Projects Troubleshooting & Gotchas

This page is the field-tested list of things that surprise people the first few times they use Projects, and what to do about each one. If you hit something that is not covered here, contact hello@jetstack.ai — we add patterns to this page as we see them.

Access and Visibility

I don’t see the Projects tab

Projects is in limited release and is disabled by default. Three things can be true:

Your organisation is not enabled. Email hello@jetstack.ai from your sign-in address with your organisation name. See Getting Access for the full request flow.
You are signed into the wrong organisation. If your account has multiple workspaces and only one is enabled, switch using the workspace selector in the top-left of the dashboard (or the header in the extension).
Your browser cached the previous state. Hard-refresh (Cmd/Ctrl-Shift-R) the dashboard or close and reopen the extension. The gate is read on first organisation load.

I had access yesterday but the tab is gone now

Limited-release organisations can be temporarily flagged off during incident response or a rollback. Existing projects are not deleted — they reappear when access is restored. Contact us if you need confirmation or an ETA.

The dashboard shows Projects but the extension doesn’t

The extension checks for access when it loads and when you switch organisations. If the two surfaces disagree, the extension is showing a cached state.

Fix: in the extension, switch organisations and switch back. That refreshes the access check and the Projects tab will appear if your organisation is enabled.

Context Gathering

Jetty keeps asking for more context

The context phase ends when Jetty believes it has enough information to design a usable blueprint. If the conversation keeps going, it usually means one of:

You described the desired outcome but not the inputs. Jetty has the “what” but is missing the “from what.” Tell it about the sources of records, the team’s current workflow, or the data already in the portal.
You described the inputs but not the outcomes. Symmetrically: Jetty has the “from what” but no measurable goals to drive the design.
The scope is very large. A project that spans Sales, Service, and Marketing motions requires more context than one focused on a single area.

If you want to move on regardless, ask explicitly: “produce the blueprint with what you have so far.” Jetty will advance and flag any sections that are under-specified.

Uploaded documents are not appearing in the chat

After dragging a document into the chat, the chat composer shows it as an attached file before you send. Make sure to send the message — until you do, the document is staged but not actually uploaded.

Once sent, Jetty shows a confirmation that the document has been parsed and lists the file’s word count and detected sections. If you do not see that confirmation, the upload failed — usually because the file is over 25 MB or in an unsupported format. Re-export to PDF/DOCX/TXT/Markdown and try again.

Blueprint Design

Jetty proposes a property name HubSpot reserves

HubSpot reserves all property names starting with hs_ (and some other prefixes) for built-in fields. Jetty’s validator knows the reserved set and rejects these names during generation, but if it sneaks one through during blueprint design, you will see a deployment failure later.

Fix: rename the offending property in the Blueprint tab before approving. Use a domain-specific prefix instead — acme_lead_source rather than hs_lead_source.

The blueprint feels generic

Jetty’s blueprints are sharper when the context phase contained specifics: named outcomes, real team-member roles, actual data sources, current pain points with examples. Generic context produces generic blueprints.

Fix: go back to chat and add specifics. Then ask Jetty to regenerate the blueprint with the new context. Often a single concrete example — “we lose deals when SDRs forget to log demo-no-shows” — reshapes an entire section of the design.

I changed the blueprint but the Assets tab still shows the old assets

The Assets tab shows whatever was last generated. Blueprint edits do not automatically trigger regeneration; they only update the document.

To regenerate after a blueprint change:

Look at the Blueprint tab toolbar — it will show “Changes pending re-approval”
Click Approve Changes
Jetty re-runs generation for the affected assets and updates the Assets tab

Asset Generation

A specific asset keeps failing to generate

If the same asset fails across multiple regeneration attempts, the issue is usually in the blueprint or its dependencies rather than the asset itself. Check:

The asset’s upstream dependencies. A workflow that needs the Lead Stage property cannot be generated successfully if the property is itself in error.
The asset’s blueprint definition. If the blueprint description is “create a workflow for lead handling” without a trigger or outcome, there is nothing concrete for Jetty to produce.
Naming conflicts. Two assets with similar names (a list and a workflow both called Lead Routing) can confuse references downstream.

Fix the upstream issue first, then regenerate the failed asset. If the failure pattern is hard to diagnose, ask Jetty in chat: “explain why the Lead Routing workflow keeps failing to generate.” Jetty has access to the full generation history and can usually point at the root cause.

Generation feels slow when iterating

A few habits keep iteration fast and predictable:

Edit, don’t regenerate. Most changes can be made directly in the asset’s detail panel (rename, change a stage probability, add a filter) without re-running generation.
Regenerate a single asset rather than the whole project. Click Regenerate on the specific asset’s detail panel — only that asset is redrafted.
Avoid blueprint rewrites in mid-stream. Big blueprint changes cascade into many asset regenerations. If you know a major rewrite is coming, finish it before approving rather than approving partial state and iterating.

Generation never finishes

Generation legitimately takes 15–25 minutes for a typical project and can stretch to 30–40 minutes for large ones — see Asset Generation — How Long Does It Take?. If generation has been “in progress” for noticeably longer than the expected range for your project size (say, more than 45 minutes on a small/medium project), it may be stuck. This is rare, but it can happen if a transient AI service issue did not surface as an error status.

Recovery: open the Assets tab and look at the progress strip. If it has not advanced in 5–10 minutes, click Cancel Generation in the strip. Generation stops cleanly and the affected assets revert to whatever state they were in before it started. You can then re-trigger generation from the Blueprint tab.

Deployment

The deploy keeps failing on the same asset

Check the asset’s error message in the Deployment tab. Common patterns:

Reserved property name — rename the property in the Assets tab
Property type mismatch on existing field — the destination already has a property with the same internal name but a different type; either rename or delete the conflicting destination property
Pipeline stage probability sum != 1.0 — regenerate the pipeline or fix the probabilities manually
Workflow references unknown list — the referenced list failed to deploy; fix that first

See Deployment Phase — Retrying After Failures for the full table.

The extension says “no pending deploy” but the dashboard says one is queued

This usually means the extension is open on a different HubSpot portal than the project’s destination. The extension only shows pending deploys for the active portal’s project queue.

Fix: in HubSpot, switch to the same portal the project is configured to deploy to (check the destination portal in the Deployment tab). The extension banner appears immediately.

Deployed assets are missing in HubSpot

A few causes, in decreasing order of likelihood:

You are looking in the wrong HubSpot portal. Confirm the destination portal is the one you have open in HubSpot.
Marketing assets need an active Marketing Hub. Emails, marketing forms, marketing dashboards, and similar assets require the destination portal to have an active Marketing Hub subscription.
Visibility rules. Reports, dashboards, and CRM cards are scoped to the deploying user by default. Open them in HubSpot and share them with other users or teams.
Workflows are paused by default. Open the workflow in HubSpot and turn it on if you want it to start enrolling records.

A redeploy created a duplicate instead of updating in place

This happens when the project loses its link to the asset in HubSpot — usually because the destination portal’s copy was deleted and recreated externally between deploys.

Fix: in the Assets tab, open the asset’s detail panel and use Reset deployment link. The next deploy creates a fresh asset rather than trying to update the now-missing one.

If you have duplicate assets in HubSpot, the originals from earlier deploys are not automatically cleaned up. Delete the duplicates by hand in HubSpot.

AI Service Errors

”AI service unavailable” during generation

This usually means the underlying AI service is rate-limiting or briefly unavailable. The Projects workspace is designed to be resilient to these — assets that hit a transient error are surfaced so you can retry them rather than failing the whole project.

Fix: wait about a minute and retry the failed assets. If you see the same error consistently across multiple retries, contact us.

Pricing

Projects is included with your existing JetStack AI subscription during limited release — there is no separate metered billing for generation or deployment. When Projects exits limited release, any pricing changes will be communicated in advance.

Reporting an Issue

If you hit a failure that is not covered here, or a behaviour that does not match this documentation:

Capture the project ID from the URL of the project workspace.
Note the time — we use this to find the relevant log entries.
Take a screenshot of the error message or the unexpected state.
Email hello@jetstack.ai with the above and a short description of what you expected to happen vs what actually happened.

Issues that include a project ID and a timestamp are usually resolved within one business day.

Next Steps

Projects Overview — Recap of the four-phase pipeline
Getting Access — Limited-release request flow
Quickstart — End-to-end first project walkthrough