Tickessa 0.11.0
Version 0.11.0 technically provides controlled automatic sending for individually approved, low-risk routine emails.
Published on the reference installation on 4 September 2026. AI_ENABLED=false and AI_AUTOMATIC_SEND=false remain unchanged, with no production rule active. Technical availability sends no automatic email.
Staged approval
Automatic replies require the external server switch, installation approval, cleared global/project stops, unchanged successful evaluation, active tasks/budgets and an exactly matching rule.
Rules bind project, category, named case, category reply profile, at least 95% confidence and a minimum source count. Quality, safety, cost and fallback are confirmed separately.
Hard safety stops
Legal, financial, privacy, complaint, security and escalation matters always require a person, as do attachments, missing information, unclear categories, conflicting/insufficient sources, provider errors and invalid output.
Emergency stops are rechecked before every provider call and immediately before sending.
Loop and duplicate protection
A ticket receives at most one automatic evaluation and one automatic reply. Parallel/repeated cron cannot create another attempt. Technical or ambiguous sending failures are never retried automatically.
A customer reply or I would like a personal review link hands the thread permanently to a person.
Knowledge and reply profile
Replies may use only approved, published, current project/version-appropriate sources with server-validated IDs. The assigned category profile controls salutation, tone, length, depth, closing and forbidden promises.
Audit without content copies
Tickessa stores state, stop reason, rule, configuration hash, AI runs, category, confidence, profile versions and content-free source references. The decision table duplicates no email, prompt, article or reply text.