GenAI Agent Design
· 6 min read
Prioritization / Motivation
- Most important thing is going to be genai in the future ..
- I want to build something mimmally myself ...
Requirements
- As an engineer, I should demonstrate my ability to construct agentic applications with LLMs.
- As an engineer, I should demonstrate the ability to leverage vector dbs to power a RAG for an LLM.
- Non functionals - this should serve as a sample design ...
- Non functionals - I should use this as an opportunity to ramp up on vector dbs ...
- I want my personal-gen-ai-agents to be examples for eventually joining or starting my own gen ai startup
- implement my own rag ...
- Design an LLM + RAG + Vector Db Applications
Use Cases
- Note Companion
- Blog Companion
- Team Doc Companion
- GenAI as a personal note taking companion ...
- GenAI agent focused on organizing notes ...
- Suggests commands to clean up notes ...
- Suggests moving ideas ...
- Discovers static ideas ...
- Groups hyperlinks ...
- Use Case: Blog Author
- I should give chatgpt instructions to ignore slugs, etc
- Can I have it review blog posts / core essence and write the post on my behlaf?
- Append a section at the end of every post to help the LLM generate content?
- Docusaurus llm plugin?
- I should make agentic setup to help me weite blog posts ...
- Use genai to write blog posts ...
- Similar to blogs, I should add a prompts blog with all my genai prompts + scripts to call chatgpt api
- I should make a welcome-to-prompts
- Auto create blog posts form essence
- I should have an offline script call and convert ... outside the blog directory structure for api key security
- I shouldn't need to move my notes around and organize them ... GenAI should be able to do it for me ... or at least tag a note ...
- If it mis-tags it ... I should be able to explain why it tagged it wrong ...
- GenAI as a dev documentation companion ...
- "Having a note taker on the team"
- Team can collaboratively curate notes ... campanion organizes them, cleans them up, etc ...
- Goes from notes / words to a puml diagram ...
- Collaborative Note Taking ...
- There is a whole genAI use case on changing how people take notes ...
- 5 people talked about the same meeting in different ways ...
- Can we find contradictions
- Can we find important things that everyone covered
- Things only 1 person covered ...
- People might be afraid to use to not make them look stupid ...
- Impact
- Helps cut through the ambiguity ...
- There is a whole genAI use case on changing how people take notes ...
Secondary Use Cases
- Interview Companion
- Financial Companion
- Rap Companion
- Teaching Companion
- Other
- Companion to help you prepare for an interview ... or hold one ...
- What interview questions should I ask people?
- What make you special, how are you different than your peers
- Are you self aware
- What tech questions should I ask?
- How would you solve it
- How would you write an X document ...
- What examples would you include in it?
- What points would you make in it?
- Whats its outline?
- What questions would it answer?
- Another agent - analyze financial data / see how much I am spending in different categories ... for different occasions ... etc ...
- Use the LLM to help build an LLM based app ...
- Genai music use voice of football players + deep introspective questions - as a song ....
- Use voice of soccer players
- What were you doing when you were 16 ... Link
- https://www.vocalize.fm/signin
Gen AI to educate ... what are your gaps ... what content should you see ... what concepts must you learn
- What is your strategy of distributing info?
- Instead of finding info ... how does info find you ...
- Are there any agentic use cases that come to mind?
- Vetting a business?
- Interview preparer?
Todos
- Figure out proper use case for agent ...
High Level Design
Components
- Is there a sample chat interface I can build / tinker with?
Essence
- ...
Approach
- iterative enhancements
- Can I run the genAI on my m1 Mac ....
- Free in that case ...
- I just need to get started in the GenAI space
- Use GenAI to develop a GenAI agent
- Using q to craft docker compose files
Resources
- Vector DBs
- AWS Offerings
- Docker Compose
- https://aws.amazon.com/blogs/machine-learning/from-rag-to-fabric-lessons-learned-from-building-real-world-rags-at-genaiic-part-1/
- https://aws.amazon.com/opensearch-service/serverless-vector-engine/
- https://docs.aws.amazon.com/opensearch-service/latest/developerguide/serverless.html
- https://opensearch.org/platform/search/vector-database.html
- https://opensearch.org/docs/latest/search-plugins/vector-search/
- https://opensearch.org/docs/latest/getting-started/quickstart/
- https://playground.opensearch.org/app/home#/
- What GenAI related services does AWS offer?
- https://aws.amazon.com/blogs/database/the-role-of-vector-datastores-in-generative-ai-applications/
- https://aws.amazon.com/opensearch-service/serverless-vector-engine/
- https://docs.aws.amazon.com/opensearch-service/latest/developerguide/serverless.html
- https://aws.amazon.com/blogs/machine-learning/from-rag-to-fabric-lessons-learned-from-building-real-world-rags-at-genaiic-part-1/
- https://aws.amazon.com/blogs/machine-learning/enhance-your-customers-omnichannel-experience-with-amazon-bedrock-and-amazon-lex/
version: "3.8"
services:
qdrant:
image: qdrant/qdrant:latest
ports:
- "6333:6333"
rag-app:
build: .
ports:
- "8000:8000"
depends_on:
- qdrant
- Amazon Q
- Running Locally
- GenAI Agents
- RAGs
- Youtube Videos
- Other GenAI Stuff
- https://huggingface.co/spaces
- https://huggingface.co/google-research-datasets
- https://huggingface.co/datasets/google-research-datasets/natural_questions/viewer/default/train?p=1
- https://huggingface.co/learn/agents-course/unit0/introduction
- https://huggingface.co/docs/transformers/en/agents
- https://huggingface.co/docs/transformers/en/llm_tutorial
- https://www.aimon.ai/posts/deep-dive-into-agentic-llm-frameworks
- https://www.deeplearning.ai/the-batch/agentic-design-patterns-part-5-multi-agent-collaboration/
- https://www.google.com/search?q=course+build+rag+app+vector+db+docker-compose&num=10&sca_esv=a46735a6b9ec06aa&pws=0&sxsrf=AHTn8zphZFXvMrr3L9QmU3WwfWLSHHM9bw%3A1738897076103&ei=tHalZ_OFBpO3wN4PlKCBoQE&ved=0ahUKEwjzr5a0yLCLAxWTG9AFHRRQIBQQ4dUDCBI&uact=5&oq=course+build+rag+app+vector+db+docker-compose&gs_lp=Egxnd3Mtd2l6LXNlcnAiLWNvdXJzZSBidWlsZCByYWcgYXBwIHZlY3RvciBkYiBkb2NrZXItY29tcG9zZTIFECEYoAEyBRAhGKABMgUQIRigATIFECEYoAEyBRAhGKABSJEPUPQBWLAOcAF4AZABAJgBzAGgAbwLqgEGMTMuMS4xuAEDyAEA-AEBmAIQoALgC8ICChAAGLADGNYEGEfCAgUQIRirApgDAIgGAZAGCJIHBjExLjQuMaAHtEA&sclient=gws-wiz-serp
- https://www.google.com/search?num=10&sca_esv=a46735a6b9ec06aa&pws=0&udm=7&sxsrf=AHTn8zoSSP3fG8TYCVyu449HSsM51Ed54Q:1738897080539&q=course+build+rag+app+vector+db+docker-compose&sa=X&ved=2ahUKEwjZjqW2yLCLAxWf8MkDHV1ZDZQQ8ccDegQIDxAH&biw=1792&bih=974&dpr=2
- https://community.openai.com/t/tutorial-a-step-by-step-guide-on-rag-using-surrealdb/701653
- https://forums.developer.nvidia.com/t/dli-course-building-rag-agents-for-llms-building-the-container-for-the-course-setup/306533
- https://qdrant.tech/documentation/quickstart/
- https://python.langchain.com/docs/how_to/
- https://arize.com/blog/understanding-agentic-rag/
- Can I make a simple Mac app ... with apple intelligence ?
- https://github.com/anthropics/anthropic-quickstarts/blob/main/computer-use-demo/README.md
- Deep Dive into LLMs like ChatGPT
- https://qdrant.tech/
Patterns
- What are the different components involved in an GenAI Based architecture ...
Options
Vector Databases
- What are various vector database offerings?
https://opensearch.org/platform/search/vector-database.html https://opensearch.org/docs/latest/search-plugins/vector-search/
