Has anyone figured out how to use LLM's on Airtables?
André J
11 replies
Or using any kind of LLM on structured tabular data. Google sheets.
What I would like to do is use perplexity to filter some structured tabular data.
Any suggestions?
Replies
Greg Z@gregz
A common practice is to employ the seq2seq approach, interpreting user input in a Natural Language Interface (NLI) and converting it into SQL commands. These commands can access specific columns and filter or aggregate them as necessary within your structured tabular data source.
The requests in NLI can also be converted into other types of API requests, for instance, GraphQL. Therefore, it essentially depends on the prompt you're using or the type of fine-tuned Large Language Model (LLM) you prefer for more in-depth analysis. You can initiate a prototype with OpenAI GPT-4 to convert NLI into API requests or SQL queries. Alternatively, you can select any appropriate models using LLM Explorer (https://llm.extractum.io/list/?q...) and run it locally either as the original model or after fine-tuning it as per your needs.
Share
Launching soon!
@gregz Im thinking more general purpose here. Something my mother could do. Go to website. Enter question, enter url to excel sheet. LLMs + Database files like AirTable would be so powerful. Like how hard can it be to just chat with this file?: https://www.airtable.com/univers...
@sentry_co Indeed, that's a great concept. Just consider, that the term "no code" always implies that someone has expended considerable effort on coding so that you don't have to. Thus, in order to let your mom work with tabular data, that integration needs to be set up. The more time and effort you invest in this process, the easier it'll be for your users. Simultaneously, if the implementation is simplified, it's likely to be less functional and more complicated. It really is that straightforward.
Launching soon!
@gregz Right, But if I go to an Airtable. Download the .csv and upload It to a GitHub gist. Then I'm able to chat with the data via perplexity by just pointing perplexity to the .csv file on GitHub gist. It's just that AirTable blocks LLMs from working on them for some reason. Probably they want this market for them selfs. 🤷
André, integrating LLMs with Airtable or Google Sheets can be achieved through custom scripts or API connections that allow the LLM to process and analyze your tabular data. For filtering based on perplexity, you might consider using a script in Google Sheets that interacts with the LLM's API to score and sort your data accordingly.
Integrating LLMs with Airtable or Google Sheets can be achieved through their respective APIs, André. For filtering tabular data using perplexity, consider creating a custom script that uses an LLM's API to analyze and sort your data based on the calculated perplexity scores.
Launching soon!
@jamin_nanthan Appreciate the suggestion. But trying to use their API and your sent down a rabbit hole you can't get out. 😅 This should be as easy as googling.
Launching soon!
@sawyer_wyatt But you do want this right? Like its going to be a thing?
André, integrating LLMs with Airtable or Google Sheets can be achieved through their respective APIs, where you can run scripts to analyze tabular data with perplexity measures. Consider creating a custom script or using automation platforms like Zapier to bridge the LLM's capabilities with your structured data.
Use LLMs on Airtable by integrating with external APIs or custom scripts to leverage their language capabilities for data processing and analysis.
https://www.chemicalbull.com/pro...