From Zero to Agent: My 2‑Month AI Project with Full Open‑Source Learning Roadmap

Learning Roadmap Overview

The AI/Agent development path is divided into five sequential phases. Each phase builds on the previous one, moving from basic programming skills to a complete, maintainable AI Agent product.

Phase 1 – Programming Foundations

Learn Python syntax, data types, control flow, functions, modules, file I/O, exception handling, and basic OOP.

Practice with small scripts (e.g., accounting app, todo list, text statistics).

Master command‑line operations (cd, dir, python xxx.py, pip install).

Use Git for version control (git init, git clone, git add, git commit, git push, git pull).

Phase 2 – AI Application Basics

Understand what large language models can and cannot do.

Learn prompt engineering: define task goal, input format, output constraints, examples, and step‑by‑step reasoning.

Call model APIs (obtain API key, install SDK, send requests, handle responses and errors).

Produce structured JSON output (fixed fields, classification labels, parameter objects).

Phase 3 – Core Agent Skills

Tool (function) calling – design tool signatures, decide when the model should invoke a tool, and feed results back to the model.

Multi‑step task decomposition – single‑round vs. multi‑round calls, workflow orchestration, sub‑task splitting.

Retrieval‑Augmented Generation (RAG) – document splitting, embedding, vector search, re‑ranking, and injecting retrieved context into prompts.

Memory & state handling – short‑term conversation history, user state, long‑term memory.

Workflow & control flow – conditional branches, loops, approvals, human‑in‑the‑loop, multi‑tool routing.

Phase 4 – Project Development

Build end‑to‑end projects that combine the above capabilities.

Typical projects: command‑line chatbot, text‑processing assistant, tool‑using Agent, knowledge‑base Q&A, web‑based Agent with UI.

Phase 5 – Engineering & Advanced Topics

Evaluation (Evals) – design test sets, compare prompts, record failures, measure tool‑call correctness.

Logging & observability – capture model inputs, selected tools, execution steps, and failure points.

Security & permission control – prevent prompt injection, protect sensitive operations, enforce role‑based access.

Multi‑Agent collaboration – split responsibilities across specialized agents (search, summarization, approval).

Deployment – FastAPI backend, database integration, Docker containerization, cloud server provisioning, environment‑variable management.

Suggested Weekly Schedule (2–3 hours per day)

Weeks 0‑3: Python basics (data types, control flow, functions, files, exceptions, OOP).

Weeks 4‑5: HTTP/JSON basics, call a public API.

Weeks 6‑7: Model API usage and simple prompt writing (chat, summarizer, extractor).

Weeks 8‑9: Tool calling (weather query, calculator, search, database query) and a multi‑step workflow demo.

Weeks 10‑11: RAG pipeline – document splitting, embedding, vector store, retrieval, prompt integration.

Weeks 12‑13: FastAPI backend, integrate RAG and tool calling, add streaming responses and logging.

Weeks 14‑15: Front‑end integration (Vue 3 + Vite + Element Plus), JWT authentication, role‑based UI for chat and knowledge‑base management.

Key Practical Commands

cd

dir

python xxx.py

pip install 包名

git init

git clone

git add

git commit

git push

git pull

Project Portfolio (Technical Highlights)

Project A – Drone Rental System

A B2C drone‑rental platform with user registration, qualification review, order lifecycle, fault reporting, maintenance workflow, and an AI consultant for model selection, rule Q&A, and order status. The backend has been upgraded to Java 21 / Spring Boot 3.5 and includes Spring AI tool‑calling, multi‑channel communication (MCP), AI memory, and optional Python LangChain/LangGraph services. Core business loop: browse devices → qualification → order creation → payment → delivery → usage → return → evaluation.

Project B – AI‑enhanced Cloud Disk (smart‑disk)

An online file‑storage service that integrates RAG and Agent capabilities. Users can upload files, create vector indexes, and interact with an AI assistant for document summarization, multi‑document retrieval, and context‑aware Q&A while preserving original files. Architecture combines a Java Spring backend (authentication, permission, file handling) with a Python FastAPI RAG service (Chroma vector store, OpenAI model calls) and MySQL for business and AI trace data.

Project C – OmniShopAI (e‑commerce with AI chatbot)

A full‑stack e‑commerce system (Vue 3 front‑end, Django + DRF back‑end) with three user roles (customer, merchant, admin). Features include product catalog, shopping cart, order processing, coupons, announcements, and a pluggable AI chatbot for customer support. The backend uses a custom User model, JWT authentication, RBAC (Role/Permission tables), and a dedicated chats app for AI conversation storage and external model integration.

Technical Architecture Summary

Client layer : Vue 3, Vite, Element Plus, Pinia for state, Axios for API calls.

API layer : Spring MVC controllers (or Django REST Framework), JWT interceptor for stateless authentication, admin interceptor for permission checks.

Business layer : Service implementations handling core domain logic (users, devices, orders, coupons, AI orchestration).

Data access layer : MyBatis‑Plus (or Django ORM) with MySQL for persistent storage, soft‑delete and pagination support.

AI orchestration layer : Spring AI ChatClient (or FastAPI) with OpenAI‑compatible provider, RAGFlow fallback, tool‑calling, and optional LangChain integration.

Data storage layer : MySQL 8 for business data, local uploads directory for media files, external RAGFlow service for vector indexes.

Core Skills and Practices

Python programming and command‑line proficiency.

Git workflow for version control.

HTTP/JSON fundamentals and REST API consumption.

Prompt engineering best practices.

Structured JSON output design.

Tool/function calling patterns.

Multi‑step workflow design and execution.

RAG pipeline implementation (splitting, embedding, vector search, re‑ranking).

Memory management (session history, user state).

Workflow control (conditions, loops, approvals, human‑in‑the‑loop).

Evaluation (Evals) and logging for reliability.

Security measures (prompt injection mitigation, permission checks).

Multi‑Agent collaboration strategies.

Deployment using FastAPI, Docker, and cloud servers.

Recommended Resources

Python Beginner’s Guide – https://wiki.python.org/python/BeginnersGuide.html

Official Python Tutorial – https://docs.python.org/3/tutorial/

Git Cheat Sheet – https://git-scm.com/cheat-sheet

FastAPI Tutorial – https://fastapi.tiangolo.com/tutorial/

OpenAI Agents Guide – https://platform.openai.com/docs/guides/agents

OpenAI Agents SDK – https://platform.openai.com/docs/guides/agents-sdk/

OpenAI Agent Evals – https://platform.openai.com/docs/guides/agent-evals

OpenAI Realtime API – https://platform.openai.com/docs/guides/realtime

OpenAI Agent Builder – https://platform.openai.com/docs/guides/agent-builder

Practical Guide to Building Agents – https://cdn.openai.com/business-guides-and-resources/a-practical-guide-to-building-agents.pdf