Agentic AI - A Developer's guide

Samir Sahoo — Wed, 05 Mar 2025 18:46:41 +0000

What Are AI Agents and How Can They Boost Your Productivity?

AI agents are automated systems designed to take actions on your behalf, making them powerful tools for boosting productivity. Unlike ChatGPT, which can generate text-based responses but can't perform tasks like emailing a copy of your conversation or adding events to your calendar, AI agents go a step further. So, what exactly makes an AI system an "agent"?

AI agents have three core capabilities:

Context or Knowledge: Agents have access to relevant information, such as your documents or files, which helps them understand the tasks you need done.
Reasoning and Planning: Powered by advanced models, agents can interpret your requests and plan a series of tasks to achieve specific goals.
Action and Integration: The key difference between a regular AI model and an AI agent is the ability to take action. Agents can interact with other apps, use tools, and execute tasks directly.

This combination of reasoning and action is what gives AI agents their "agency." It’s also known as the ReAct framework—a system that allows agents to both understand your requests and act on them. This distinction is what makes AI agents more than just chatbots and positions them as valuable assistants in managing tasks efficiently.

So, can AI agents help you get more done? Absolutely—by handling routine tasks and integrating seamlessly with your apps, they free up your time for more strategic work.

Ways to Create Your Own Agent

There are several approaches to building your own agent, each offering different levels of flexibility and technical requirements. Here are the main options:

1. Use Pre-Built Agents or Agentic Workflows
Best for: Beginners or those who need quick solutions.
Tools: Platforms like Jasper or ChatGPT offer pre-built agents with integrated capabilities such as content generation and marketing automation.

Pros: Easy to use and quick to set up.
Cons: Limited customization and flexibility.

2. Leverage No-Code or Low-Code Agent Builders
Best for: Users with limited coding skills who want more customization.
Tools: Zapier, Make (formerly Integromat), Bubble, and Microsoft Power Automate.

How it Works:
Use drag-and-drop interfaces to connect apps and create workflows.
Build agents that can automate tasks, process information, and interact across multiple platforms.

Pros: Requires no programming skills, offers flexibility, and supports thousands of apps.
Cons: Complex workflows might require paid plans or additional learning.

3. Custom Agents with APIs (Intermediate Approach)
Best for: Users with basic coding skills who want to extend capabilities.
Tools: Postman for testing APIs, Zapier Webhooks, or writing simple scripts in JavaScript or Python.

How it Works:
Integrate different services using APIs.
Create agents that perform specific tasks, like pulling data from a CRM or sending automated messages.

Pros: More customization than no-code tools.
Cons: Requires some coding knowledge.

4. Build Agents from Scratch (Advanced Approach)
Best for: Developers with coding experience.
Tools:
Languages: Python, JavaScript (Node.js), or Go.
Frameworks: Rasa, LangChain for AI agents, or Flask/Django for web-based agents.
Libraries: OpenAI API for AI capabilities, TensorFlow for ML, or Selenium for web automation.

How it Works:
Code every aspect of the agent—from logic and data handling to API integrations.
Train custom AI models if needed.

Pros: Full control and customization.
Cons: Time-consuming and requires advanced skills.

5. Use AI Platforms for Custom Agents
Best for: Those wanting AI-driven agents without coding from scratch.
Tools:
OpenAI API: For natural language processing and chatbots.
Dialogflow: For conversational agents integrated with Google services.
Microsoft Bot Framework: For building enterprise-level bots.

How it Works:
Train AI models to understand and respond to user input.
Integrate with various platforms like Slack, WhatsApp, or your website.

Pros: Powerful AI capabilities with moderate coding.
Cons: Costs can rise with usage.

Summary: Choosing the Right Approach
For quick and easy setup: Pre-built agents or no-code tools.
For moderate customization: No-code tools with API integration.
For full control and advanced features: Build from scratch or use AI platforms.

In the next post I'll share how to delegate task to your Agent and how we can create our first agent.

The Generative AI Stack

Samir Sahoo — Sun, 02 Mar 2025 17:39:09 +0000

The generative AI tech stack is a comprehensive collection of tools, technologies, and frameworks widely used in building AI systems. It acts as the backbone for developing generative AI, providing essential guidance in transforming theoretical ideas into practical and innovative results.

Generative AI Stack are divided into 4 categories.

Models & APIs
Vector Databases
LLM Frameworks
Deployment

1. Models & APIs

2. Vector Databases

Vector databases provide a new way of structuring data that improves the performance of the database in multiple ways.

In a nutshell, vector databases store the data in the form of high-dimensional vectors which can be achieved using a transformation or an embedding function. These high-dimensional vectors allow for a faster search using a technique called semantic similarity. Instead of relying on conventional approaches such as querying databases using precise matches or predetermined criteria, a vector database allows you to locate data that is most similar or relevant by considering their semantic or contextual significance.

Some examples of vector databases are given below:

Pinecone
Chroma
Qdrant

3. LLM Frameworks

LLM frameworks allow developers to build complex applications using LLMs that can execute a series of tasks and access external tools to complete complex tasks accurately. Specifically, they allow large language models to access other data sources and APIs and allow them to interact with the software’s environment.

As large language models are trained on text data, and as they don’t really have built-in capabilities to solve complex maths problems, we may want to use an external API such as WolframAlpha to solve the maths problem.

Some examples of LLM frameworks are given below:

Langchain
LlamaIndex
Anarchy

4. Deployment

Finally, deploying generative AI applications such that they are scalable. As an example of such deployment infrastructure, the suite of tools provided by Microsoft in its Azure OpenAI Services. This product by Microsoft provides a single place to deploy your application in the cloud and to easily access the ChatGPT API.
Other examples of deployment infrastructure are Vertex AI by Google and HuggingFace Inference Endpoints.

To summarize, the four levels in the generative AI software stack that are given below.

DEV Community: Samir Sahoo