> ## Documentation Index
> Fetch the complete documentation index at: https://help.1nspect.app/llms.txt
> Use this file to discover all available pages before exploring further.

# Pre-publish checks

> Before a report publishes, 1nspecT runs rule-based checks against the standard of practice. The Narrative Review Panel surfaces every flagged narrative, and Publish is blocked on required items.

Before a report is published, 1nspecT runs rule-based checks (not AI) against the standard of practice the inspection follows. Flags surface in the Report Workspace so you can act on them before Publish. Required items block Publish entirely until resolved. This page describes what the checks do, how to read them, and how to resolve the common ones.

## Where the checks live

Open the Report Workspace: **Inspections** top nav › the order › the inspection › Report Workspace.

The workspace has two tabs across the top:

* **Findings** — every finding on the inspection, with ratings, narratives, photos and the Narrative Review Panel
* **Legal Preamble** — the SoP-required legal preamble

The pre-publish checks surface on the Findings tab.

## The Narrative Review Panel

As you work through the report, the Narrative Review Panel evaluates each narrative against the SoP's writing rules and surfaces any issue:

* Numbers that don't come from captured readings
* Causes stated as certain where the SoP requires hedging
* Jargon that should be plain-language
* Hedging dropped by the AI that your field note included
* Specialist rules for HVAC, electrical, plumbing, site and grading, gas and fire safety
* A passing finding with no Recommendation when your field note asks for an action (see below)
* A narrative written by the backup AI model, whose headline tag needs a second look (see below)
* Headline and movement wording, punctuation, and findings that may belong in the summary (see below)
* Recommendation wording and equipment data (see below)

Each flag links to the finding so you can jump straight to it, fix the wording, and move on.

### Passing finding is missing the action your note asks for

This flag is for review only and never blocks Publish. It appears when a passing finding tagged **Acceptable Performance** has no Recommendation, but your field note asks for an action ("recommend", "replace", "repair", "service", "monitor", "adjust", "clean" and similar). The message reads: *"This passing finding has no Recommendation, but the field note asks for an action. Add the Recommendation the note asks for."*

If the note doesn't ask for an action, the draft gets *"No action is required at this time."* as the Recommendation automatically.

### Drafted by the backup AI

If the main AI model is unavailable when a narrative is drafted, a backup model writes it instead, so the draft still arrives. This is rare. The finding then shows the flag **Drafted by the backup AI — check the Safety / Structural tag**, with the message: *"Drafted by the backup AI model while Gemini was unavailable. Check the headline tag — especially whether it should be a Safety or Structural Deficiency — before publishing."*

The backup model isn't tuned as closely to your Standard of Practice as the main model, so check the headline tag, especially whether it should be a Safety or Structural Deficiency. Fix the tag if it's wrong. The flag is for review only and never blocks Publish.

The flag shows only in the web Report Workspace, not in the mobile app.

### Headline, movement and summary flags

| Flag | Blocks Publish? | What it means |
| - | - | - |
| **Cause named in the headline** | Yes, until you acknowledge it | The headline names a cause, such as "causing…". *"A headline states the condition only — describe each condition, and leave the connection to the Analysis (hedged) or the trade."* "Not causing" isn't flagged. |
| Movement dated as active, ongoing or historical | On a Movement Indicator, or when your field note says "active": review only. Otherwise: yes | A narrative never says whether movement is active, ongoing or historical, even when your field note does. Describe what you observed. The engineer decides whether movement is active. |
| **Doubled period (..)** | No | Two periods in a row, usually from pasting one sentence after another. An ellipsis (…) isn't flagged. |
| **Headline ends in a period** | No | *"Headlines are labels, not sentences — remove it."* |
| **Safety Deficiency not in the summary** | No | A Safety Deficiency isn't in the Deficiency Summary. Add it, or leave it out on purpose. |
| **Movement Indicator not in the summary (at threshold)** | No | The report has reached your Movement Indicator threshold, and this one isn't in the Deficiency Summary. |
| **Non-functional system not in the summary** | No | A system didn't operate, and the finding isn't in the Deficiency Summary. |

A finding gets at most one of the three summary reminders.

### Recommendation and equipment flags

A Recommendation names the trade and one direct action, for example *"A qualified contractor should repair the caulking as needed."* Use "should evaluate … and repair as needed" only when the cause or scope is unknown. None of these flags blocks Publish.

| Flag | What it means |
| - | - |
| **Recommendation adds a purpose or verify step** | The Recommendation ends with a purpose or verification clause. Name the trade and one action. A new AI draft has this clause removed automatically, so the flag mostly appears on text you've edited by hand. |
| **Recommendation lists repair steps** | The Recommendation lists several repair steps. Use one direct verb without describing the method ("should repair … as needed"). This one is never rewritten for you. |
| **Unit has no serial number** | An equipment unit has a capacity or manufacturer recorded but no serial number. A repeated data-plate scan on an older app version could add a unit that isn't there. Remove the extra unit, or add the serial number from the data plate. |
| **Auxiliary heat may have staged** | In heat pump mode, the supply temperature (125°F or more) or the temperature rise (45°F or more) is high enough that auxiliary heat may have come on during the test. The reading is still judged against the 15°F heat pump floor. You can add *"Auxiliary heat staging during the test could not be ruled out."* to the Condition if you choose. |

## Completeness gate

Publish is a hard gate. Before Publish will accept the report, the completeness gate checks:

* Subsection ratings (every subsection that requires a rating has one)
* Required fields. On TREC 7-6 templates these include **Location of water meter** and **Location of main water supply valve**.
* The FPO check
* Every finding has either a narrative, or an explicit "publish without narrative" sign-off

Unresolved narrative violations from the review panel also **block** Publish until you acknowledge them. Flags marked for review only, such as [Passing finding is missing the action your note asks for](#passing-finding-is-missing-the-action-your-note-asks-for) and [Drafted by the backup AI](#drafted-by-the-backup-ai), never block Publish.

If a required field is empty, Publish stops with *"Cannot publish — N required field(s) empty"* and lists each one as **Section → Subsection → field**. Use NP for a subsection that isn't present or doesn't apply.

Any missing items are listed as **publish blockers**. Each blocker shows a **Section → Subsection** label so you know exactly where in the report the missing item belongs. Resolve all of them and Publish will go.

## Findings without a narrative

If a finding has no narrative when you try to Publish, the Publish action refuses with a clear message:

> *"Cannot publish — N finding(s) have no narrative. Draft them (AI Draft & Polish), or choose 'Publish without narrative' for each one in the review panel."*

Two ways to resolve:

1. **Draft the narrative.** Run **AI Draft & Polish** (drafts every eligible finding, then Polish Report) or draft the finding individually.
2. **Publish without narrative.** For a finding that doesn't need a narrative (an observation that's already documented in a related finding, say), open the review panel for that finding and sign off on Publish without narrative. The sign-off is recorded against the caller on the server, so there's an audit trail of who approved publishing without a narrative on which finding.

## Aged-systems check

Separately from narrative review, 1nspecT runs an aged-systems check that looks at equipment age fields (HVAC units, water heaters) and flags any equipment past its typical service life that doesn't have a related finding. This is informational, not a publish blocker: it's a prompt to confirm you intended to publish without flagging the aged equipment.

### Aged Equipment Advisory findings

A finding about aged equipment, such as a water heater or air handler, is headed **Aged Equipment Advisory**. It's informational, not a deficiency. The tag is applied when the caption:

* is "Aged Equipment Advisory";
* says the unit is past, beyond or exceeds its expected, typical or estimated service life;
* or is an age caption such as "Water Heater Age — 14", which the phone suggests when an age field is above your template's range.

The Recommendation is always: *"Budget for replacement of the \[unit]. A licensed plumber (or HVAC technician, or qualified contractor) should perform routine maintenance in the interim."* Aged piping stays Informational.

## Preview PDF before publishing

Use **Preview PDF** to generate the PDF without locking the inspection. It warns you upfront if any publish blockers exist, but it still generates the PDF so you can walk the client through a draft, catch anything the review panel missed, or review layout. The real Publish action is blocked until the required items are resolved.

(Not to be confused with **Publish Report**, which does publish the report to the client and locks that version.)

## What Publish actually does

Once every required item is resolved, Publish:

* Locks the inspection (no more edits without a Republish)
* Uploads the final PDF to secure storage
* Generates the client's permanent report URL slug and the agent's view
* Saves a version record for this publish (version number and publish time)
* Fires the matching workflow trigger (`report.published` when the order is paid, which sends the configured delivery emails, or the payment-pending variant when it is unpaid)

See [Publishing and delivery](https://help.1nspect.app/1nspect/workflows/publishing-and-delivery.md) for what happens after Publish.

## Related

* [Publishing and delivery](https://help.1nspect.app/1nspect/workflows/publishing-and-delivery.md)
* [AI features](https://help.1nspect.app/1nspect/features/ai-features.md)
* [AI Credits](https://help.1nspect.app/1nspect/features/it-tokens.md) (polish, spell check, bundle costs)
* [The order lifecycle](https://help.1nspect.app/1nspect/workflows/booking-to-published-report.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.