Auto-Analysis¶

Auto-analysis lets Patcherly skip the manual Analyze with AI click. New errors go straight to AI analysis, so you don't have to start analysis on each one yourself. It's a paid plan feature (see the Pricing page), and you turn it on per website/app (target).

Auto-analysis is a separate feature from auto-apply (Patcherly applying a fix without you clicking Approve). They're independent paid features with their own per-target toggles — you can have one ON and the other OFF on the same target. See Auto-apply: when Patcherly applies a fix without your click in Approving Fixes.

How it works¶

By default, each error a connector finds lands in Pending and waits for you to click Analyze with AI before AI looks at it. With auto-analysis turned on, that click is skipped: the connector sends the error to AI as soon as it arrives.

With auto-analysis turned on:

1. The connector finds an error.
2. The error is received and stored.
3. AI analysis starts right away — no click needed.
4. A fix suggestion is generated.
5. Fix approval (this is where auto-apply may take over): if the target also has Auto Apply turned on (separate paid feature, see below), the fix's confidence is high enough, and a few other conditions are met, Patcherly can also skip the "Approve the Fix" step for you (this is auto-apply). If Auto Apply is OFF, or any condition isn't met, the fix waits for your manual approval. See Auto-apply: when Patcherly applies a fix without your click for the full picture.

Without auto-analysis (default on every plan):

1. The connector finds an error.
2. The error is recorded as Pending.
3. You click Analyze with AI in the dashboard.
4. AI analysis starts.
5. A fix suggestion is generated.
6. You review and approve, or reject the patch.

Auto-analysis vs auto-apply at a glance¶

These two are related but separate paid features. Auto-analysis can be ON without auto-apply ever happening — analysis still runs automatically, you just keep manual control over every approve. Auto-apply only kicks in when the target also has auto-analysis ON, because the connector needs an analysed fix to act on.

Read more in Approving Fixes → Auto-apply.

Requirements¶

Auto-analysis needs two things to be active at the same time:

1. A plan that includes it. Check the Pricing page to see which plans include auto-analysis. If yours doesn't, you'll see an upgrade prompt in the dashboard instead of the toggles.
2. The Auto-Analysis toggle turned on for the target. Even on a plan that includes auto-analysis, you turn it on for each website/app one at a time. It's off by default on every new target, so nothing changes until you flip it.

Both must be true. If either is missing, the error waits for you to click Analyze with AI as usual.

Enabling Auto-Analysis¶

From User Profile¶

1. Go to Profile in the sidebar
2. Switch to the AI tab
3. Find the Auto-Analysis Preferences card
4. Toggle auto-analysis On for the targets you want

If your plan doesn't include auto-analysis, you'll see an upgrade prompt instead of the toggles.

From the Targets Page¶

1. Go to Targets in the sidebar
2. Expand a target row (click the chevron)
3. Find the Auto Analysis toggle at the bottom of the expanded area
4. Toggle it On or Off

Both locations control the same setting — changing it in one place updates the other.

Ignored Errors and Auto-Analysis¶

If you previously ignored an error (by error signature and file path), that same error will remain ignored even when auto-analysis is enabled. Auto-analysis respects your ignore rules:

• Error is detected by the connector
• When an error first arrives, Patcherly checks it against your ignored patterns
• If a match is found, the error is set to Ignored status
• Auto-analysis is not triggered for ignored errors
• These pattern-matched rows do not write Business Metrics events (same as pre-analysis Hide)

This prevents noise from known, unimportant errors.

Analysis Details¶

After an error is analyzed (whether manually or via auto-analysis), you can view analysis metadata in the error detail panel:

• Template: Which AI prompt template was used (shown to members with approve permission)
• Target Context Used: Whether target-specific context was included
• Analysis Time: How long the analysis took
• Confidence: AI's confidence in the generated fix

Workspace members without approve permission still see confidence and timing; prompt/template and quarantine diagnostic detail stay limited to approvers.

Analysis Failures¶

If the AI service is temporarily unavailable (rate limits, gateway issues, or platform provider credits), Patcherly retries analysis automatically up to six times using escalating back-off (about 30 s, 2 m, 15 m, 1 h, then up to 6 h between attempts). Automatic retries run only while you are actively using the dashboard (any page, with the tab visible) or while a connector is waiting on analysis via its long-poll session during auto-analyze. If you close the dashboard and no connector is waiting, the retry schedule pauses in the background until you sign in again or a connector reconnects — nothing is lost, and the next attempt resumes from the saved schedule. Connectors wait for the final outcome via the central API without blocking on a single long request (each analysis-wait long-poll holds at most 30 seconds; connectors sleep on retry_after_seconds between calls). If all retries fail, the error moves to Analysis Failed status and you receive one email and in-app notification (if enabled in your profile) with generic copy — no provider or billing details.

Don't want to wait for the next scheduled retry? While an error is waiting between automatic attempts (its status still shows analysing), you can click Retry now on the error row to trigger an immediate attempt without waiting for the back-off timer — useful when you know the AI provider is back online. Doing so starts a fresh retry budget, so if the manual attempt also fails it still gets the full set of automatic retries afterwards.

What happens when analysis fails permanently: - The error is kept with all its original data — nothing is lost. - While automatic retries are still running, the Errors row may show Retrying analysis (n/max) and a short tenant-safe reason (not raw provider details). - You receive one notification (email and in-app, if enabled) when analysis fails permanently — not during intermediate retries.

When analysis finishes without a patch (Not patchable): - Status stays Analyzed with a Not patchable badge — Approve / Reject are hidden. - Durable reason codes (shown as short tenant-safe copy): insufficient_context (could not gather enough source), ai_declined (model would not propose a safe change), or empty_fix (no draft returned). - Approval-needed notifications are not sent for these rows (only when a real draft is ready for review). - Use Mark as manually fixed or Delete (Hide is not available on Not patchable rows). Re-analyze is not offered on Not patchable — Core/Pro may already have run one automatic extra-file follow-up during the same analysis when the model named missing files.

File context during analysis: - First pass (all plans): up to 3 source files from the traceback plus shallow related-path hints (same directory or one parent level). - Core/Pro (advanced_fixes): if that pass is Not patchable for insufficient context and the model names 1–2 extra project-relative files, Patcherly may fetch them once and re-analyze once in the same run. Personal stays single-pass. To retry after a permanent failure: 1. Go to Dashboard → Errors (or the Patcherly plugin Errors screen in WordPress). 2. Filter by the Analysis Failed status. 3. Click Retry analysis on the error row. 4. Analysis starts again with a fresh retry budget (up to six more automatic attempts).

For permanently failed errors, use Hide Error & Ignore or Delete when you do not plan to retry analysis.

Disabling Auto-Analysis¶

To stop auto-analysis for a target, toggle it Off from either the Profile AI tab or the Targets page. New errors for that target will return to requiring a manual Analyze with AI click.

Errors that were already auto-analyzed are not affected — their results remain.

Next Steps¶

• Understanding Errors — Learn about error states and the full lifecycle
• Approving Fixes — How to review and approve generated fixes
• Pricing page — Compare plans and features