FFF filament Evidence Lab · Playbook 2026.09.21

Turn a failed filament print into a controlled incident record.

This workflow is for filament-extrusion (FFF/FDM) jobs. Mark what is visible, record exact machine and slicer facts, run one bounded test, and choose what to save or hand off. Resin and other processes need their own process-specific workflow.

Reviewed failure patterns

Choose up to three patterns to investigate.

Select FFF observations that resemble the failed print. These remain hypotheses until a controlled test produces evidence.

No patterns selected.
Loose, unfused first-layer lines can be observed; the image cannot establish which setup variable caused them.
Lifted corners can be observed; adhesion, cooling, drafts, material shrinkage, geometry, and process settings still require isolation.
A narrow bottom-layer flare can be observed; Z-axis calibration, bed heat, first-layer flow, geometry compensation, and measurement method require controlled tests.
Hairs between features can be observed; moisture, heat, retraction, travel, and residual pressure can overlap.
Bumps and seam marks can be observed; correlate them with seam or pause positions before naming a cause.
An abrupt X/Y offset can be observed; belts, pulleys, collision, motion settings, cables, and material feed still require inspection.
Separated lines and gaps can be observed; the restricted or miscalibrated source is not visible.
Swollen ridges suggest excess delivered material; measurement and flow calibration are needed before assigning the cause.
A line that thins and stops supports a flow interruption; it does not identify debris, temperature, heat creep, drive, or feed-path cause.
Clean lower layers and later extrusion loss are consistent with heat creep, but other restrictions overlap.
Ripples trailing a sharp edge are consistent with ringing; motion settings and mechanical looseness still need isolation.
Regular horizontal bands can be observed; band spacing and inspection are needed to find a Z-axis or extrusion source.
Unsupported strands can visibly sag; bridge length, flow, cooling, speed, material, and orientation require review.
Collapsed support and underside damage can be observed; fused support is a separate branch that may require a different contact strategy.
Upturned edges can be observed; cooling, temperature, layer time, angle, orientation, and material require comparison.
The thinnest features can be observed failing; the minimum generated feature depends on nozzle, line width, material, orientation, geometry, and slicer behavior.
Excess support and a rough supported face can be observed; the preferred orientation depends on documented load, cosmetics, fit, time, material, process, and machine.
An open seam and missing wall are illustrations. Mesh checks can report open-boundary and non-manifold counts, but no single check proves printability or every topology condition.
Two sizes can illustrate a mismatch; intended real-world dimensions must come from owner or source evidence.
Pits, bubbles, and stringing are consistent with moisture but not unique to it.

Evidence Lab is ready. Open any capability to begin.

Durable incident workspace

Keep this evidence trail recoverable.

Before Save incident, the incident draft exists only in this open page's memory and is lost on reload or navigation. Save incident sends the title, status, playbook identifiers, selected pattern IDs, chronology, four baseline fields, normalized photo-region coordinates, allowlisted slicer values, and calibration plan to the account endpoint. It does not send the selected photo or slicer file bytes or names. A successful signed-in save is associated with the account, carries release/version and hash metadata, and uses a stale-tab conflict check. Save baseline writes to this browser; Save incident also includes those baseline fields. Signed-in outcomes are saved to the selected account incident; guest outcomes stay in the browser.

Browser-only mode. Sign in to attempt an account save for an incident.
Phase 01

Observe

Preserve what can be seen and when it changed before naming a cause.

01
Browser-local photo regionMark the failed area without uploading the image.

The current marking step keeps the selected image in this open page's browser memory. Current incident, draft, export, and journal payloads use normalized region data rather than the image bytes or filename; review a handoff before sending it.

The photo stays in this browser for marking. Sending photos to Print Analysis is temporarily unavailable; use its written-symptom preview instead.

Describe the print problem No photo selected.
02
Symptom chronologyRecord the last known-good point and recent change.
No chronology captured.
Phase 02

Context

Attach only the machine and slicer facts that can be stated exactly.

03
Printer and material baselineSave an owner-approved local profile.

Saving is explicit. Only these four fields are retained locally; no account or server record is created.

No local baseline saved.
04
Slicer receipt importRead an allowlisted subset and cite the source keys.

Parsing is local. The original file, its name, and unknown keys are not retained. Allowed values may include private profile labels; review the displayed values before saving or sharing.

No slicer receipt selected.
Phase 03

Test

Generate one short run with a stop condition, then record the actual outcome.

05
Calibration run packBuild a bounded sequence—not machine-ready G-code.

This pack never generates G-code or overrides printer safety limits. Confirm every value against the machine, material, and manufacturer guidance.

06
Outcome journalRecord improvement without declaring causation.

This button saves the selected outcome in this browser only. Each entry is bound to playbook 2026.09.21. Clear local journal removes browser-only entries, not saved account outcomes.

No outcomes loaded.

Phase 04

Learn

Group owner-recorded incidents and choose evidence for a help draft.

07
Recorded-outcome clustersGroup recorded printer/material/timing combinations.

Clusters group up to 50 outcomes from this browser's journal. Several outcomes can belong to one incident. Counts are not independent incidents or proof of causation.

Add at least one journal outcome to build a cluster.

08
Community evidence handoffChoose exactly which facts enter a help draft.
Include only approved sections

Selected text is copied as entered, not automatically redacted. Remove private details in the source fields or the separate Community review before publishing. Building keeps the draft in this browser; nothing is published.

Review in Community opens a separate editable review flow. It never publishes the question automatically.

Phase 05

Handoff

Export observed facts separately from hypotheses and bind them to reviewed guidance.

09
Provider repair briefExport a non-certifying technician packet.
Include in the provider packet

Selected text is copied as entered, not automatically redacted. Remove private details from the source fields before copying or downloading. Building keeps the brief in this browser; nothing is sent to a provider.

The packet organizes supplied evidence. It is not a repair authorization, safety certification, warranty finding, or proof of root cause.

10
Reviewed playbook releaseVersion and fingerprint the exact guidance in use.
Active release
2026.09.21
Reviewed patterns
20
Content fingerprint
Calculating…
Release authority
Checking server release…
Change set
Evidence Lab: local photo regions, chronology, baseline, slicer receipt, calibration packs, journal, clusters, and reviewed handoffs.

Saved outcomes, calibration sheets, drafts, and briefs name this release so later guidance changes cannot silently rewrite the evidence used at the time.