Sync Issue Between Airtable AI Fields and Noloco

Bugs
1

1. Background & Context

Rutger Hofste reported an issue with syncing AI-generated fields from Airtable into Noloco. Specifically, the values from Airtable’s AI fields are not being correctly reflected in Noloco after modifications to prompts in Airtable. The application involved is hosted at hidden

2. Initial Troubleshooting

Rutger attempted the following before reaching out:

  • Performed a manual schema sync via Noloco.
  • Triggered table syncs within Noloco’s interface.

Observations:

  • Partial success: Newly generated field values do show up in Noloco.
  • Failure scenario: When a prompt is updated in Airtable without regenerating the field value, Noloco removes the previous value from the synced table.

This is problematic because updating a prompt in Airtable does not clear the value in the cell — meaning the data still exists on the Airtable side.

3. Technical Analysis

Airtable Behavior

  • AI fields in Airtable retain their values even if prompts are updated.
  • Values may be marked as "isStale": true to indicate outdated information, but the value is still available via API.

Noloco Behavior

  • When a prompt is updated without re-running the AI generation, Noloco deletes the existing value from its sync.
  • This discrepancy leads to data loss or visual inconsistency within Noloco, despite Airtable retaining the information.

4. User Expectations vs. Actual Behavior

Aspect Expected Behavior Actual Behavior
Sync logic Retain previous value (even if stale) Deletes value if AI field isn’t regenerated
Staleness handling Display with a “stale” tag or flag Removes value completely
Prompt update Should not trigger removal Triggers null sync if not regenerated
Alignment with Airtable Mimic Airtable’s data retention Discards outdated data

5. Use Cases Shared

Rutger outlined several use cases supporting retention of stale data:

  • AI responses that are “good enough” don’t always require regeneration.
  • Prompt changes are used to fine-tune formatting (e.g., from "1.000 euros" to "1,000€").
  • AI randomness means regenerating the same prompt/input combo can yield different results.
  • Users may want to compare old and new outputs.

6. Proposed Solution

Feature Request Summary:

  • Allow Noloco to preserve stale AI field values from Airtable instead of deleting them during sync.
  • Use isStale: true as a signal to retain but not regenerate.
  • Possibly map AI field data into a specialized object or field type in Noloco (rather than plain long text) to maintain metadata such as staleness.

Rutger provided supporting API documentation:
:paperclip: Airtable API – AI Text Field Model

Example from Airtable API:

{
 "Name": "Rutger",
 "Notes": {
   "state": "generated",
   "value": "Why did Rutger bring a ladder to the low-code conference? Because he heard the solutions were on another level!",
   "isStale": true
 }
}

Even with "isStale": true, the "value" is present and should be retained.

7. Additional Suggestions

Domhnall suggested a workaround:

  • Create a separate text field in Airtable to cache the AI output value before syncing.
  • However, Rutger noted that this should ideally be handled on Noloco’s end, since the sync relationship is maintained there.

8. Multimedia Reference

Rutger also provided a video walkthrough of the issue:
:movie_camera: Watch Video

9. Conclusion

This is a critical sync behavior misalignment between Airtable and Noloco, particularly with the rise of AI-generated content workflows. While Airtable’s API provides access to stale values, Noloco currently deletes them, causing potential data loss or confusion.

Recommendation: Noloco should consider updating their sync engine to:

  • Detect isStale flags,
  • Retain values marked as stale,
  • Possibly introduce a new metadata-aware AI field type for enhanced compatibility.