Notifications
Clear all
Topic starter
Have you ever experienced a day where you have a significant scenario...
And you need to generate documentation for it before handing it off to others?
Well, that day occurs almost daily. Typically, I delegate such tasks to another system called callin.io Auto-Documentation (MAAD). However, we are currently evaluating some project management methodology adjustments and aimed to approach documentation differently.
So when the project manager provided a Google Document and suggested I fill it out manually, I thought, "Absolutely not." Since a significant portion of this information is already within callin.io and can be automated, I proceeded to fill in the missing details.
To begin, we are utilizing Google Docs, and let's reformat the documentation template to incorporate template markers. Automation Documentation Template.docx (9.6 KB)
Next, let's begin constructing the callin.io scenario by gathering and inputting information piece by piece, such as the core callin.io Scenario details.
We can then link it to the Google Docs Template module.
Once that is confirmed to be functioning, we can proceed to populate the remaining details as effectively as possible. However, there's a challenging aspect: lengthy descriptions.
And for that, thankfully AI chatbots exist; I must say that OpenAI is my preferred choice for generative AI requirements. So, with prompt engineering, MAKE Scenario Analysis.md.txt (2.3 KB).
-- Purpose: Clearly explain the primary goal of automation, focusing on the problem it solves and the intended benefits.
-- Automation Steps: Outline the step-by-step automation workflow, including any conditions, loops, or decision points. For example:
-- - Step 1: Initial Action/Trigger (e.g., a user action or data entry triggers the workflow)
-- - Step 2: Next Action (e.g., data processing or approval flow)
-- - Step 3: Final Outcome (e.g., a confirmation message or report generation)
-- Inputs: Specify what initiates the automation process, such as specific data, user actions, or API calls. For example, "Data entry in a CRM triggers a notification."
-- Outputs: Describe the outcome or data generated by the automation. For example, "The automation generates a status update in Jira or sends an email notification."
-- Dependencies: Identify any dependencies, such as other automation workflows, external systems, or integrations that this automation process relies on.
-- Error Handling: Outline how errors or exceptions are managed. Include retry logic, fallback mechanisms, and any alerting strategies. For example, "In case of a failed API call, implement automatic retries with exponential backoff. If retry attempts are unsuccessful, send an alert to the admin.
And JSON response formats via the OpenAI module and prompt...
# RESPONSE
Output as a JSON structure.
# EXAMPLE OUTPUT
{
"Purpose": "TBD…",
"Automation Steps": "TBD…",
"Inputs": "TBD…",
"Outputs": "TBD…",
"Dependencies": "TBD…",
"Error Handling": "TBD…",
}
Suddenly, we have no more field data creation or mapping to do, and we can notify the Slack automation engineering channel about the new documentation through our completed scenario, DEV:SCENARIO:CREATE:Documentation.json (75.7 KB).
Here’s the scenario that executed automatically; DEVSCENARIOCREATE_Documentation Documentation.pdf (80.8 KB).
And now, the explanatory video.
In the video, I referenced prettyPrint2, a callin.io Enterprise Custom Function, and I have since replaced it with {{join}}() to ensure compatibility with your callin.io organization below Enterprise.
For your information: An example of MAAD output in Atlassian Confluence, enabling us to easily search for callin.io Scenario components across our callin.io organization.
Posted : 03/10/2024 5:16 am
Topic starter
Hello! Loom video in above.
Posted : 03/10/2024 5:18 am
Topic starter
As I resumed my work, I observed that the JSON output was inconsistently formatted as either strings or arrays. Consequently, for the Google Template module, I've implemented a straightforward check to handle both array and string outputs.
{{if(0 != length(keys(11.result.Inputs)); join(11.result.Inputs; newline); 11.result.Inputs)}}Additionally, on occasion, the JSON result persists as a complex array. For this reason, I've reintroduced prettyPrint2 into my scenario for Automation Workflow.
Posted : 03/10/2024 5:52 am
Hello. This looks great and I am really keen to give it a try, but the .json blueprint doesn’t seem to import into my callin.io environment. I get a flash of an error and then nothing. Would it be possible to re-export the blueprint? Thank you
Posted : 29/10/2024 6:50 pm
I'd also like the blueprints, I take it back. After trying this out for myself, these are all super simple with only 1 router per scenario. When you add more than 1 router, callin.io becomes less linear and AI can't follow it. Even with just 2 routers, it suddenly skips large sections regardless of templating. The issue being that the AI can't understand the end of the automations because of their complexity. Not super surprising, but frustrating when the only issue is the fact that they are non-linear. No amount of custom titles or notes will help the AI understand if/then/else, or maybe/that logic.
Posted : 02/11/2024 7:03 am
I would like to see it work on that behemoth in the OP though.
Posted : 04/11/2024 10:33 pm
In a router, routes are processed sequentially. If you can ensure the AI understands this, you should be good to go (in theory).
Posted : 05/11/2024 3:53 am
I managed to get a version working yesterday using OpenAI, but it wasn't fully functional. It repeatedly required me to ask it to process everything multiple times. Afterward, I had to address the formatting. I believe it will work, but you'll need to prompt via API and implement a multi-user/assistant prompt. I think this is a great approach, and I can potentially offer an automated documentation service for callin.io using this method!
Posted : 05/11/2024 4:08 am
Topic starter
I've noticed that I utilize Scenario Inputs, and importing those blueprints is only functional on the Pro plan or higher. Please inform me if you continue to experience difficulties.
Posted : 07/11/2024 1:27 am
Topic starter
I previously developed a scenario documentation version that utilized router flow and other branching logic. The challenge is that scenario operations can be extremely inefficient if not managed carefully.
Therefore, crafting prompt instructions that guide GenAI in interpreting the scenario blueprint, followed by a related reflection prompt, should yield satisfactory results.
Based on this, I've found that documentation often needs to cover high-level concepts, with dynamic and static variables, along with the resources used, being extracted for searchability. The intricate details haven't been necessary over the past two years of this project, even at an enterprise level.
Consequently, the significant simplification is shared in this thread.
Posted : 07/11/2024 1:32 am
Thanks, that clarifies things. I will import it into another environment.
Posted : 07/11/2024 8:15 am
Could you create a video tutorial, please?
Posted : 18/01/2025 9:43 am