compsci_ai_herk/build n8n ai agents - 8 hr course/_subsections/lesson-01.org

#+title: Lesson 01 | Course Agenda
#+HTML_HEAD: <link rel="stylesheet" type="text/css" href="../../_share/media/css/herk.css" />
#+HTML_HEAD: <link rel="stylesheet" type="text/css" href="../../_share/media/css/org-media-sass/collapsible.css" />
#+HTML_HEAD: <script src="../../_share/media/js/collapsible.js"></script>

* Links
#+attr_html: :class links
- [[../toc.org][TOC | Build n8n ai agents 8 hr course]]
- [[https://www.youtube.com/watch?v=Ey18PDiaAYI][Course Video]]

*** timestamps
:PROPERTIES:
:CUSTOM_ID: timestamp
:END:

#+attr_html: :class playlist
- [[https://www.youtube.com/watch?v=Ey18PDiaAYI][0:00:00 Course Agenda]]
- [[https://www.youtube.com/watch?v=Ey18PDiaAYI&t=103s][0:01:43 AI Agents Simplified]]

* Notes
** course agenda
- no need for code
- walk through step by step builds
- 15 builds

*** course topics
1. ai agents
2. n8n foundations
3. step by step workflows
4. apis & http requests
5. ai agents tools & memory
6. multi agent architecture
7. prompting
8. webhooks
9. mcp servers
   - what it is
   - hosting setup
10. lessons

** what are ai agents
- what is it at its core
- what can it do
- why do we need them

#+caption: what is an ai agent
#+attr_html: :width 600px
[[../../_share/media/img/compsci/ai/herk/build_8hr/ch01_e01.png]]

*** what it does
- take input
- process input
- get output

*** how to use it
- send the output to a tool
  - gmail
  - latex
  - anything

- when we add an LLM to a tool we get
  - ai workflow
  - ai agent

*** workflow vs agent
**** ai workflow
#+caption: ai workflow
#+attr_html: :width 600px
[[../../_share/media/img/compsci/ai/herk/build_8hr/ch01_e02a.png]]

***** pros
- reliability
- cost efficiency
- easier debugging and maintenance
- scalability

***** what we are doing
- input
- tools to process input
- call the LLM
- tool chain
  - process output before final output
- final output

***** practical example
- hubspot tool
  - passes in lead

- perplexity tool
  - does research

- LLM
  - takes research

- send email

**** ai agent
#+caption: ai agent
#+attr_html: :width 600px
[[../../_share/media/img/compsci/ai/herk/build_8hr/ch01_e02b.png]]

***** how it works
- takes input
- has a set of tools at it's disposal
- uses internal logic to decide what to do

***** disadvantages
- not linear
- expensive

***** when to use it
- if the task is unpredicatable

**** anatomy of an ai agent
#+caption: anatomy of an ai agent
#+attr_html: :width 600px
[[../../_share/media/img/compsci/ai/herk/build_8hr/ch01_e02c.png]]

***** outside features
- input
- LLM
- output

***** in the agent
- brain
  - LLM
    - anthropic
    - gemini
  - memory
    - long term
    - short term
    - it won't forget what we need it to remember

- instructions
  - system prompt
    - differnt than input in that it stays the same
    - input changes regularly, ie every interaction you have with chatgpt
  - what is your role
  - what do you do
  - this is what you got to work with