My Tech Learnings

https://platform.claude.com/docs/en/build-with-claude/handling-stop-reasons

https://www.youtube.com/watch?v=ldqOnljDINc

Three Types of Roles

Reference : https://medium.com/@chiwai.kiriba/the-anatomy-of-a-prompt-system-user-and-assistant-roles-d514cbc621ce

User : what the user says
System : the persona you are giving to AI
Assistant : the model’s output. It is what the AI generates in response to the System and User messages

Bottom line :

Each message or prompt has to be associated with a role
Role defines what an entity does

Example

User Prompt ( Ask ) : How is weather
Assistant Prompt (Response) : it is sunny

What is Agent ?

https://huggingface.co/docs/smolagents/en/conceptual_guides/intro_agents

AI Agents are programs where LLM outputs control the workflow

In practice defines an AI solution that involved any of the following

Multiple LLM calls

LLMs with ability to use tools

An environment where LLMs interact

A planner to coordinate activities

Autonomy

Chat Vs Agent

different agentic levels

Anthropic distinguishes two types of Agentic Systems

Workflows are systems where LLMs and tools are orchestrated through predefined code paths
Agents are systems where LLMs dynamically their own process and tool usage maintaining control over how they accomplish their task

✅ When to use agents / ⛔ When to avoid them

Agents are useful when you need an LLM to determine the workflow of an app. But they’re often overkill. The question is: do I really need flexibility in the workflow to efficiently solve the task at hand? If the pre-determined workflow falls short too often, that means you need more flexibility. Let’s take an example: say you’re making an app that handles customer requests on a surfing trip website.

You could know in advance that the requests will belong to either of 2 buckets (based on user choice), and you have a predefined workflow for each of these 2 cases.

Want some knowledge on the trips? ⇒ give them access to a search bar to search your knowledge base
Wants to talk to sales? ⇒ let them type in a contact form.

If that deterministic workflow fits all queries, by all means just code everything! This will give you a 100% reliable system with no risk of error introduced by letting unpredictable LLMs meddle in your workflow. For the sake of simplicity and robustness, it’s advised to regularize towards not using any agentic behaviour.

But what if the workflow can’t be determined that well in advance?

For instance, a user wants to ask: "I can come on Monday, but I forgot my passport so risk being delayed to Wednesday, is it possible to take me and my stuff to surf on Tuesday morning, with cancellation insurance?" This question hinges on many factors, and probably none of the predetermined criteria above will suffice for this request.

If the pre-determined workflow falls short too often, that means you need more flexibility.

That is where an agentic setup helps

Stop_Reason

When calling the Claude Messages API, the stop_reason field in a successful (http 200) response explains why the model finished generating.

You should evaluate this field to determine your next step: accepting the response, continuing the conversation, retrying, or falling back to an alternative model. Note that this field is completely omitted if the request encounters an API error or processing failure.

How to use it:- Check this field to decide if you should use the output as-is, continue the prompt sequence, retry, or switch to a backup model.

Exceptions:- This field is not included if the API request fails or throws an error.