: Paper title
• authors: List of author names
• year: Publication year
• url: Link to ArXiv page

BuiltSimpleResearchRetriever

Searches both PubMed and ArXiv, interleaving results.

BuiltSimpleResearchRetriever(
    pubmed_url: str = "https://pubmed.built-simple.ai",
    arxiv_url: str = "https://arxiv.built-simple.ai",
    limit_per_source: int = 5,
    timeout: float = 30.0
)

Tools

All tools inherit from langchain_core.tools.BaseTool and can be used with LangChain agents.

BuiltSimplePubMedTool

• Name: pubmed_search
• Description: Search PubMed for peer-reviewed biomedical literature
• Input: query (str), limit (int, default=5)

BuiltSimpleArxivTool

• Name: arxiv_search
• Description: Search ArXiv for preprints in physics, math, CS, ML
• Input: query (str), limit (int, default=5)

BuiltSimpleResearchTool

• Name: research_search
• Description: Search both PubMed and ArXiv simultaneously
• Input: query (str), limit (int, default=5)

Examples

See the examples/ directory for complete working examples:

• basic_retrieval.py - Simple retriever usage
• rag_chain.py - RAG chain with ChatOpenAI
• agent_with_tools.py - Agent with research tools

License

MIT License - see LICENSE for details.

Links

• Built-Simple
• PubMed API
• ArXiv API
• LangChain