meeting-from-transcription

Meeting from Transcription

Create meeting notes from transcription sources with intelligent metadata extraction.

Input Detection

Workflow

Step 1: Resolve Input

No input provided:

First, prime the index to ensure it's up-to-date:

para_index_prime({ dir: "00 Inbox", response_format: "json" })

Then query for transcriptions:

para_index_query({ frontmatter: "type=transcription", dir: "00 Inbox", response_format: "json" })

• If transcriptions found → Display numbered list with recorded dates, then stop
• If none found → Use AskUserQuestion to prompt for file path

Input provided:

• .vtt extension → See vtt-conversion.md — REQUIRES --date flag
• .md or bare name → Continue to Step 2

Path expansion: If no / in path, prepend 00 Inbox/. If no .md extension, append it.

Step 2: Read & Validate Transcription

para_read({ file: "<PATH>", response_format: "json" })

Validate frontmatter:

• Has frontmatter block
• Has type: transcription
• Has recorded field OR user provided --date

On validation error → See error-handling.md

Extract:

• recorded → Meeting datetime (or use --date override)
• summary → Use if present, otherwise generate from content
• Body content → Everything AFTER the --- frontmatter block

Step 3: Determine Meeting Type

If --type override provided, use it. Otherwise infer from transcript content.

See meeting-types.md for enum values and inference signals.

Step 3b: Determine Area or Project

REQUIRED: Meeting notes must have either area or project in frontmatter.

1. Check transcription frontmatter for existing area or project field
2. If present → Use that value (already a wikilink)
3. If not present → Use AskUserQuestion to ask user:
   • "Which area or project does this meeting belong to?"
   • Provide options based on context from the transcript content
   • Include "Other" option for custom input

The value must be a wikilink format: [[Area Name]] or [[Project Name]]

Step 4: Commit Uncommitted Changes

para_commit({ response_format: "json" })

Step 5: Create Meeting Note

cd ${CLAUDE_PLUGIN_ROOT} && bun src/cli.ts create \
  --template meeting \
  --source-text "<TRANSCRIPT_BODY>" \
  --arg "meeting_date=<DATE>" \
  --arg "meeting_type=<TYPE>" \
  --arg "transcription=[[<NOTE_NAME>]]" \
  --arg "summary=<SUMMARY>" \
  --arg "area=[[<AREA_NAME>]]"

Or if project instead of area:

  --arg "project=[[<PROJECT_NAME>]]"

Arguments:

• meeting_date — ISO format: YYYY-MM-DDTHH:mm:ss
• meeting_type — One of: 1-on-1, standup, planning, retro, review, interview, stakeholder, general
• transcription — Note name WITHOUT path or .md extension, wrapped in [[...]]
• summary — Concise 1-line description (max 100 chars)
• area OR project — REQUIRED (one of these must be provided) — Wikilink to parent area or project

Step 6: Link Transcription to Meeting

Extract meeting filename from CLI output, then:

para_frontmatter_set({
  file: "<TRANSCRIPTION_PATH>",
  set: { "meeting": "[[<MEETING_NOTE_NAME>]]" },
  response_format: "json"
})

Step 7: Final Commit

para_commit({ response_format: "json" })

Step 8: Report Results

Return:

• Meeting note path created
• Meeting type (inferred or overridden)
• Summary used
• Confirmation of bi-directional links

References

Load as needed based on input type or errors:

• VTT files: references/vtt-conversion.md — Conversion command, date requirement
• Meeting types: references/meeting-types.md — Enum values, inference rules
• Errors: references/error-handling.md — Validation errors, recovery steps