Travelling alone in this journey of Gen-AI we used to built things in a messy 😵‍💫 manner, and locally it is ok. But what if we properly modularize , Scalable it and can magically modify quickly for more optimize output 🤖and even easily understanding complex work flow ⚙️.

So, In this article we will look closely how could we achieve and even get better tool calling than those usual prompts 😎. Also at end I would teach how could you also code 🧑‍💻 and get visualizations from jupyter notebook. So Stay Tune. 🤗

🤔 Pre-Requisites

How to work with Basic Models
Prompting basics
Pydantic and TypeDict (I will try to give a basic idea)

🌌 Introduction: The Galactic Need for Agent Orchestration

In a galaxy of LLMs, APIs, and tools, not all intelligence is centralized. Some agents specialize in star mapping (retrieval), others in alien translation (generation), and others in anomaly detection (tools & logic). How do we coordinate these cosmic entities?

So till now we may have included different use cases or either used functions and wrote if-else, but we might not easily have better control over a flow or organization. LangGraph provides tools where still under functions we can make graphs structures, and organize them as we want (think of like n8n but in code)

Introduce the idea of agents as autonomous crew members.
State the challenge: They need coordination to complete a mission.
Enter: LangGraph, your mission control system in the stars.

🛰️ What Is LangGraph? (Mission Control on the Moon)

LangGraph is like NASA’s mission control for your agents. It uses graphs, not just pipelines, letting you dynamically route tasks between agents, tools, or memory systems.

Built on stateful computation over graphs.
Inspired by ideas from LangChain, but more explicit and programmable.
Think of each node as a satellite, each edge as a signal.

👩‍🚀 What Are Agents? (Autonomous Astronauts with Special Roles)

Agents are your intelligent astronauts—each equipped with certain skills and knowledge.

Examples:
- DataScout: Retrieval agent scanning nebulae of documents.
- CommsOfficer: Handles language generation and user queries.
- EngineerBot: Executes tools, fixes logic circuits, or calls APIs.

Each agent is like an astronaut with a unique role on the spacecraft.

🛸 Designing the Agent Workflow with LangGraph

🚦 Step 1: Define the States (Defining Galactic Coordinates)

from typing_extensions import TypedDict

class State(TypedDict):
    query: str
    llm_result: str | None

🌌 Step 2: Define Nodes (Mapping Constellation)

import os
from dotenv import load_dotenv
from openai import OpenAI

load_dotenv()

client = OpenAI(
    api_key=os.getenv("GEMINI_API_KEY"),
    base_url="https://generativelanguage.googleapis.com/v1beta/openai/"
)

# Like this we can make more nodes
def chat_bot(state: State): # this is a node

    query = state['query']
    response = client.chat.completions.create(
        model="gemini-2.0-flash",
        messages={"role": "user", "content": query}
    )
    result = response.choices[0].message.content
    state["llm_result"] = result

    return state

🔁 Step 3: Define Graph Building (Wormholes Between Agents)

from langgraph.graph import StateGraph, START, END

graph_builder = StateGraph(State)# initialized a graph

graph_builder.add_node("chat_bot", chat_bot)# added a node

#Building edges
graph_builder.add_edge(START, "chat_bot")
graph_builder.add_edge("chat_bot", END)

# compile the graph
graph = graph_builder.compile()

💫 Step 4: Initializing our Graph (Running through Stars)

def main():
    user = input("> ")

    #invoke the graph
    _state = {
        "query": user,
        "llm_result": None
    }
    graph_result = graph.invoke(_state)

    print("graph_result", graph_result)


main()

🤝 More advance workflow (Tool Calling Integration): (ChatGroq)

Our Dependencies:

from typing_extensions import TypedDict
from langchain_core.messages import AnyMessage #allowing both Human or ai messages
from typing import Annotated
from langgraph.graph.message import add_messages
from langgraph.graph import StateGraph, START, END
from langgraph.prebuilt import ToolNode
from langgraph.prebuilt import tools_condition # when should we use tool node and which path i need to take
from tools import llm_with_tools, available_tools
from IPython.display import display, Image

Defining State:

class State(TypedDict):
    messages: Annotated[list[AnyMessage], add_messages] # return a dict

Defining some Tools:

Here I will take arxiv (Research Related Data), Wiki (General Info), Tavily (Web searches)

from langchain_community.tools import ArxivQueryRun, WikipediaQueryRun
from langchain_community.utilities import ArxivAPIWrapper, WikipediaAPIWrapper
from langchain_tavily import TavilySearch
from langchain_groq import ChatGroq

api_wrapper_arxiv = ArxivAPIWrapper(top_k_results=2, doc_content_chars_max=500)
arxiv = ArxivQueryRun(api_wrapper=api_wrapper_arxiv, description="Query arxiv papers")

api_wrapper_wiki = WikipediaAPIWrapper(top_k_results=1, doc_content_chars_max=500)
wiki = WikipediaQueryRun(api_wrapper=api_wrapper_wiki)

tavily=TavilySearch()

Combining Tools:

# Combining all the tools in the list
available_tools = [arxiv, wiki, tavily]

llm=ChatGroq(model="qwen-qwq-32b")
# print(llm.invoke("what is ai"))

llm_with_tools = llm.bind_tools(tools=available_tools)

def tool_calling_llm(state: State):
    return {"messages": [llm_with_tools.invoke(state["messages"])]}

Automating a Hybrid setup through (LangGraph):

#Build graph
graph_builder = StateGraph(State)
graph_builder.add_node("tool_calling_llm", tool_calling_llm)
graph_builder.add_node("tools", ToolNode(available_tools)) # passing available tools in

#edges
graph_builder.add_edge(START, "tool_calling_llm")
graph_builder.add_conditional_edges("tool_calling_llm", tools_condition)
# if latest message(result) from assistant is a tool -> tools_condition routes to tools
# if latest message(result) from assistant is not a tool -> tools_condition routes to END

graph_builder.add_edge("tools", "tool_calling_llm")

graph = graph_builder.compile()

Visualizing the node (in Jupyter notebook)

display(Image(graph.get_graph().draw_mermaid_png()))

Start Conversations:

messages = graph.invoke({"messages": "what is the ai news and then what is the current quantum research progress?"})
for m in messages["messages"]:
    m.pretty_print()

End Result of AI (Hybrid Thinking → Both Automated by Human and Decision taken by AI for Tool calling decisions):

================================[1m Human Message [0m=================================

what is the ai news and from recent research paper on quantum computing?
==================================[1m Ai Message [0m==================================
Tool Calls:
  tavily_search (0pyj552g2)
 Call ID: 0pyj552g2
  Args:
    query: AI and quantum computing news
    search_depth: advanced
    topic: news
  arxiv (4t2m1ben1)
 Call ID: 4t2m1ben1
  Args:
    query: quantum computing and artificial intelligence recent research
===========h======================[1m Tool Message [0m=================================
Name: tavily_search

{"query": "AI and quantum computing news", "follow_up_questions": null, "answer": null, "images": [], "results": [{"url": "https://www.axios.com/2025/06/11/nvidia-nvda-ai-jensen-huang", "title": "Nvidia CEO Jensen Huang is getting quantum computing fever - Axios", "score": 0.7295395, "published_date": "Thu, 12 Jun 2025 02:25:45 GMT", "content": "Driving the news: Huang told VivaTech 2025 in Paris Wednesday that quantum computing is raring to go — something of a reversal from the outright skepticism he expressed earlier this year.\n\n   \"Today I can tell you there's an inflection point happening,\" Huang said. \n   \"It is clear now we are within reach of being able to apply quantum computing — quantum classical computing — in areas that can solve some interesting problems in the coming years,\" he added. \"This is a really exciting time.\" [...] Image 2: A person in a leather jacket and silvery hair holds their hands up while speaking\n\nNvidia CEO Jensen Huang speaks at the 2025 VivaTech conference in Paris on Wednesday. Photo: Nathan Laine/Bloomberg via Getty Images\n\nJensen Huang's doubts about quantum computing's potential — or at least the speed with which it'll develop — appear to be fading.\n\nWhy it matters:The Nvidia CEO's perspectives on any technologies touching the AI economy are market moving and narrative shaping. [...] Quantum computers use the quantum state of an object to produce what's known as qubits. The complex math behind these qubits can be plugged into special algorithms to do calculations that would be impossible for a classical computer to perform.\n   This could have huge implications for everything from drug discovery, to financial modeling, to AI development.", "raw_content": null}, {"url": "https://thequantuminsider.com/2025/06/15/researchers-use-trapped-ion-quantum-computer-to-tackle-tricky-protein-folding-problems/", "title": "Researchers Use Trapped-Ion Quantum Computer to Tackle Tricky Protein Folding Problems - The Quantum Insider", "score": 0.52144736, "published_date": "Sun, 15 Jun 2025 11:59:19 GMT", "content": "However, protein folding is an incredibly complicated phenomenon, requiring calculations that are too complex for classical computers to practically solve, although progress, particularly through new artificial intelligence techniques, is being made. The trickiness of protein folding, however, makes it an interesting use case for quantum computing. [...] Image 14\n\n   News\n       Capital Markets\n       National\n       Quantum Computing Business\n       Research\n\n   Resources\n       Education\n       Reports\n       Featured\n       Insights\n       Interviews\n       Media\n       Women In Quantum\n\n   Product\n   Advisory\n   Marketing\n   About Us\n\n   News\n       Capital Markets\n       National\n       Quantum Computing Business\n       Research [...] News\n       Capital Markets\n       National\n       Quantum Computing Business\n       Research\n\n   Resources\n       Education\n       Reports\n       Featured\n       Insights\n       Interviews\n       Media\n       Women In Quantum\n\n   Product\n   Advisory\n   Marketing\n   About Us\n\nSubscribe\n\nGet Access\n\nImage 9\n\nImage 11\n\n   News\n       Capital Markets\n       National\n       Quantum Computing Business\n       Research", "raw_content": null}, {"url": "https://www.sciencedaily.com/releases/2025/06/250611054144.htm", "title": "This mind-bending physics breakthrough could redefine timekeeping - ScienceDaily", "score": 0.5100334, "published_date": "Wed, 11 Jun 2025 10:14:49 GMT", "content": "Feb. 17, 2021 — Scientists have used cutting-edge research in quantum computation and quantum technology to pioneer a radical new approach to determining how our Universe works at its most fundamental ... \n\nPrintEmailShare\n\n  \n\nBreaking\n--------------------------------------------------\n\nthis hour [...] Artificial Intelligence\n           Mathematics\n           Quantum Computers\n           Robotics\n           Virtual Reality\n           ... _more topics_\n\n   Enviro\n       \nView all the latest top news in the environmental sciences, \n\nor browse the topics below: \n\nPlants & Animals \n           Agriculture and Food\n           Biology\n           Biotechnology\n           Extinction\n           Microbes and More\n           ... _more topics_\n\nEarth & Climate [...] Dec. 4, 2023 — The prevailing assumption has been that Einstein's theory of gravity must be modified, or 'quantized', in order to fit within quantum theory. This is the approach of two leading ... \n\nFinding Coherence in Quantum Chaos\n\nMay 26, 2022 — A theoretical breakthrough in understanding quantum chaos could open new paths into researching quantum information and quantum computing, many-body physics, black holes, and the still-elusive ...", "raw_content": null}, {"url": "https://www.forbes.com/sites/stevemcdowell/2025/06/12/ibm-promises-enterprise-ready-quantum-computing-by-2029/", "title": "IBM Promises Enterprise-Ready Quantum Computing By 2029 - Forbes", "score": 0.477969, "published_date": "Fri, 13 Jun 2025 00:35:09 GMT", "content": "It’s early days for quantum computing and the competitive landscape remains fractured. Startups like QuEra and PsiQuantum pursue different technical approaches but lack IBM's enterprise relationships and infrastructure capabilities. Google and Amazon possess the resources to compete, but they have not committed to IBM's aggressive commercialization timeline or its enterprise-focused architecture. [...] The quantum computing market is at an inflection point. IBM's Starling system will transform quantum computing from an expensive research curiosity into enterprise infrastructure that delivers measurable business value. This requires IBM to execute, but the company has built credibility by hitting every public milestone its put on its quantum roadmap. [...] For executives evaluating quantum computing strategies, the question has shifted from whether quantum computing will impact their industries to how quickly they can integrate quantum capabilities into competitive advantage. Choosing a partner to help with that journey is a critical first step, with IBM taking an early leadership position.", "raw_content": null}, {"url": "https://www.forbes.com/sites/petercohan/2025/06/12/rigetti-computing-stock-down-37-as-nvidia-ceo-sees-inflection-point/", "title": "Rigetti Computing Stock Down 37% As Nvidia CEO Sees ‘Inflection Point’ - Forbes", "score": 0.4757764, "published_date": "Thu, 12 Jun 2025 15:39:04 GMT", "content": "Perhaps Huang was motivated to make his Wednesday remarks by IBM — which on June 10 announced plans to launch in 2029 a large-scale QC capable of “operating without errors,” according to IEEE Spectrum.\n\nGoogle released a quantum computing chip called Willow in December, saying the technology \"paves the way to a useful, large-scale quantum computer,\" noted Yahoo! Finance. While Amazon and Microsoft announced two quantum chips in February. [...] It remains to be seen which problems will be more amenable to QC than classical computing. “It is unclear whether QC will help make large language models easier to train and operate. Who could know what AI would end up doing?” [...] The reason? “Quantum computing is reaching an inflection point,” Nvidia CEO Jensen Huang told a crowd at Nvidia GTC Paris Wednesday, reported Yahoo! Finance. “We are within reach” of using quantum computers for “areas that can solve some interesting problems in the coming years,” Huang added.", "raw_content": null}], "response_time": 1.28}
=================================[1m Tool Message [0m=================================
Name: arxiv

Published: 2007-05-23
Title: The Road to Quantum Artificial Intelligence
Authors: Kyriakos N. Sgarbas
Summary: This paper overviews the basic principles and recent advances in the emerging
field of Quantum Computation (QC), highlighting its potential application to
Artificial Intelligence (AI). The paper provides a very brief introduction to
basic QC issues like quantum registers, quantum gates and quantum algorithms
and then it presents references, ideas and research guidelines on how QC can be
==================================[1m Ai Message [0m==================================

Here’s a synthesis of the latest news and research at the intersection of AI and quantum computing:

---

### **Recent News Highlights**  
1. **Nvidia’s Shift on Quantum Computing (Axios, June 2025)**  
   - NVIDIA CEO Jensen Huang announced an "inflection point" in quantum computing, signaling growing confidence in its practical applications for solving complex problems, including those relevant to AI. This reverses his earlier skepticism, indicating industry momentum.  
   - Focus: Hybrid quantum-classical systems could accelerate AI workloads, though scalable integration remains challenging.  

2. **IBM’s 2029 Quantum Roadmap (Forbes, June 2025)**  
   - IBM plans to launch enterprise-ready quantum systems by 2029, targeting error-corrected quantum computing. This could enable applications in AI optimization, drug discovery, and financial modeling.  

3. **Quantum for Protein Folding (The Quantum Insider, June 2025)**  
   - Researchers used trapped-ion quantum computers to tackle protein folding—a problem AI has already made progress on. Quantum computing could complement AI by handling even more complex molecular simulations.  

---

### **Recent Research Insights**  
- The provided arXiv example paper (2007) is outdated, but **more recent arXiv papers** (not shown in the tool response) likely explore:  
  - **Hybrid AI-Quantum Models**: Leveraging quantum algorithms to enhance machine learning tasks (e.g., quantum neural networks).  
  - **Error Mitigation**: Improving quantum computing’s reliability for real-world AI applications.  
  - **Optimization Problems**: Quantum algorithms for training large AI models faster.  

---

### **Key Takeaways**  
- **Industry Momentum**: Tech giants like IBM and NVIDIA are prioritizing quantum-AI integration.  
- **Near-Term Applications**: Drug discovery (e.g., protein folding), financial modeling, and AI optimization stand out as promising areas.  
- **Challenges**: Error-prone quantum hardware and the need for specialized algorithms remain hurdles.  

Would you like deeper insights into a specific angle, like use cases or technical challenges?

🛰️ Example: A LangGraph-powered RAG Agent Workflow

Mission Objective: Answer any question using PDFs, tools, and live web data

graph LR
    A[User Query Received] --> B{Do I need context?}
    B -->|Yes| C[Search in Vector DB]
    B -->|No| D[Use Memory or History]
    C --> E[Call LLM for Answer]
    D --> E
    E --> F{Tool Needed?}
    F -->|Yes| G[Call Tool Agent]
    F -->|No| H[Return Answer]
    G --> H

Each of these nodes is a LangGraph state, each decision a transition.

🌍 Conclusion: The Sky Is Not the Limit

With LangGraph, you don’t just build agents—you orchestrate them like stars in a galaxy. Each one does its job, but together, they enable deep-space intelligence.

LangGraph Agents: Orchestrating Intelligence Across the Stars | Gen AI-5