tableQA
tableQA copied to clipboard
abhijithneilabraham
* Support more dataframes
-Currently the feature supports csv files only. However, integrating more dataframes is easy. Go through the get_dataframe() method in data_utils.py and include support to detect the incoming file and parse the dataframe from it.
It'd be good to just be able to do something like:
agent = Agent(df)
Sure. We will update it asap.
@abhijithneilabraham keep up the great work, looking forward to contributing
Thanks! @rhamnett The repo is updated with the changes you suggested, kindly go through the readme and examples and report any bugs.
Hello @abhijithneilabraham! That's great! I gave it a quick test on a sample dataset i created from my own columns and the basic functionality seems to work!
Some support for date types etc - datetime64 etc would be really good and then getting the model to understand what today/yesterday/this week/last week means from the current date.
agent.query_db("how many website visits were there in the last week?") agent.query_db("how many orders have I had today?") agent.query_db("how many orders have I had for the week beginning 31st August 2020?")
This kind of thing 👍
Makes sense. In fact, I was already building something like that, but the issue is to map the natural language to datetime without approaching it in a rule based way. Let me think about this and will update asap.
Makes sense. In fact, I was already building something like that, but the issue is to map the natural language to datetime without approaching it in a rule based way. Let me think about this and will update asap.
Yes, that's whats going to make it interesting :)
Haven't implemeted multi-table support yet. We need to join tables in that case, it seems. Will plan that for future. Thanks 😊 .
thanks for the quick reply. Then df should be select * from table, right?
That's correct. Download a table as a dataframe and pass it to this.
You could also contribute a feature if you can enable something like agent=Agent(db_url, table_name). Refer agent.py,data_utils.py
Hello @abhijithneilabraham! That's great! I gave it a quick test on a sample dataset i created from my own columns and the basic functionality seems to work!
Some support for date types etc - datetime64 etc would be really good and then getting the model to understand what today/yesterday/this week/last week means from the current date.
agent.query_db("how many website visits were there in the last week?") agent.query_db("how many orders have I had today?") agent.query_db("how many orders have I had for the week beginning 31st August 2020?")
This kind of thing 👍
This has been done too! We have a Date class in column_types.py
Hi Abhijit Have used this tool, it is an excellent tool. Have tested some of my scenarios, they are working fine. Have one query, do joins works in this tool like left join or right join
Hi @dharmesh2002 , Can you be more specific where you have tested a scenario with SQL joins?
Hi @abhijithneilabraham , i haven't tested sql queries, what i need is, to connect sql and than to join with tables. So that's why i was checking with you whether that is feasible?
@abhijithneilabraham - is it possible to add new functions like group by or having clause?
Yes. Can you add support for them?
On Thu, Apr 21, 2022, 7:48 AM dharmesh2002 @.***> wrote:
@abhijithneilabraham https://github.com/abhijithneilabraham - is it possible to add new functions like group by or having clause?
— Reply to this email directly, view it on GitHub https://github.com/abhijithneilabraham/tableQA/issues/23#issuecomment-1104637670, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIOHO46G666EDB72L7LWKMDVGC3FTANCNFSM4RZTADAQ . You are receiving this because you were mentioned.Message ID: @.***>
@abhijithneilabraham sorry I didn’t understood, can u pls help me where to add group by function, I tried adding but it didn’t work
@dharmesh2002 Can you make a pull request with your code and tag it here?
@dharmesh2002 Can you make a pull request with your code and tag it here? hi, @abhijithneilabraham https://github.com/abhijithneilabraham/tableQA/pull/61
@abhijithneilabraham please if you can have a look on pool request we shared above
Hi, I reviewed the code. Have you tried testing it on an actual database and see if the results suit your needs in a natural language query?
On Fri, Apr 22, 2022, 9:27 PM dharmesh2002 @.***> wrote:
@abhijithneilabraham https://github.com/abhijithneilabraham please if you can have a look on pool request we shared above
— Reply to this email directly, view it on GitHub https://github.com/abhijithneilabraham/tableQA/issues/23#issuecomment-1106649895, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIOHO44IC3HVSCJZKISVCWTVGLD6XANCNFSM4RZTADAQ . You are receiving this because you were mentioned.Message ID: @.***>
Sure, will check that. And send you a pr if get anything solid. @TheurgicDuke771 , I am working on a similar use case of connecting the Agent with Postgresql db. Are you able to crack anything for making a connection with db?
