Make agent text-to-sql cookbook #140
Conversation
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
| @@ -0,0 +1,317 @@ | |||
| { | |||
There was a problem hiding this comment.
The link to the notebook leads to a 404.
"...to learn more about agents, start with..."
Maybe briefly recap what the llm_engine is for a gentler introduction.
Reply via ReviewNB
There was a problem hiding this comment.
The 404 is because the notebook is not live yet, but once uploaded on our blog it won't exist anymore.
Just added more detail about llm_engine!
| @@ -0,0 +1,317 @@ | |||
| { | |||
There was a problem hiding this comment.
"Hugging Face Serverless API and dedicated Endpoints"
I think it'd be helpful to add some links to the Serverless API and Endpoint docs.
"The Serverless API provides a built-in HfEngine class that lets you use any model on the Hub via the Serverless API or your own dedicated Endpoint."
Reply via ReviewNB
There was a problem hiding this comment.
also if there's any docs to where someone can create their endpoint would be nice
preferred*
| @@ -0,0 +1,317 @@ | |||
| { | |||
There was a problem hiding this comment.
"We directly pass the initialized HfEngine to the llm_engine init of the agent."
"This callable is the heart of the LLM engine."
I don't think you need to nest those bullet points.
Reply via ReviewNB
| @@ -0,0 +1,317 @@ | |||
| { | |||
There was a problem hiding this comment.
| @@ -0,0 +1,317 @@ | |||
| { | |||
There was a problem hiding this comment.
Change "Anything else" to "Next steps" or something like that to encourage the reader to try something else
Reply via ReviewNB
| @@ -0,0 +1,437 @@ | |||
| { | |||
There was a problem hiding this comment.
Add a link to transformers.agents so curious readers can learn more.
Keep formatting consistent for how we refer to text-to-SQL:
"What's the advantage over a classical text-to-SQL pipeline?"
"A classical text-to-SQL pipeline..."
"An agent system is able to critically inspect outputs and decide if the query needs to be changed or not..."
Reply via ReviewNB
| @@ -0,0 +1,437 @@ | |||
| { | |||
There was a problem hiding this comment.
"...by the agent system. It gives the LLM information about how to use the tool. So this is where we want..."
Reply via ReviewNB
| @@ -0,0 +1,437 @@ | |||
| { | |||
There was a problem hiding this comment.
Maybe also quickly (1-2 sentences) recap the necessary parts of building a tool and then provide a link to the more in-depth doc describing it.
Reply via ReviewNB
There was a problem hiding this comment.
Thank you, I've made an exhaustive list for this!
| @@ -0,0 +1,437 @@ | |||
| { | |||
There was a problem hiding this comment.
Could also probably add a bit more description here about the ReactCodeAgent and llm_engine (for example, you can use another model if you like or another LLM engineer like OpenAI/Anthropic etc.)
Reply via ReviewNB
| @@ -0,0 +1,437 @@ | |||
| { | |||
There was a problem hiding this comment.
|
Thank you for the review @stevhliu. I've modified the cookbooks following the ideas in your comments! 😃 |
What does this PR do?
Add two cookbooks: Agentic text-to-SQL + How to use any LLM with transformers agents.
cc @merveenoyan and @stevhliu.