Support
How to run the package, what usually goes wrong, and how to report a problem so it can be reproduced.
A — Using the package
- 1 — Name the workflowSay which single workflow the agent would act in, and who is accountable for it. One workflow at a time produces a usable draft; several at once produces a vague one.
- 2 — Supply the mandate fieldsGive the monthly ceiling in EUR, the per-action ceiling, the actions in scope and the review threshold. Where you do not know a value, say so and leave it blank.
- 3 — State provenanceEach supplied value is recorded with a supporting quote or source location. A value without supporting evidence is flagged for human review; the package does not invent a source.
- 4 — Review the flagsThe package flags missing and conflicting inputs before compiling. Resolve or acknowledge each flag; unresolved fields stay blank in the outputs.
- 5 — Read the four outputsThe inactive JSON draft, permission matrix, readable report and acceptance-test specifications. Drafts are labelled DRAFT · NOT ACTIVE and cases NOT RUN.
B — Common causes
- No JSON draft or test specifications appeared
- The compiler step needs Python 3 in the host environment. Without it the readable review is still produced, but the compiled artifacts are not.
- Fields came back blank
- Unknown values remain blank and need human input. Supplied values without evidence remain visible with a provenance warning.
- Outputs contradict each other
- Review the flagged input conflicts first. If the files disagree about the same supplied field, send a redacted minimal example so the issue can be investigated.
C — Reporting a problem
Redacted, minimal, reproducible.
- The step you were on and what you expected instead.
- The smallest set of mandate fields that still reproduces it, with real figures replaced by placeholders.
- Whether Python 3 execution is available in your host environment.
- The exact wording of any error, with names, identifiers and internal detail removed.
Please do not send secrets, credentials, personal data or full chat logs. A redacted minimal example is more useful and safer for you.
Email: hello@iov.agency