Custom Passable variables in Taskade
khanocp
I need to create a flow where , I can pass few variables from a an input form to Taskade, where taskade can pass it to my desired LLM, and Taskade should also be able to hold the result in another temp memory area where the same result i need to pass to LLM again and the result should be displayed in some output format. Basically I want to create a flow where i provide my input(s) to Taskade , then Taskade sends it to LLM and Back to Taskade and again to LLM and then output in Taskade or sent to Google sheet.
Log In
r
ryantaskade
Hi Khan,
To clarify, we do support referencing variables in our automation, so you can use the @ symbol to pass information from one action to another, more information on our automations can be found here:
Do let me know if i misunderstood you.
Best Regards,
khanocp
ryantaskade thank you for guidance, i create my initial flow successfully. But there are few problems.
1. I am unable to select the LLM version or Provider (in some cases i may need BYOK).
- I am unable to process an image in the file section of the variables. see attached.
- What if instead of few variables for single run, I need to run in a Batch using plenty of different variable sets. I mean is there any table concept where i can store my all variable in the beginning and let the workflow run in Batch.
- Sometime i may need to store my data in a storage somewhere instead of taking it to Sheets or docs. Is there any option to do it?
r
ryantaskade
khanocp: Hi there, thank you for the clarification.
- Yes, we don't support BYOK or LLM selection for the automation actions, in the case of agents, they will use the one set for the respective agent in the Edit agent modal.
- Yes, we currently don't have support for OCR with images, you can only do so with PDFs at the moment.
- and 4. I don't quite understand what you mean by storing variables for a batch run, if you want to run an automation over multiple tasks then you need to store the output for each run somewhere, for example in Taskade project, after that in another automation, you can extract the information from this project use the information there for a "batch run"
We don't support storing multiple variables over a series of runs without saving it in a specific location like a file or Taskade project.
Do let me know if i misunderstood you.
khanocp
ryantaskade Thank you.. 3 and 4: Let me try to explain it again, In many scenarios, you might want to run the same task multiple times but with different input values each time. For example:
Running a report for multiple regions.
Processing different datasets using the same pipeline.
Testing software with various input parameters.
Instead of manually changing the variables each time, it’s more efficient to store all these variable sets in a structured format like a table or spreadsheet. The workflow can then read this table and execute the process once for each row (each set of variables).
Example
Imagine you have a simple workflow to generate learning objectives or questions based on Bloom's Taxonomy for various combinations of:
Grade (e.g., Grade 3, Grade 4)
Subject (e.g., Math, Science)
Topic (e.g., Fractions, Plants)
Bloom’s Taxonomy Workflow: Single Run vs Batch Run
- Single Run Mode
You provide one set of inputs:
Grade: Grade 3
Subject: Math
Topic: Fractions
The workflow generates Bloom’s Taxonomy-based learning objectives or questions for this single combination only.
Example output for single run:
Bloom Level Learning Objective
Remember List different types of fractions.
Understand Explain what a fraction represents.
Apply Solve addition problems involving fractions.
Analyze Compare proper and improper fractions.
Evaluate Judge the correctness of fraction solutions.
Create Design your own fraction word problem.
Continuing to next comment due to chaaracter limitations...
- Batch Run Mode
You provide multiple sets of inputs stored in a table:
Grade Subject Topic
khanocp
Continued from previouse comment due to chaaracter limitations...
- Batch Run Mode
You provide multiple sets of inputs stored in a table:
Grade Subject Topic
Grade 3 Math Fractions
Grade 4 Science Plants
Grade 3 Math Decimals
The workflow automatically processes each row one by one (or in parallel), generating learning objectives for each combination.
Example outputs for batch run:
For Grade 3, Math, Fractions → Generates objectives like in the single run example.
For Grade 4, Science, Plants → Generates objectives related to plant biology.
For Grade 3, Math, Decimals → Generates objectives on understanding and applying decimals.
Key Difference
Single Run: You manually input one set of variables and get one output.
Batch Run: You feed the workflow a table containing many variable sets, and it runs the process repeatedly for each set automatically.
Why Batch Run?
Saves time by automating multiple executions.
Ensures consistent output for many combinations.
Useful when you have large data or many topics to cover.
I guess we might need some cutomizable table structure inside taskade like this or you may have some other type of its implementation.
Table Structure for Batch Processing Bloom’s Taxonomy Inputs
ID Grade Subject Topic Status ResultLink
1 Grade 3 Math Fractions Pending
2 Grade 4 Science Plants Pending
3 Grade 3 Math Decimals Pending
4 Grade 5 Science Ecosystems Pending