Troubleshooting
Catalyst Narrative Risk: Troubleshooting and Recovery
Diagnose and recover common Catalyst Narrative Risk failures without losing state.
Catalyst Narrative Risk · Troubleshooting
Claims, sources, evidence, narrative mapping, risk dimensions, scoring, uncertainty, cases, comparison, review, approval, governance, and handoffs.
Purpose
This guide uses an evidence-first diagnostic sequence for Catalyst Narrative Risk. Preserve state before changing configuration. Determine whether the failure comes from input data, permissions, version identity, browser assets, backend health, schema compatibility, external dependencies, or a product defect.
Prerequisites
- Exact error message and time
- Product, backend, package, and schema versions
- Affected route, feature, record ID, and user role
- A safe backup or export
- The synthetic sample for a reference test
Sample data
This article uses a fictional Resilient Community Retrofit Program. The files contain one demonstration row for every documented feature, so each capability can be tested without private or production data.
Download CSV feature examples · Download JSON workspace example
Preview
| # | Feature | Sample action | Expected result | Classification |
|---|---|---|---|---|
| 1 | Case Workspace | create NR-RETROFIT-001 | a persistent narrative-risk case | synthetic-demo |
| 2 | Artifact Intake | register a public retrofit announcement with URL, date, version, and custody | a source-preserved communication artifact | synthetic-demo |
| 3 | Claim Decomposition | separate cost, emissions, comfort, and timeline claims | atomic claim records | synthetic-demo |
| 4 | Narrative Mapping | map frames, causal stories, heroes, harms, omissions, and implied action | a structured narrative map | synthetic-demo |
| 5 | Actor and Audience Context | record speaker, institution, intended audience, affected groups, and power context | actor and audience records | synthetic-demo |
The downloadable files include all 16 feature rows.
Procedure
- Stop repeated imports, submissions, migrations, or destructive retries.
- Record URL, feature, action, role, versions, browser/runtime, and expected result.
- Run the same feature row in a new synthetic workspace.
- If the sample fails, inspect health, logs, browser console, network response, and validation details.
- If the sample succeeds, compare real input schema, units, permissions, size, source, and status with the sample.
- Check version and schema identity before clearing caches or reinstalling.
- Apply the smallest reversible change.
- Repeat the sample and real workflow, then document the result.
Common problems
| Symptom | Likely cause | Safe response |
|---|---|---|
| Version identity mismatch | packaged contracts report inconsistent versions | regenerate identity files and run integrity tests |
| Score without evidence | dimension has no cited rationale | add evidence or leave unresolved |
| Overbroad conclusion | case scope does not support the generalization | narrow the claim and state exclusions |
| Sensitive material exposed | private or harmful content entered public workflow | restrict access and follow incident procedure |
Escalation package
- Product and version
- Feature and component
- Exact error text
- Reproduction steps using synthetic data
- Expected and actual result
- Redacted log excerpt without secrets or personal data
- Whether a clean sample reproduces the problem
Expected output
- A reproducible diagnosis or bounded unresolved issue
- Preserved user and system state
- A documented workaround when available
- A support-ready issue summary without private data
Verification
- The synthetic feature succeeds after the fix.
- The original workflow succeeds or remains safely isolated.
- Version and health identity are consistent.
- No duplicates, stale locks, or unintended cache artifacts remain.
- The resolution is reflected in a support article, known issue, or release record when appropriate.
Responsible use
Avoid automated accusation, profiling, or definitive judgments; preserve context, counterevidence, uncertainty, and human review.
Never place credentials, API keys, private URLs, personal records, client documents, or confidential logs in a public article or issue.
