AI-powered call-support platform. Transcribes calls, detects intent, generates actions, and escalates conversations that need human intervention.
callsup-web/ React 19 + TypeScript dashboard (frontend)
consolidated/
callsup-platform/ Platform service
callsup-audio-engine/ Whisper transcription · port 8010
callsup-intelligence-engine/ NLU + action engine · port 8011
svc-llm-adapter/ GitHub Copilot proxy · port 9100
callsup-specs/ OpenAPI contracts + governance
| Service | Port | Description |
|---|---|---|
| Audio Engine | 8010 | Ingest audio, redact PII, retrieve transcripts |
| Intelligence Engine | 8011 | Intent detection, action decisions, escalation |
| LLM Adapter | 9100 | Authenticated GitHub Copilot / OpenAI proxy |
Located at callsup-web/. Features:
- Live service health — online/offline status and version per service
- Analytics stats — pending escalation tickets, resolved today, total escalations, services online
- Task queue — escalated conversations that require human intervention, with one-click Resolve
- Audio ingest — upload recordings for transcription
- Transcripts — browse conversation segments with PII redacted
- Intelligence — step-by-step analysis tool
- Call simulation — test the full pipeline with a scripted conversation
- Context management — manage business knowledge used by the intelligence engine
See RUNNING_THE_SYSTEM.md for full instructions.
# 1. Start backend services (3 terminals — see RUNNING_THE_SYSTEM.md)
# 2. Start frontend
Set-Location callsup-web
npm install
npm run dev # http://localhost:5173- PII is redacted before any third-party LLM call (phone, email, SSN, card numbers)
- Audio encrypted at rest with Fernet
- TLS enforced in transit (configurable per service)
- Raw service URLs are never exposed in the frontend UI
# Backend smoke test (requires all 3 services running)
& "C:\Users\nyaga\Documents\.venv\Scripts\python.exe" smoke_test.py
# Module unit tests — see PROJECT_STATUS_SUMMARY.md § 3