March 2025 PDX Qrew Meetup Notes
We started the March meeting with sharing updates on development progress during the previous month. Jen went first by showing a report designed to track progress and work complete on Jobs. The use case is to provide an easily digestible view for anyone with access to the realm. The report looks really nice and provides a lot of information with links to details. The goal of the report is to replace the spreadsheets used in the past. Next Kennedy updated Notifications sent to customers around warranty information. Seeing how others use the features of Quickbase to perform work is interesting. Jim talked about a Pipeline using the Restful API to get a report from Quickbase and then again using the Restful API to UPSERT changes to the table. The Pipeline takes three steps and uses the batch method in place of loops. He also shared progress on using QuNect to UPSERT to a UNIQUE lookup on a child table. Progress came to a halt when it was discovered that QuNect doesn't allow writing to lookup fields. Prior to a recent change by Quickbase to allow mapping tables using a Unique lookup field, this would have been an error. QuNect correctly made certain the builder didn't attempt to write to a lookup field. Now that writing to a lookup is allowed, QuNect is working on a solution and should have an update next week. Elena shared her work with Fast Field forms. She is building a solution for a customer. One problem was found problem with inserting a string value to a field. There is an option on the Fast Field form builder to insert a string literal. The string isn't written to the cell correctly. The workaround is to add a hidden field to the Fast Field form to take the place of the literal and then upsert that to QB. As we discussed Fast Field we moved into a general discussion of sharing our experience with Fast Field. The general talk about Fast Field includes a brief overview of the upcoming OCR feature. We have some interesting use cases to test out and determine if the Fast Field feature is preferred to the other offerings available in the market. During the conversation, Kennedy asked about Fast Field in general as her company doesn't have access. We gave a brief overview of Fast Field and then went over the features and differences between Fast Field and Quickbase. It seems like Jen has the most production level Fast Field experience amongst the PDX Qrew. Her experience in general is the separation between the two platforms causes duplicate work during development. We all acknowledge that while this integration is only a couple years old, improvements are taking place on a regular basis. As the platforms come together we expect improvements in the integration. As the Fast Field conversation died down, we started talking about the benefits and experience of using AI. We talked about how we use AI to write some code. We also shared our requirements & best practices for using AI. The conversation then moved towards an interesting AI use case. Elena shared a code page for designing a gardening layout planner tool. The page uses the Open AI pipeline channel to fill out record data based upon how the garden is drawn by the User. The Code page is a draggable quadrant like a square celled excel spreadsheet. Users draw a garden plan into the quadrant. Then using a header with drop down lists to define plants, soil type, that are in the selected cells on the quadrant. Next the code page generates values to be inserted into a table in Quickbase. Some of the feature ideas were to include suggestions from AI about plant placement, watering, soil all based upon the plant type in the generated records. The User now has a document with facts about the components of the garden design. It's a neat idea! The problem appears when the amount of data becomes overwhelming to the AI. The code page is generating a json array to send to OpenAI and that is where it seems to not work. The problem with the example is the number of key value pairs became too large and the AI was unable to process the data easily and at a reasonable cost. Good feedback from Jacob is to use the image itself or convert the data values into a series of questions and send the questions to open AI to produce the results. The reasoning is that AI is build upon language and may be more effective/ efficient than sending chunks of data. We all were pretty excited by this observation. The meetup began to wind down and we were beginning to say goodbye when Tammie King appeared in the Teams call. Since Tammie snuck in late, she asked about the cool excel like code page. So we talked about it some more. Again, we all were pretty excited by the garden designer code page. Then we talked about hanging out at Empower and then we called the meeting to a close. Jacob recommended the Pocket Pub as he felt it was nice the last time. While at the Pocket Pub we were witness to a storm front rolling over our heads. The news forecast was for tornadoes, lightening. The weather event was windy for about ten minutes. Some of us got dust blown in our beer before we could cover them with our hands. Then it rained and we continued our conversations.
