Templates

Templates control how Noctune formats a SOAP note. You can use the built-in defaults, customize them, or build your own from scratch. A well-written template is the single biggest lever you have over note quality — Noctune’s AI follows your template carefully, so the quality of the output is bounded by the quality of the prompt you give it.

System vs. user templates

• System templates are the Noctune-provided defaults (Wellness Exam, Emergency Triage, Dental, General SOAP). You can’t edit them, but you can duplicate one and modify the copy.
• My templates are yours. Edit, reorder, and delete freely.

Creating a template

Click New Template. A template is a structured prompt: markdown headings for each section (Subjective, Objective, Assessment, Plan) plus one or two sentences describing what belongs in each.

Editing your own templates

Open any template under My Templates and click Edit. You’re editing plain markdown — headings (##, ###), bullets, and instruction prose.

Save when you’re done. Your next encounter using that template will reflect the changes immediately; past encounters keep whatever was generated at the time.

Auto-review on save

When you save a template, Noctune automatically runs it through an AI reviewer and shows you suggestions: missing sections, vague instructions, sample data that might leak into real notes, and structural issues that make the output worse. Each suggestion has a severity (error / warning / info) and a suggested fix you can accept with one click.

You’re free to ignore any suggestion — nothing is enforced. But the reviewer catches real problems that make notes worse, and applying its suggestions is the fastest way to get measurably better output. Think of it as a linter for your prompts.

Common categories the reviewer flags:

• Sample data — e.g. “Patient: Fluffy” in the template; the AI sometimes copies example names into real notes
• Missing section — a standard section (Plan, Assessment) is absent
• Structure — inconsistent heading levels, missing blank lines, or overlapping section purposes
• Placeholder — a {like_this} placeholder without an instruction for what to fill in
• Instruction — a section header with no guidance for what to write
• Best practice — vet-specific advice like “specify metric vs. imperial units”

Writing LLM-friendly templates

A few patterns reliably produce better notes. These are why our system templates look the way they do.

1. Explicit section instructions

Bad:

## Subjective

Good:

## Subjective
 
Summarize the pet owner's concerns, symptoms, and history as provided.

Why: LLMs don’t infer context from section names alone — especially the first time they see your template. One-sentence instructions shape what gets included and what’s skipped.

2. Say what to do when data is missing

Good:

### Objective
 
Extract and summarize **all** examination findings, tests, and vital signs.
If a finding is not mentioned in the transcript, omit it. Do not invent values.

Why: Models default to filling gaps plausibly — which in medical context means hallucinating vitals. Telling them explicitly to omit is the single most important anti-hallucination instruction.

3. Use markdown headings, not custom formatting

Bad: SUBJECTIVE: | OBJECTIVE: | ...

Good: ### Subjective on its own line

Why: Models respect ## / ### more reliably than ALL-CAPS or custom delimiters. The output ends up better-structured with no extra effort.

4. Anti-truncation instructions

Good:

### Objective
 
Include every exam finding — do not truncate. If the visit contains multiple
exam descriptions (e.g. separate sections on soft tissue and vitals), merge
them into one comprehensive summary.

Why: Default model behavior is to summarize and compress. For clinical notes, you usually want the opposite. Tell it.

5. Put meta-instructions at the top

Good:

## Formatting conventions
 
- Use **metric units** (kg, cm, °C).
- Bullet Plan items; prose everywhere else.
- Reference the patient by name, not "the dog."
 
## SOAP Note
 
### Subjective
 
...

Why: Instructions near the top carry more weight than instructions buried in section 4. If you want a convention to apply globally, write it globally.

6. No sample data

Bad:

**Patient:** Fluffy
**Weight:** 12.3 kg

Good:

**Patient:** (patient's name from transcript)
**Weight:** (weight in kg, or omit if not mentioned)

Why: Sample names and numbers sometimes get copied into real notes. Parenthesized instructions are clearly not data and the model fills them with real content.

7. One section, one purpose

If Assessment asks for “diagnosis and recommended treatments” and Plan asks for “treatments and follow-ups,” you’ll get duplicate content in both. Each section should have one job.

Example: an annotated template

Here’s a minimal wellness-exam template that follows all seven rules:

## Formatting conventions
 
- Use metric units (kg, cm, °C).
- Omit any section where the transcript has no information — do not invent values.
- Reference the patient by name.
 
## SOAP Note
 
### Subjective
 
Summarize the owner's reported concerns, symptom timeline, and relevant history.
 
### Objective
 
All physical exam findings, weight, vitals (HR, RR, Temp, CRT, MM color),
and point-of-care test results mentioned in the transcript. Merge duplicate
findings into one comprehensive summary.
 
### Assessment
 
The veterinarian's working diagnosis and any differentials mentioned.
 
### Plan
 
- **Treatments and medications** (dose, route, duration)
- **Follow-up** (timing and trigger)
- **Client education points** covered during the visit
 
## Client Communications
 
Summarize what was discussed with the owner:
 
- Findings shared
- Risks or side effects explained
- Questions asked and answered
- Recommendations accepted or declined

What each piece is doing:

• Formatting conventions at the top cover units, anti-hallucination, and voice globally — so each section doesn’t need to repeat them.
• Section instructions are one sentence, action-oriented.
• Objective has an explicit anti-truncation instruction because incomplete vitals is a common failure mode.
• Plan uses bullets (matches PIMS paste targets); Subjective and Assessment stay as prose.
• Client Communications is separate from Plan — Plan is what you’re doing, Client Communications is what you said. Separating them keeps both clean.

Reordering

Drag templates by the handle on the left to change their order in the picker that appears during recording. The top template becomes your default.

Duplicating

Every template (yours or a system template) has a Duplicate action. The copy lands in My Templates and is fully editable.

Community templates

See templates other vets have shared in the Community Template Library. Fork anything that looks useful and edit the copy.