readme for agent app sample and main readme updates#56
Conversation
prithvikannan
left a comment
prithvikannan
left a comment
There was a problem hiding this comment.
Thanks a lot @FMurray!! LGTM with a handful of comments
| - [RAG Getting Started](./rag_app_sample_code/README.md) | ||
| - Start with `agent_app_sample_code/A_POC_app` to build a proof of concept | ||
| - Then explore `agent_app_sample_code/B_quality_iteration` to improve quality | ||
| - Uses Databricks' Mosaic AI Agent Framework for enterprise features |
There was a problem hiding this comment.
what does "enterprise features" mean here? can we be more explicit and say something like "agent serving" and "agent eval"
| - **For an agent that uses a retriever tool:** | ||
| - Check out `agent_app_sample_code` | ||
|
|
||
| - **For OpenAI SDK Integration:** |
There was a problem hiding this comment.
| - **For OpenAI SDK Integration:** | |
| - **For Agent Application in pure Python + OpenAI:** |
| ## Repository Structure | ||
|
|
||
| ``` | ||
| ├── agent_app_sample_code/ # Sample code for agent applications | ||
| │ ├── agents/ # Agent code | ||
| │ ├── 03_agent_proof_of_concept.py # Example of a proof of concept agent | ||
| │ └── ... # Additional directories and files | ||
| ├── openai_sdk_agent_app_sample_code/ # Sample code using OpenAI SDK | ||
| │ └── ... # Directories and files | ||
| ├── rag_app_sample_code/ # Sample code for RAG applications | ||
| │ ├── A_POC_app/ # Proof-of-Concept applications | ||
| │ ├── pdf_uc_volume/ # Example of a RAG application using a PDFs | ||
| │ ├── B_quality_iteration/ # Code for quality iteration | ||
| │ └── ... # Additional directories and files | ||
| ├── genai_cookbook/ # Documentation and learning materials | ||
| ├── data/ # Sample data for testing and development | ||
| ├── dev/ # Development tools and scripts | ||
| └── README.md # This README file |
There was a problem hiding this comment.
can we drop this part? seems hard to maintain as we iterate on this repo
|  No newline at end of file | ||
| ## TL;DR: | ||
|
|
||
| Choose the recipe that best matches your needs: |
There was a problem hiding this comment.
can we emphasize the 10 minute demo?
| - Follow the numbered notebooks in sequence | ||
| - Each notebook contains detailed instructions and explanations | ||
|
|
||
| ### Option 2: Running Locally |
There was a problem hiding this comment.
is this fully supported? we dont recommend this in our docs anywhere, so just want to make sure.
There was a problem hiding this comment.
Or it means for running via VS Code but using Databricks' runtime? I feel a bit concerned as we are cloud platform and customers may feel confusing if we suggest run stuff locally.
There was a problem hiding this comment.
That's fair - I think we need to document it, but it might not work consistently across all "recipes". We should probably let each one define their local development workflows if applicable.
There was a problem hiding this comment.
Or it means for running via VS Code but using Databricks' runtime? I feel a bit concerned as we are cloud platform and customers may feel confusing if we suggest run stuff locally.
Yes, this means running in VSCode with databricks connect.
|
|
||
|  | ||
|
|
||
| ## Getting Started |
There was a problem hiding this comment.
maybe worth mentioning prereqs like UC
|
|
||
| ### Prerequisites | ||
|
|
||
| - Databricks Runtime 14.0 or higher |
There was a problem hiding this comment.
| - Databricks Runtime 14.0 or higher | |
| - Databricks Runtime 14.0 or higher or serverless |
There was a problem hiding this comment.
I am not sure if that could be done by purely Serverless. In the AI cookbook documentation, it mentions both https://ai-cookbook.io/nbs/6-implement-overview.html#requirements.
My understanding is the AI Cookbook includes many components and integrates various products, each with its own requirements. Some products, like Mosaic AI vector search, use Serverless runtime, while others rely on the DBR environment. This is why both DBR 14.3 (Non-ML) and Serverless runtime are mentioned. Not sure if that's correct.
There was a problem hiding this comment.
Yeah we should specify 'serverless notebook'
|
|
||
| ```python | ||
| # 1. Configure your agent | ||
| from agents.agent_config import AgentConfig, RetrieverToolConfig, LLMConfig |
There was a problem hiding this comment.
not related to this PR, but i feel importing from agents... for cookbook util code, and then using from databricks_agents... for the actual SDK is confusing.
|
|
||
| The `rag_app_sample_code` directory contains sample code for Retrieval-Augmented Generation (RAG) applications. | ||
|
|
||
| The `genai_cookbook` directory contains a 10 minute getting started guide. |
There was a problem hiding this comment.
would dev and quick_start_demo got a word to say here, or they are good to go?
There was a problem hiding this comment.
I wonder which part this 10 minute demo has been simplified compared to agent_app_sample_code?
There was a problem hiding this comment.
dev is for publishing this as a jupyter book, IMO we need to remove it from this repo
| ### Getting Help | ||
|
|
||
| - For bugs or feature requests, [create an issue](https://github.com/databricks/genai-cookbook/issues) | ||
| - For questions, start a [GitHub Discussion](https://github.com/databricks/genai-cookbook/discussions) |
There was a problem hiding this comment.
nit: the link doesn't work. maybe remove or update the link?
jiayi-wu-3150
left a comment
jiayi-wu-3150
left a comment
There was a problem hiding this comment.
LGTM with comments. Can you take another look at the local running part? I know some code developers may prefer that way. Not sure how we would recommend that.
I ensured that the general instructions are valid but not that every example runs locally. I think it can be left to contributors of to test that independently. |
4 participants
Related Issues/PRs
General chore item
What changes are proposed in this pull request?
Adding and updating readmes incl. getting started, contributing, local development
How is this PR tested?