🤖 Agentic AI Tools

A collection of production-inspired Agentic AI workflows built from scratch

Build real-world AI systems using LangGraph, RAG, Planning, Tool Calling, Preflight Validation, State Management, Synthesizers, and more.

---

Overview

This repository demonstrates how modern Agentic AI systems are actually built.

Instead of isolated examples, each project combines multiple components into complete AI workflows similar to those used in production systems.

Examples include:

• Retrieval-Augmented Generation (RAG)
• Multi-step Planning
• LangGraph State Machines
• Tool Orchestration
• Preflight Validation
• LLM Routing
• Context Management
• Response Synthesis
• Memory
• Vector Search
• Structured Outputs

The goal is to help developers understand how the entire pipeline works together, not just individual libraries.

---

Architecture

User Query
│
▼
Planner
│
├──────────────┐
│              │
▼              ▼
Preflight   Direct Answer
│
▼
Retriever (RAG)
│
▼
Context Builder
│
▼
LLM
│
▼
Synthesizer
│
▼
Final Response

Depending on the workflow, additional nodes may include:

• Tool Executor
• Memory
• Re-ranking
• Query Rewriting
• Guardrails
• Human-in-the-loop

---

Features

✅ Retrieval-Augmented Generation (RAG)

✅ LangGraph Workflows

✅ Planner-Based Execution

✅ Preflight Validation

✅ Context Engineering

✅ Tool Calling

✅ Conditional Routing

✅ State Management

✅ Multi-Step Reasoning

✅ Response Synthesis

✅ Production-Oriented Folder Structure

---

Workflow Example

Question

↓

Preflight Validation

↓

Planner

↓

Should I use tools?

↓

Yes

↓

Which Tools?

↓

Retrieve Documents

↓

Build Context

↓

LLM Generation

↓

Synthesizer

↓

Final Answer

---

Technologies Used

• FastApi
• LangGraph
• OpenAI / Ollama
• pgvector / Pinecone / Chroma
• AI Agents
• Pydantic
• PostgreSQL / Neon DB
• Retrieval Augmented Generation
• LLM Orchestration
• Vector RAG / Graph RAG
• Hybrid Search / Vector Search
• Code Reranker

---

Why This Repository?

Many tutorials show only a single concept.

This repository focuses on combining multiple concepts into complete AI workflows.

For example, instead of showing only RAG, a workflow may include:

Planner → Preflight Validation → Query Rewriting → Retrieval → LLM → Synthesizer

This better reflects how production AI systems are designed.

---

Current Workflows

Workflow	Status
Advanced RAG	✅
LangGraph Planner	✅
Preflight Validation	✅
LLM Router	✅
Response Synthesizer	✅
Multi-step Agent	🚧
Memory	🚧
Reflection	🚧
Human Feedback	🚧

---

Getting Started

Clone the repository

git clone https://github.com/pt413/Agentic-AI-Tools.git

cd Tools

Install dependencies

pip install -r requirements.txt

Configure environment

OPENAI_API_KEY=your_key

Run a workflow

python main.py

---

Learning Path

If you're new to Agentic AI, follow this order:

1. Basic RAG
2. LangGraph Fundamentals
3. Planner
4. Tool Calling
5. Preflight Validation
6. Context Engineering
7. State Management
8. Synthesizer
9. Multi-Agent Workflows

---

Inspiration

This repository is inspired by production AI architectures used across modern LLM applications, rather than toy chatbot examples.

The implementation emphasizes clarity, modularity, and extensibility.

---

Contributing

Contributions are welcome.

If you'd like to add new workflows, improve documentation, or fix issues, feel free to open a Pull Request.

---

License

MIT License

---

⭐ If you found this repository useful, consider starring it.

Happy Building!