After building multi-agent AI with CrewAI, the OpenAI Agents SDK, and pure Python, I finally started learning LangGraph. First project: a GPT-4o-mini chatbot with a StateGraph pipeline Gradio UI. 1/ #LangGraph #AgenticAI #Python #LLM #OpenAI

Splitting your #Terraform state should not be best practice. It does not fix the problem. It shards it. The state file is one flat JSON file that acts as a global lock in a queue. People tell you to split it and call that best practice. I understand why they say it, but it is bad design. You still cannot plan across two stacks at the same time. You are stuck doing the plan-and-apply wave dance. Terragrunt exists, but it is a band-aid that mocks outputs. Orchestration and other wrappers exist too. None of it addresses the core problem. Flat state made sense in 2014, when a single engineer with a laptop was building the infrastructure. It does not make sense in 2026, which is about 100 years later in internet time. We are building Stategraph to solve this. Come watch our deep dive at Demo Day in a couple of weeks: stategraph.com/demo-day

Implemented CRAG with rule-based routing and transitioning toward LLM-based decision making! Key concepts : Nodes & Edges StateGraph (LangGraph) Conditional Edges for dynamic workflow routing 🔥 If you're interested in #AI #ML #GenAI #LangGraph, let's connect! #letsconnect

The next Stategraph demo is about our architecture and what falls out of it. At the core, it's a database. And we all know how to build on databases. We've been doing it as engineers for decades. That's what makes products like Cost so easy to build inside Stategraph. You can estimate costs right inside plan operations, then reconcile them against your cloud provider for the actuals. Can't wait to show the world. stategraph.com/demo-day #terraform

2024 = RAG 2025 = Agents 2026 = Stateful Orchestration LangGraph 1.2: agent runs are durable graph executions, not function calls. StateGraph backbone. Nodes read/write shared state. Steps auto-checkpointed. Chains pass outputs. Graphs maintain state. tanziro.com

W27 is the amplifier. W26 was the wave. W26 said: the state graph is not optional. W27 says: the state graph is the primitive only if it is bound to operations. Capability grant = the law. Skill = inert resource. Witness = the protocol. #W27 #AgentSecurity #StateGraph

WEEKLY UPDATE This week I: Got 2 HR calls from a Bangalore startup 👀 let's see if I hear back > LangGraph — learned about: → StateGraph, nodes & edges → Tool integration ReAct agent architecture → Persistent memory → human-in-the-loop → MCP (Model Context Protocol)

🛠️🧭 How to Build AI Agents from Scratch - Even If You’ve Never Done It Before. Using my 𝗦𝗠𝗔𝗥𝗧 𝗖𝗢𝗠𝗣𝗔𝗦𝗦 𝗿𝗼𝗮𝗱𝗺𝗮𝗽 I built 𝗮 𝗖𝘆𝗯𝗲𝗿 𝗗𝗲𝗳𝗲𝗻𝘀𝗲 𝗦𝘆𝘀𝘁𝗲𝗺 in 12 minutes in Langgraph. Here is my roadmap so you can build it Yourself: 𝗣𝗵𝗮𝘀𝗲 𝟭: 𝗦𝗠𝗔𝗥𝗧 (The Core) --- 》𝗦𝘁𝗲𝗽 𝟭: 𝗦etup & Stack ✸ Don't start from zero. Initialize the core requirements. ✸ Choose cost-effective models for high-volume analysis. → Tools: LangGraph, LangChain, OpenAI 》𝗦𝘁𝗲𝗽 𝟮: 𝗠ission ✸ Define the exact goal. We aren't building a chatbot. ✸ Goal: "Ingest logs, detect anomalies, and classify risks." → Outcome: A clear objective for the agent system. 》𝗦𝘁𝗲𝗽 𝟯: 𝗔rchitecture ✸ Map out the team structure (Nodes). ✸ Assign specific roles: Ingest, Detect, Classify, Report. → Concept: Multi-Agent Nodes instead of one massive prompt. 》𝗦𝘁𝗲𝗽 𝟰: 𝗥easoning ✸ Equip agents with "ReAct" (Reason Act) capabilities. ✸ Allow them to think before they decide to use a tool. → Framework: ReAct Agent Construction 》𝗦𝘁𝗲𝗽 𝟱: 𝗧ools ✸ Create Python functions the AI can call upon. ✸ Specific skills: Pattern Detector, Anomaly Detector, Threat Lookup. → Code: @ tool decorated functions 𝗣𝗵𝗮𝘀𝗲 𝟮: 𝗖𝗢𝗠𝗣𝗔𝗦𝗦 (𝗧𝗵𝗲 𝗡𝗮𝘃𝗶𝗴𝗮𝘁𝗶𝗼𝗻) --- 》𝗦𝘁𝗲𝗽 𝟲: 𝗖ollaboration ✸ Build the Graph (The Workflow). ✸ Define Edges: Connect Ingest → Detect → Classify. ✸ Add Conditional Logic: If no threats found → End. → Tool: LangGraph StateGraph 》𝗦𝘁𝗲𝗽 𝟳: 𝗢perate ✸ Run the initial pass to validate the data flow. ✸ Ensure the graph compiles and agents communicate. → Action: Initial execution & debugging. 》𝗦𝘁𝗲𝗽 𝟴: 𝗠emory ✸ Add persistence so agents share context. ✸ Use a Checkpointer so Agent B knows what Agent A saw. → Tool: LangGraph MemorySaver 》𝗦𝘁𝗲𝗽 𝟵: 𝗣resent & Prompt ✸ Force outputs into readable formats (Markdown). ✸ Refine personas: "You are a Senior Security Analyst." → Technique: System Prompt Engineering 》𝗦𝘁𝗲𝗽 𝟭𝟬: 𝗔ssess ✸ Validate the output against business logic. ✸ Check Agent Confidence Scores against real data. → Metric: Precision/Recall on anomalies. 》𝗦𝘁𝗲𝗽 𝟭𝟭: 𝗦ense (The UI) ✸ Stop working in the terminal. ✸ Build a simple frontend for file uploads. → Tool: Gradio Interface 》𝗦𝘁𝗲𝗽 𝟭𝟮: 𝗦how (Live Dashboard) ✸ Deploy the system. ✸ Allow users to upload logs and see real-time agent collaboration. → Result: A production-ready Cyber Defense Dashboard. --- 🎥𝗦𝗠𝗔𝗥𝗧 𝗖𝗢𝗠𝗣𝗔𝗦𝗦 𝗿𝗼𝗮𝗱𝗺𝗮𝗽 𝗖𝘆𝗯𝗲𝗿 𝗗𝗲𝗳𝗲𝗻𝘀𝗲 𝗦𝘆𝘀𝘁𝗲𝗺: youtube.com/watch?v=p5ms-d_c… 🎥For 9 more Real-World AI Agents Projects (incl. architecture, tools, design) Watch this: youtube.com/watch?v=NcnYOLoC… -- 📥 Skip the trial-and-error of building production AI agents. Watch my free 30-min training get 88 pages of production guides. Join 46,000 AI engineers, architects, and directors. 👉 maryammiradi.com/free-ai-age…

stategraph implementations primarily. plug and play agents and sort of works in prod

🛠️🧭 How to Build AI Agents from Scratch – Even If You’ve Never Done It Before. Using my 𝗦𝗠𝗔𝗥𝗧 𝗖𝗢𝗠𝗣𝗔𝗦𝗦 𝗿𝗼𝗮𝗱𝗺𝗮𝗽 I built 𝗮 𝗖𝘆𝗯𝗲𝗿 𝗗𝗲𝗳𝗲𝗻𝘀𝗲 𝗦𝘆𝘀𝘁𝗲𝗺 in 12 minutes in Langgraph. My Hands-on Tutorial: youtube.com/watch?v=p5ms-d_c… Here is my roadmap so you can build it Yourself: 𝗣𝗵𝗮𝘀𝗲 𝟭: 𝗦𝗠𝗔𝗥𝗧 (The Core) 》𝗦𝘁𝗲𝗽 𝟭: 𝗦etup & Stack ✸ Don't start from zero. Initialize the core requirements. ✸ Choose cost-effective models for high-volume analysis. → Tools: LangGraph, LangChain, OpenAI 》𝗦𝘁𝗲𝗽 𝟮: 𝗠ission ✸ Define the exact goal. We aren't building a chatbot. ✸ Goal: "Ingest logs, detect anomalies, and classify risks." → Outcome: A clear objective for the agent system. 》𝗦𝘁𝗲𝗽 𝟯: 𝗔rchitecture ✸ Map out the team structure (Nodes). ✸ Assign specific roles: Ingest, Detect, Classify, Report. → Concept: Multi-Agent Nodes instead of one massive prompt. 》𝗦𝘁𝗲𝗽 𝟰: 𝗥easoning ✸ Equip agents with "ReAct" (Reason Act) capabilities. ✸ Allow them to think before they decide to use a tool. → Framework: ReAct Agent Construction 》𝗦𝘁𝗲𝗽 𝟱: 𝗧ools ✸ Create Python functions the AI can call upon. ✸ Specific skills: Pattern Detector, Anomaly Detector, Threat Lookup. → Code: @ tool decorated functions For 9 more Real-World AI Agents Projects (incl. architecture, tools, design) Watch this: youtube.com/watch?v=NcnYOLoC… 𝗣𝗵𝗮𝘀𝗲 𝟮: 𝗖𝗢𝗠𝗣𝗔𝗦𝗦 (𝗧𝗵𝗲 𝗡𝗮𝘃𝗶𝗴𝗮𝘁𝗶𝗼𝗻) 》𝗦𝘁𝗲𝗽 𝟲: 𝗖ollaboration ✸ Build the Graph (The Workflow). ✸ Define Edges: Connect Ingest → Detect → Classify. ✸ Add Conditional Logic: If no threats found → End. → Tool: LangGraph StateGraph 》𝗦𝘁𝗲𝗽 𝟳: 𝗢perate ✸ Run the initial pass to validate the data flow. ✸ Ensure the graph compiles and agents communicate. → Action: Initial execution & debugging. 》𝗦𝘁𝗲𝗽 𝟴: 𝗠emory ✸ Add persistence so agents share context. ✸ Use a Checkpointer so Agent B knows what Agent A saw. → Tool: LangGraph MemorySaver 》𝗦𝘁𝗲𝗽 𝟵: 𝗣resent & Prompt ✸ Force outputs into readable formats (Markdown). ✸ Refine personas: "You are a Senior Security Analyst." → Technique: System Prompt Engineering 》𝗦𝘁𝗲𝗽 𝟭𝟬: 𝗔ssess ✸ Validate the output against business logic. ✸ Check Agent Confidence Scores against real data. → Metric: Precision/Recall on anomalies. 》𝗦𝘁𝗲𝗽 𝟭𝟭: 𝗦ense (The UI) ✸ Stop working in the terminal. ✸ Build a simple frontend for file uploads. → Tool: Gradio Interface 》𝗦𝘁𝗲𝗽 𝟭𝟮: 𝗦how (Live Dashboard) ✸ Deploy the system. ✸ Allow users to upload logs and see real-time agent collaboration. → Result: A production-ready Cyber Defense Dashboard. —- ꆛ Join 46,000 AI Agents Builders - free: Zero to Hero: How to Build AI Agents (30-min training) 2 Guides: 88 pages of real-world agent best practices → maryammiradi.com/free-ai-age…

Lets Build a Grok-Powered agentic research assistant with LangGraph — here's the architecture that makes it work. Most agent demos stop at a single LLM call. This one goes deeper: < Inference: Groq's OpenAI-compatible endpoint with `llama-3.3-70b-versatile` — just swap the base URL, no wrapper changes needed < Agent loop: LangGraph `StateGraph` cycling between an `agent` node and a `ToolNode` until no tool calls remain < Sub-agent delegation: the lead agent spawns isolated assistants with scoped tool sets for focused subtasks — keeps the main context lean < Skill-based dispatch: structured `SKILL.md` files define reusable workflows. Agent calls `list_skills` -> `load_skill` before tackling complex tasks < Persistent memory: a flat JSON store handles cross-session fact retention via `remember()` / `recall()` — no vector DB needed for basic continuity < Sandboxed execution: all file I/O and Python execution are path-constrained with explicit escape prevention The graph topology is simple. The real complexity lives in the tools and the system prompt — which is the right place for it. Check out the full Tutorial Article here: marktechpost.com/2026/05/06/… Full notebook available: github.com/Marktechpost/AI-A…

# Raptor Foil Full Simulation – Engineering Nexus (Realistic v6) import context7_mcp as c7 from langgraph import StateGraph def run_full_raptor_foil_simulation(state): material = c7.get("/kist/2026-cnt-bnnt-pdms-shielding") # CFD – Hochpräzise cfd = FluentModel( geometry="raptor_chamber_axial_segment_v6.stl", mesh="poly_hex_32M_cells", turbulence="k_omega_SST", combustion="EDC_methane_LOX_full_chemistry", radiation="DO" ) cfd.add_layer("outer_ceramic", thickness=1.2e-3, material="YSZ_SiC") cfd.add_layer("foil", thickness=0.08e-3, k_axial=material["k_axial"], k_radial=material["k_radial"], anisotropy=True) cfd.add_layer("inner_copper", thickness=3.66e-3, material="CuCrZr") cfd_results = cfd.solve( T_hot_gas=3300, pressure=300e5, transient="full_ignition_to_shutdown_12s_profile", coolant="supercritical_methane" ) # FEA – Struktur Fatigue fea = AbaqusModel(geometry=cfd.export_mesh()) fea.material_model = "hyperelastic_orthotropic_anisotropic" fea.fatigue_model = "Brown_Miller_mean_stress_correction" fea.apply_loads( thermal=cfd_results.temp_profile, pressure=300e5, vibration="random_50_500Hz_12g", thermal_cycling=3000_cycles, transient_thermal_shock=True ) fea_results = fea.solve(analysis_type="fully_coupled_thermo_mechanical_transient") return { "weight_reduction": "51–64 %", "max_foil_temperature": f"{cfd_results.max_foil_temp:.1f} °C", "max_von_mises_stress": f"{fea_results.max_stress:.1f} MPa", "predicted_fatigue_life": f">{fea_results.cycles_to_failure} cycles (95% confidence)", "emi_shielding": f">{calculate_emi_shielding(material)} dB", "neutron_attenuation": f"{calculate_neutron_attenuation(material):.1f} %", "delamination_risk": fea.assess_delamination_risk(), "recommendation": "viable – recommend 0.12 mm ceramic topcoat optimized cooling channel geometry", "critical_risks": ["foil_delamination_at_cycle_\~1800", "foil_thickness_tolerance"], "next_step": "Phase 1 physical prototype testing" }

Terraform の backend って何？という基礎的な解説から運用面での考慮事項などがまとまってる😀最終的にはフラットファイルの terraform.tfstate だと限界だから Stategraph もあるよ！っていう記事 Terraform backends explained: Types, config, and limits stategraph.com/blog/terrafor…

2/25 の Demo Day に参加したけど，今度は 4/22 に Stategraph AI の Demo Day あるとのこと❗️さっそく登録した📅Claude Code Stategraph CLI で Terraform をゴニョゴニョするライブデモ気になる〜

🛠️🧭 How to Build AI Agents from Scratch – Even If You’ve Never Done It Before. Using my 𝗦𝗠𝗔𝗥𝗧 𝗖𝗢𝗠𝗣𝗔𝗦𝗦 𝗿𝗼𝗮𝗱𝗺𝗮𝗽 I built 𝗮 𝗖𝘆𝗯𝗲𝗿 𝗗𝗲𝗳𝗲𝗻𝘀𝗲 𝗦𝘆𝘀𝘁𝗲𝗺 in 12 minutes in Langgraph. My Hands-on Tutorial: youtu.be/p5ms-d_cViU?is=mXoL… Here is my roadmap so you can build it Yourself: 𝗣𝗵𝗮𝘀𝗲 𝟭: 𝗦𝗠𝗔𝗥𝗧 (The Core) 》𝗦𝘁𝗲𝗽 𝟭: 𝗦etup & Stack ✸ Don't start from zero. Initialize the core requirements. ✸ Choose cost-effective models for high-volume analysis. → Tools: LangGraph, LangChain, OpenAI 》𝗦𝘁𝗲𝗽 𝟮: 𝗠ission ✸ Define the exact goal. We aren't building a chatbot. ✸ Goal: "Ingest logs, detect anomalies, and classify risks." → Outcome: A clear objective for the agent system. 》𝗦𝘁𝗲𝗽 𝟯: 𝗔rchitecture ✸ Map out the team structure (Nodes). ✸ Assign specific roles: Ingest, Detect, Classify, Report. → Concept: Multi-Agent Nodes instead of one massive prompt. 》𝗦𝘁𝗲𝗽 𝟰: 𝗥easoning ✸ Equip agents with "ReAct" (Reason Act) capabilities. ✸ Allow them to think before they decide to use a tool. → Framework: ReAct Agent Construction 》𝗦𝘁𝗲𝗽 𝟱: 𝗧ools ✸ Create Python functions the AI can call upon. ✸ Specific skills: Pattern Detector, Anomaly Detector, Threat Lookup. → Code: @ tool decorated functions For 9 more Real-World AI Agents Projects (incl. architecture, tools, design) Watch this: youtu.be/NcnYOLoCmxU?is=MGeU… 𝗣𝗵𝗮𝘀𝗲 𝟮: 𝗖𝗢𝗠𝗣𝗔𝗦𝗦 (𝗧𝗵𝗲 𝗡𝗮𝘃𝗶𝗴𝗮𝘁𝗶𝗼𝗻) 》𝗦𝘁𝗲𝗽 𝟲: 𝗖ollaboration ✸ Build the Graph (The Workflow). ✸ Define Edges: Connect Ingest → Detect → Classify. ✸ Add Conditional Logic: If no threats found → End. → Tool: LangGraph StateGraph 》𝗦𝘁𝗲𝗽 𝟳: 𝗢perate ✸ Run the initial pass to validate the data flow. ✸ Ensure the graph compiles and agents communicate. → Action: Initial execution & debugging. 》𝗦𝘁𝗲𝗽 𝟴: 𝗠emory ✸ Add persistence so agents share context. ✸ Use a Checkpointer so Agent B knows what Agent A saw. → Tool: LangGraph MemorySaver 》𝗦𝘁𝗲𝗽 𝟵: 𝗣resent & Prompt ✸ Force outputs into readable formats (Markdown). ✸ Refine personas: "You are a Senior Security Analyst." → Technique: System Prompt Engineering 》𝗦𝘁𝗲𝗽 𝟭𝟬: 𝗔ssess ✸ Validate the output against business logic. ✸ Check Agent Confidence Scores against real data. → Metric: Precision/Recall on anomalies. 》𝗦𝘁𝗲𝗽 𝟭𝟭: 𝗦ense (The UI) ✸ Stop working in the terminal. ✸ Build a simple frontend for file uploads. → Tool: Gradio Interface 》𝗦𝘁𝗲𝗽 𝟭𝟮: 𝗦how (Live Dashboard) ✸ Deploy the system. ✸ Allow users to upload logs and see real-time agent collaboration. → Result: A production-ready Cyber Defense Dashboard. —- ꆛ Join 46,000 AI Agents Builders - free: Zero to Hero: How to Build AI Agents (30-min training) 2 Guides: 88 pages of real-world agent best practices → maryammiradi.com/free-ai-age…

Stategraph did too: - github.com/stategraph/argus

LangGraphで注目すべき3つのポイント ・StateGraphで複数エージェントの分岐・合流フローを明示的に設計できる ・Human-in-the-loopと永続化チェックポイントで長時間タスクの再開がしやすい ・LangSmith連携で実行トレースを可視化し、障害箇所を数分で特定しやすい PoC止まりのAIエージェントを運用設計まで持っていくなら、この構成はかなり実用的。 #AIエージェント #生成AI