You are helping me build the following AI agent workflow.

## Goal
Call Analyzer Insights: Transcribes sales calls with AssemblyAI and uses OpenAI to extract client intent, interest scores, and upsell opportunities into Supabase.

## Specification
- What it does: Get clear insights from your sales calls by automatically transcribing audio with AssemblyAI and extracting client intent, interest scores, and upsell opportunities using OpenAI. The system then saves both the raw transcripts and these actionable metrics in Supabase for your immediate review.
- Trigger: Triggered by an event (Manual · on new audio file)
- Autonomy: Fully hands-off
- Expected setup effort: a few focused days
- Tools/services involved:
  - N8N: Workflow orchestration
  - AssemblyAI: Audio transcription and speaker diarization
  - OpenAI: Analyzing call transcripts for intent and upsell opportunities
  - Supabase: Storing transcription and analysis data

## Known pitfalls, handle each one explicitly in your implementation
1. AssemblyAI webhooks may arrive out of order if multiple calls are processed simultaneously, so add a unique call ID field to correlate results correctly.
2. OpenAI responses can occasionally fail JSON validation due to trailing commas or markdown formatting, so implement a retry loop with regex cleanup logic.
3. Supabase insertions might fail if the AI returns null values for required fields, so add default value mappings in the workflow before the database step.

## Reference implementation
https://n8n.io/workflows/2547 (workflow template)
Fetch and inspect this before building. If it matches my stack, adapt it;
if not, rebuild the pattern with my tools.

## Process requirements
1. Before building: ask me which of the listed tools I actually use and
   what my platform is (n8n / Make / code / other). Do not assume.
2. Adapt the pattern to my answers; do not force the reference stack.
3. Address every pitfall above; tell me how you handled each.
4. Provide a test plan I can run before letting this touch real data.
5. Ask before any step that sends messages, modifies data, or spends money.

Source: https://usecasesforagents.com/use-case/call-analyzer-with-ai-insights/ via usecasesforagents.com