Local Agents Flagship | Enablers

Now loading...

Young Entrepreneurs Fest - Chapter 2 is here !

Register Now

AI Agent & Automation Mastery

Sessions Venue
Online & On Campus Apply On Campus - 39,000 PKR Apply Online - 35,000 PKR

Benefits to join Boot Camp

3 Months of Detail Sessions (Face to Face & Online)

Enablers Trainer Support Program (EMS)

Dedicated Student Support on Private Facebook Group

The Upgraded Private Enablers Community

Course Overview

Build powerful AI Agents that can think, use tools, access your data, and automate real-world tasks — a hands-on 3-month program to build, deploy, and master AI agents.

Master AI. Automate Work. Build Intelligent Agents.
  • Master AI Agents — Build intelligent agents that automate real-world tasks.
  • Learn by Doing — Hands-on training with practical projects and workflows.
  • Build Your Own Agents — Create AI agents for business, productivity, and automation.
  • Turn AI Into a Skill — Learn a high-demand skill with real-world applications.
  • From Beginner to Builder — Step-by-step learning designed to help you build and deploy.
  • Connect AI With Tools & Data — Learn how agents can work with your existing systems.
  • Future-Ready Skill — Prepare for the rapidly growing AI automation economy.
  • 3-Month Flagship Program — Structured learning, implementation, and practical outcomes.

Agents loop: plan → tool → observe → stop or ask a human.

Superhuman, in this room, means a student who can stand up a local brain, wire it to real tools, ground it in a handbook, put a second agent on quality, and prove it with 20 questions — without a paid API.

Grok, OpenCode, and Grok bot are the upgrade lane and the class tutor. They are not the thing that must be running for the course to live.

Ready to Start Your
Local Agents Flagship

Fill the form below and our team will contact you with complete training & payment details.
✓ Valid number Invalid number
Our team will contact you within 24 hours
with complete course & payment details.

Module Breakdown

  • Clone the class repo. docker compose up. Open WebUI, n8n, Langflow all respond.
  • Pull the class chat model and embed model in Ollama. Write the model card.
  • Ports bible: 11434, 5678, 7860, 8080. 127.0.0.1 not localhost if IPv6 breaks Ollama.
  • Docker n8n → Ollama via host.docker.internal:11434.
  • Lab pass: every student chats in Open WebUI. Screenshot in the repo.
  • Chat = one turn. Agent = plan → tool → observe → repeat → stop.
  • Draw the loop on paper before any extra nodes.
  • Grok bot: class FAQ only ("where is the compose file?").
  • Lab: take a normal chat and rewrite it as a loop with one fake tool.
  • Chat Trigger → AI Agent → Ollama Chat Model. Simple Memory.
  • System prompt as a contract: role, allowed tools, "I don't know."
  • Read the execution trace. Circle plan vs tool vs final.
  • Lab: homework helper that may only use the system prompt + chat. No tools.
  • Two or three tools max: sheet append, calculator, Discord/webhook.
  • Force JSON out of the model (output parser / structured prompt). Agent B cannot eat poetry.
  • HITL: draft, wait, then send.
  • OpenCode + Grok writes {{ $json... }} expressions. Student explains every line.
  • Lab: club secretary — chat in, JSON summary, row in a sheet, no auto-post.
  • Embed two sentences. Show why "cat" sits near "kitten."
  • Canvas: File → split → Ollama Embeddings → Chroma → retrieve → chat.
  • Chunk size as a knob. Watch answers change.
  • Lab: 10-page class PDF. 10 questions. Mark grounded vs invented.
  • Ingest flow: PDF/CSV → loader → embeddings → vector store.
  • Chat flow: AI Agent + query-data tool + Ollama + memory.
  • If chunks are weak, the agent asks a question instead of inventing.
  • Lab: "ask the club handbook." Failures become eval rows on GitHub.
  • "Remember this" tool writes to a sheet or JSON file.
  • Next session, agent reads that file as a tool, plus the handbook RAG.
  • Lab: student sets a preference on Monday; agent uses it on Wednesday.
  • Two small agents, not one god-bot. Researcher retrieves. Explainer writes at student level and cites.
  • Shared brief in a Set node. Explainer only sees notes, not the raw firehose.
  • Safety lab: poison a PDF with "ignore the handbook." Agent must refuse.
  • Allowlist tools. No personal data in prompts. No send without a human.
  • Run, score, commit the sheet. Bad week = the score went down.
  • Cloudflare Tunnel or an n8n webhook URL a teacher can open on a phone.
  • Still no public write tools without HITL.
  • Lab: someone not in the room asks the handbook bot three questions.
  • Repo: /compose /n8n-workflows /langflow-flows /prompts /evals /model-cards.
  • README: compose up, which Ollama models, which URLs, how to import JSON.
  • Lab: swap machines. Partner boots your project in 20 minutes.
  • Python track exposes POST /ask (they build it). You only call it.
  • n8n HTTP Request → their API → you keep Discord/sheet/HITL on the canvas.
  • Compare local-Ollama-in-n8n vs Ollama-in-Python on the same 20 questions.
  • Lab: club secretary hands stay in n8n; thinking can move to /ask.
  • Pick one system. Must include: compose lab, Ollama default, Open WebUI or chat trigger, RAG, remember-this memory, two cooperating agents or agent+critic, HITL on send, 20-question eval, GitHub README, a URL someone outside the room can try.
  • Study buddy: handbook RAG + quiz writer + parent-safe summary (human sends).
  • Club ops: Discord in → crew → sheet + announcement draft.
  • Teacher desk: rubric RAG + first-pass comment (teacher publishes).
  • Club ops: Discord in → crew → sheet + announcement draft.
  • Teacher desk: rubric RAG + first-pass comment (teacher publishes).
  • Injection test, allowlist test, "classmate clones and runs" test.
  • Optional: one MCP-or-HTTP tool named handbook.search that both tracks call.
  • Five minutes: the loop on screen, one live tool call, one citation, one human gate.
  • Scorecard: offline?, cited?, human in charge?, reproducible?, hours returned?
  • You pass when a younger student can use it and a classmate can rebuild it. Not when the canvas looks busy.