# Sentifix

AI-powered UX signal intelligence platform.
Tagline: From Noise to Insights.

## Resources

- Full integration guide: https://sentifix.app/AGENTS.md
- Expanded LLM context: https://sentifix.app/llms-full.txt
- Site: https://sentifix.app

## What it does

Sentifix ingests raw user behavior events from web applications, clusters them into actionable signals, and generates AI assessments with severity ratings and fix recommendations.

## Pipeline

events → signals → assessments

1. Events: Raw telemetry (errors, crashes, rage clicks, dead ends, workarounds, feature patterns, performance issues)
2. Signals: Clustered events grouped by type and category, with severity (low/medium/high) and frequency
3. Assessments: AI-generated root cause analysis, impact assessment, and recommended fixes

## Use cases

- Surface the highest-impact bugs and crashes from production traffic without manual triage.
- Spot product opportunities (workarounds, repeated patterns, dead ends) that hint at missing features.
- Replace generic analytics dashboards with a prioritized list of issues and opportunities.
- Give AI coding agents a structured feed of real user pain to act on.

## API

POST /functions/v1/ingest-event
Body: { api_key, type, payload?, session_id?, user_fingerprint?, idempotency_key?, context? }
Batch: { api_key, events: [{ type, payload?, session_id?, user_fingerprint?, idempotency_key? }], context? }
Auth: Project API key (sfx_ prefix) in request body
Requires: Active subscription

Valid event types: error, crash, rage_click, dead_end, workaround, feature_pattern, performance
Signal types: issue, opportunity
Severity levels: low, medium, high

Context fields: page_url, user_agent, viewport, locale, referrer, timestamp
Batch limit: 50 events per request
Idempotency: Optional idempotency_key prevents duplicate ingestion on retries

Rate limits: Builder 2k/min, Scale 10k/min, Enterprise 25k/min

## Pricing

- Builder: 19/mo, 3 projects, 2k events/min
- Scale: 39/mo, 10 projects, 10k events/min
- Enterprise: custom, unlimited projects, 25k events/min

Currency symbol (€/$) swaps by visitor locale; the numeric value is unchanged. No free tier.

## Trust and compliance

GDPR compliant, based in the Netherlands. Anonymous user fingerprints, 30-day auto-delete of raw events. BYOK for Anthropic and OpenAI; Sentifix never stores customer LLM keys server-side beyond the encrypted record needed to call the provider.

## Integration

See AGENTS.md (https://sentifix.app/AGENTS.md) for the full integration guide with curl examples.

## Smart AI Routing

Sentifix routes each signal to the cheapest model in your BYOK keys that can solve it. Three tiers:

- Triage: Haiku 4, GPT-5 nano. Classification, dedupe, severity.
- Analysis: Sonnet 4.5, GPT-5. Root cause, repro steps, fix suggestion.
- Escalation: Opus 4, GPT-5 Pro. Only when triage confidence is low.

Typical cost reduction: ~60% versus running every signal on a frontier model. No quality loss because escalation kicks in automatically on hard cases.