Infrastructure Demo

Resilience in action

Watch Presence's multi-provider LLM gateway handle real failures in real time. No reloads. No downtime. Just seamless failover.

Live System
Multi-provider LLM Gateway

Network Topology

Live traffic flow visualization

Streaming
AIGateway⚡GROQ120ms✦GEMINI340ms◎OPENAI580ms→Output

Chaos Controls

Inject faults, watch recovery

3

Active

0

Failed

groq

Primary

Calls real backend endpoints. TrueFoundry routes automatically when a provider fails.

Failover chain

  1. 1

    Groq (primary)

    Fastest — llama3-70b at ~120ms. Handles all traffic while healthy.

  2. 2

    Gemini (fallback)

    Auto-promoted if Groq errors or latency spikes. gemini-flash-latest.

  3. 3

    OpenAI (tertiary)

    Last resort. gpt-4o-mini. Ensures availability even if two providers fail.

  4. 4

    Zero user downtime

    The companion chat continues responding — no error state shown to user.

For judges

  1. 1.

    Click "Kill Primary (Groq)" — watch traffic reroute to Gemini instantly

  2. 2.

    Send a message in the chat tab — system responds from Gemini seamlessly

  3. 3.

    Click "Kill Both" — OpenAI takes over, user never sees an error

  4. 4.

    Click "Restore All" — system self-heals and returns to Groq

Provider Status

PRIMARY
⚡

Groq

Active

Latency

120ms

Requests

0

served
Uptime99.9%
llama3-70b
✦

Gemini

Standby

Latency

340ms

Requests

0

served
Uptime98.7%
gemini-flash-latest
◎

OpenAI

Standby

Latency

580ms

Requests

0

served
Uptime99.5%
gpt-4o-mini