summarizer

Zoom AI Services Summarizer

Implementation guidance for Zoom AI Services Summarizer across:

• synchronous inline transcript summarization (POST /aiservices/summarizer/summarize)
• asynchronous batch transcript-file jobs (/aiservices/summarizer/jobs*)
• recap, action item, detailed summary, and combined summary tasks
• Build-platform JWT generation and server-side credential handling
• webhook-driven batch status updates

Official docs:

• https://developers.zoom.us/docs/ai-services/
• https://developers.zoom.us/docs/ai-services/summarizer/
• https://developers.zoom.us/docs/ai-services/summarizer/fast-mode/
• https://developers.zoom.us/docs/ai-services/summarizer/batch-mode/
• https://developers.zoom.us/api-hub/ai-services/methods/endpoints.json
• Quickstart sample: https://github.com/zoom/ai-services-quickstart/

Routing Guardrail

• If the user already has transcript or dialogue text and needs a recap, summary, action items, or full summary, route here first.
• If the user has audio/video media and needs text first, route to ../scribe/SKILL.md, then chain here.
• If the user needs plain text translated, route to ../translator/SKILL.md.
• If the user needs live meeting media before summarization, route to ../rtms/SKILL.md.
• If the user needs the raw AI Services endpoint inventory, chain ../rest-api/SKILL.md.
• If the user needs webhook signature patterns, optionally chain ../webhooks/SKILL.md.

Quick Links

1. scenarios/high-level-scenarios.md
2. examples/fast-mode-node.md
3. examples/batch-webhook-pipeline.md
4. references/api-reference.md
5. references/environment-variables.md
6. references/samples-validation.md
7. troubleshooting/common-drift-and-breaks.md
8. RUNBOOK.md

Core Workflow

1. Get Build-platform API key and API secret.
2. Generate an HS256 JWT server-side.
3. Choose fast mode for inline transcript text or batch mode for stored transcript files.
4. Select the summarization task: recap, summary, action_items, or full_summary.
5. Submit the request and persist result.text or batch output files.
6. For batch jobs, poll job/file status or receive webhook notifications.

Endpoint Surface

Mode	Method	Path	Use
Fast	POST	/aiservices/summarizer/summarize	Synchronous summarization for inline transcript text
Batch	POST	/aiservices/summarizer/jobs	Submit asynchronous batch summarization job
Batch	GET	/aiservices/summarizer/jobs	List batch summarization jobs
Batch	GET	/aiservices/summarizer/jobs/{jobId}	Inspect job summary/state
Batch	DELETE	/aiservices/summarizer/jobs/{jobId}	Cancel queued/processing job
Batch	GET	/aiservices/summarizer/jobs/{jobId}/files	Inspect per-file results
Batch	GET	/aiservices/summarizer/jobs/{jobId}/files/{fileId}	Inspect one per-file result

Mode Selection

Need	Use
One bounded transcript string under 96 KB	Fast mode
Many .vtt, .srt, or .txt transcript files in S3	Batch mode
Summary immediately in app UI	Fast mode
Archive or compliance pipeline	Batch mode plus polling/webhooks
Audio/video input	scribe first, then summarizer

Output Contract

Summarizer returns rendered text in result.text. Treat it as display-ready text or markdown-like content, depending on the requested task:

• recap: concise recap
• summary: detailed summary
• action_items: action item sections grouped by owner
• full_summary: recap, summary, and action items together

Chaining

• Media to transcript to summary -> ../scribe/SKILL.md + summarizer
• Summary to localized output -> summarizer + ../translator/SKILL.md
• Batch webhook receiver -> summarizer + ../webhooks/SKILL.md
• Raw endpoint discovery -> ../rest-api/SKILL.md
• Cross-product routing -> ../general/SKILL.md

Operations

• RUNBOOK.md - 5-minute preflight and debugging checklist.