Skip to main content

Voice and corrections

A voice profile is a machine-readable description of how your content should read: the vocabulary it should use, the terms it should avoid, and the tone and style that make it recognizable. Most workspaces call their default profile brand. A workspace with several audiences carries several profiles, each bound at the point it governs. A profile works as a set of checks. kapi and Bowrain run it over content and report what drifts off-voice, in the way a test reports what broke.

Bowrain aggregates repeated corrections into candidate rules. Reviewers can inspect their evidence, preview their effect on stored content and promote them into the workspace's terms. The Voice section of the Context hub provides this workflow.

The loop​

Repeated corrections can reveal gaps in a profile, such as an unwanted phrase or a term reviewers consistently replace. Bowrain records these corrections for review:

  1. Correct. A person in the review session or the editor, or an AI assistant working through MCP, fixes content the profile should have caught. Bowrain records the correction with provenance: the original text, the fix, the voice dimension, the point it was made at, and who made it.
  2. Surface. When the same (original → corrected) correction recurs across the project, it surfaces as a candidate rule. Candidates are recomputed live from the correction stream and reflect the current evidence.
  3. Preview. Before a candidate is enforced, Bowrain computes its blast radius. It runs the rule over your stored content and shows how many blocks it would newly flag, how many it resolves, and how many of the new findings fail a check.
  4. Promote. A reviewer promotes the candidate. Bowrain writes it into the workspace's terms as a term rule and records who promoted it and when. From then on, a rule-based check enforces it on every generation, in every language the terms cover. The promoted rule is concept-backed: it references the concept its term denotes, so the rule and the term it governs share one identity in the Context hub. Promoting a voice rule is a governed edit, and a concept's story shows the rules that reference it.
  5. Monitor. Bowrain tracks aggregate compliance and raises a drift alert when it declines.
Correctperson or AI (MCP)recursSurfacecandidate rulecompute impactPreviewblast radiusreviewer approvesPromoteinto the profileenforceMonitoraggregate compliancedrift alert when compliance erodesfeeds the next correction

Recurring corrections provide evidence for candidate rules. Reviewers preview and promote them into the terms the workspace shares.

Rules inherit and expire​

A rule is bound at a point and applies to everything beneath it until a narrower profile overrides it. A rule can also carry a validity: it starts and stops on a date. Use validity windows for campaign vocabulary or a rename's transition period.

Progressive autonomy​

Every profile starts fully manual: each candidate waits for a person to promote it. When a team trusts the candidates a dimension produces, it can raise that dimension's auto-promote threshold. A correction that pushes a term past the threshold then promotes its rule automatically and announces it. You turn autonomy on per profile, one dimension at a time.

Where it shows up​

The browser, an AI assistant and the command line all work with the same voice profiles:

  • Web: each voice profile's review lists its candidates beside a detail pane, in groups: needs you, suggested, approved, established and rejected. A phrasing your team corrected to more than one form waits under needs you until you choose the form. Select a candidate to see the corrections behind it and who decided it, then promote, change or reject it from the pane's toolbar, or press a, c or d. Choose a project in the page header to preview a candidate's blast radius over that project's content. The Context hub's dashboard links each profile's review and marks a drifting project on its row.
  • MCP: an AI assistant can read the candidates its own corrections produced, preview a candidate's impact, and promote it.
  • CLI / CI: a promoted rule is an ordinary check. kapi enforces it locally and in CI through its exit code, without contacting the server.

The open engine detects and scores. Bowrain keeps the decisions, profile versions, provenance, autonomy settings and drift history, and provides the screens that work with them.

For larger voice changes, such as banning a term across the workspace or swapping a preferred term, the Context hub groups the voice-rule edits into a reviewable change-set. Bowrain measures its blast radius before anyone approves it.