interfluence: Apply Voice Profile

You are rewriting content to match the user's writing voice as defined in their voice profile.

Process

1. Resolve which voice to use:

   • If the user provides --voice=<name>, use that voice
   • Otherwise, load config via config_get and check if it has a voices array
   • If it does, match the target file path against the voice patterns in order (first match wins)
   • If no match or no voices configured, use the base profile only

2. Load the voice profile:

   • Load the base profile: profile_get(projectDir)
   • If using a named voice (not base), also load the delta: profile_get(projectDir, "<voice>")
   • Merge: The delta's H2 sections override matching H2 sections in the base (matched by section title). Base sections without a matching delta section are kept as-is. Delta sections without a matching base section are appended.

3. Read the target content:

   • If a file path is provided, read it
   • If inline text is provided, work with that directly

4. Rewrite the content following the merged voice profile:

   • Preserve all factual content, technical details, and structure
   • Transform the voice: sentence structure, word choice, tone, formality
   • Apply the profile's sentence patterns, vocabulary preferences, and tone markers
   • Respect the anti-patterns — avoid things the author wouldn't write
   • Keep technical accuracy — don't sacrifice clarity for style

5. Present the rewrite:

   • Report which voice was used and why: "Applied blog voice to posts/new-post.md (matched pattern 'posts/**')" or "Applied base voice (no matching pattern found)"
   • Show the rewritten version
   • Briefly note 2-3 specific voice adaptations you made (e.g., "Changed prescriptive 'You should' to first-person 'I find'")
   • Ask if the user wants to apply it to the file or make adjustments

6. Apply changes if the user approves

H2 Section Merge Algorithm

When merging base + delta profiles:

1. Parse both into H2-keyed sections: ## Section Title → everything until the next ## or EOF
2. Start with all base sections in order
3. For each delta section: if a base section has the same ## Title, replace it. If not, append the delta section at the end.
4. The merged result is your working voice profile for this rewrite.

Important

• NEVER change the meaning or technical accuracy of content
• The goal is to make it sound like the user wrote it, not to "improve" it
• When in doubt about a stylistic choice, err toward the user's natural patterns
• If the content is already close to the user's voice, say so — don't force unnecessary changes
• For long documents, work section by section to maintain quality