Guide to Agentic AI: The Reasoning + Acting Loop from Scratch (Python & FastAPI)
Here is a complete template for your description. It includes the mandatory links to both the Python and Go implementations and follows the timestamped agenda from your episode.
***
*Description:*
In this episode, we break down **Agentic AI**, an open-source conversational assistant built entirely from first principles using the Reasoning + Acting (ReAct) loop. We look at why building without heavy orchestration frameworks like LangChain can lead to a more robust, decoupled architecture.
We deep dive into the Python (v3) implementation, exploring its three independent FastAPI services, token-budgeted conversational memory, document-aware routing, and how it uses MongoDB as the ultimate source of truth alongside a Pinecone vector layer.
🔗 *Project Links:*
*Python Showcase (Discussed in this video):* https://dharmendrra.github.io/agentic-ai-py/
*Go Version (Equivalent architecture with static typing & goroutine-safe dispatch):* https://dharmendrra.github.io/agentic-ai/v2/
⏱️ *Timestamps:*
*0:00 – Part 1: First-Principles Engineering & Architecture*
Exploring the core while-loop that drives the Thought → Action → Observation sequence, handling API calls to Anthropic Claude or local Ollama models. We also cover the three FastAPI services (Agent, Ingestion, MCP) and why MongoDB is the source of truth.
*4:00 – Part 2: The Agentic AI Framework*
How the system handles runtime routing with source toggles (Web vs. My Library), honest provenance labeling for citations, and "clarify-back" logic to prevent the model from guessing ambiguous queries.
*7:00 – Part 3: System Design & Replicating the Build*
A look into the decoupled vector layer (Pinecone), the book ingestion pipeline (text extraction, chunking, and nomic-embed-text embedding), and testing methodologies.
*9:00 – Part 4: Getting Started*
How to run the architecture locally. Using the idempotent `setup.sh` script to install dependencies, run on Ollama without API keys, and explore the UI.
Note: For developers who prefer Go's static typing and performance, an exact architectural equivalent of this system—featuring the same MongoDB schema and ReAct loop compiled to a single binary—is available via the Go version link above.
On this page of the site you can watch the video online Guide to Agentic AI: The Reasoning + Acting Loop from Scratch (Python & FastAPI) with a duration of hours minute second in good quality, which was uploaded by the user TechGyan w/ AI 11 June 2026, share the link with friends and acquaintances, this video has already been watched 18 times on youtube and it was liked by 0 viewers. Enjoy your viewing!
