What is AuditorIA

AuditorIA is an AI-powered phone call auditing platform built by LinkSolution. It transcribes, analyzes, and evaluates conversations automatically — eliminating the need to listen to recordings manually and letting you audit 100% of your contact center volume.

The problem it solves

In contact centers, auditing calls is slow, expensive, and subjective. Supervisors typically review less than 5% of total volume, which leads to:

• Lack of visibility into real service quality.
• Inconsistency between human auditors (low inter-rater reliability).
• Delays in detecting critical incidents.
• Inability to scale the quality process with operational volume.

AuditorIA automates this entire workflow with AI: every call, same criteria, in real time.

What you get with AuditorIA

Capability	What it does	Technology
Transcription	Converts audio to text, word by word	WhisperX / OpenAI / Deepgram
Diarization	Separates each speaker's turns	PyAnnote 3.1
Sentiment analysis	Evaluates tone and emotion per segment	pysentimiento (Spanish-optimized)
Auto-tagging	Classifies calls by topic and outcome	GPT-4o
Agent identification	Detects who is agent vs customer	GPT + heuristics
AI Chat	Natural-language questions about the call	GPT-4o + RAG
Quality auditing	Automatic scoring against configurable criteria	Rules engine + LLM
Reports	Dashboards, trends, ranking, emotion radar	Recharts + SQL aggregations
Notifications	Real-time alerts on tasks and events	SSE + Redis pub/sub

Primary use cases

• Contact centers — mass auditing of customer service, sales, technical support.
• Continuous quality monitoring — daily performance evaluation for every agent.
• Compliance — automatic verification of regulatory scripts (greeting, consent, legal close).
• Coaching — evidence-based improvement opportunities for each agent.
• Business intelligence — trend detection, recurring drivers, friction spikes.
• Retention & upsell — spotting missed opportunities and retention attempts.

High-level architecture

AuditorIA is a microservices architecture with asynchronous queues:

End-to-end flow

1. User uploads an audio file (UI or API).
2. Backend stores the audio in MinIO and creates a task in PostgreSQL.
3. Task is queued in Redis toward the appropriate worker (based on device).
4. Worker processes transcription + diarization.
5. GPT worker analyzes sentiment, generates tags, and identifies agent/customer.
6. Results are persisted in PostgreSQL.
7. Frontend receives SSE notification and updates the viewer.

Two available APIs

API	Port	Authentication	Primary use
Internal	8000	JWT (Keycloak OAuth2/OIDC)	Frontend, internal apps
External	8001	Integration credential	Third-party integrations

info

See Authentication for detailed examples of both methods.

Supported audio formats

Format	Extension	Notes
WAV	.wav	Uncompressed, max quality
MP3	.mp3	Compressed, most common
FLAC	.flac	Lossless compression
OGG	.ogg	Free format
M4A	.m4a	Apple AAC

warning

Quality recommendation

For best results: 16 kHz or higher sample rate, mono or stereo with one channel per speaker, ≥128 kbps bitrate for compressed formats. See Uploading audio for details.

Supported languages

• Spanish (es) — primary, tuned for Río de la Plata, Mexican, Colombian variants.
• English (en), Portuguese (pt), French (fr), German (de), Italian (it).
• 90+ additional languages via Whisper (ja, zh, ko, ar, ru, nl, etc.).

See Language selection for the full guide.

Where do I start?

End user (auditor, supervisor, admin)

Start with the User Guide — covers login, dashboard, how to upload audio from the UI, how to review transcriptions, and how to interpret reports.

Developer or integrator

Go to Quickstart for your first API transcription in 5 minutes.

Platform admin

See Roles and permissions and Managing campaigns.

---

Next step

• Quickstart — your first transcription in 5 minutes (API)
• User Guide — login & navigation (UI)
• Key concepts — understand tasks, workers, queues, and events