Sql context

PPREDATOR

I'm trying to add context to a structured table using the same method as shown below, but it doesn't seem to work. Has anyone successfully injected context to a sql table?

Plain Text

# manually set context text
city_stats_text = (
    "This table gives information regarding the population and country of a given city.\n"
    "The user will query with codewords, where 'foo' corresponds to population and 'bar'"
    "corresponds to city."
)

from llama_index.objects import SQLTableNodeMapping, SQLTableSchema

table_node_mapping = SQLTableNodeMapping(sql_database)
table_schema_objs = [
    (SQLTableSchema(table_name="city_stats", context_str=city_stats_text))
]

Sources:
https://github.com/jerryjliu/llama_index/blob/main/docs/examples/index_structs/struct_indices/SQLIndexDemo.ipynb
https://gpt-index.readthedocs.io/en/latest/end_to_end_tutorials/structured_data/sql_guide.htm

14 comments

LLogan M

How do you know it doesn't work?

LLogan M

Is there an error?

PPREDATOR

I tried running a new query in the same ipynb, it was not providing any response as per the context.

LLogan M

So that text is only used for embeddings and retirveing the tables actually

LLogan M

Lol wait I'm wrong again, I have not touched any of this code before

PPREDATOR

But "the user will query with codeworss, where 'foo' corresponds to population" what is it for ?

LLogan M

Tbh I have no idea. Seems like a really strange example notebook

LLogan M

If you turn on debug logs, you can see the inputs the LLM inputs are

Plain Text

import openai
openai.log = "debug"

PPREDATOR

Thanks, I will check.

PPREDATOR

If we do not manually add context how does llm understand the table schema?

PPREDATOR

Is there anyway to check what type of data/information it stores about the table?

LLogan M

There is a step that pulls the sql schema for the table automtaically

PPREDATOR

@Logan M my usecase is to create a chatbot for large datasets, and i would like to add context for each table.

What approach should I take?

PPREDATOR

For example: I have an ecommerce dataset, and I would like to explain how to perform calculations such as gross margin, LTV, calculating new customers etcs, all based in the columns in the table

Add a reply

Find answers from the community

Sql context