Corabot Ai Center

adi-coralogix

CoraBot AI Center Demo — Svelte chatbot with Coralogix RUM + OTLP observability

svelteRum

SvelteKit chatbot with Coralogix RUM (Real User Monitoring), OpenTelemetry tracing, and Google Gemini 2.5 Flash Lite. Includes session replay, trace propagation from frontend to backend, and AI span tracing for observability in Coralogix.

Architecture

Frontend: SvelteKit app with Coralogix RUM (session replay, trace propagation to backend)
Backend: Python FastAPI + Gemini 2.5 Flash Lite, with OpenTelemetry GenAI instrumentation and OTLP export to a local collector (collector forwards to Coralogix)
Source Viewer: Browser-based source code viewer with syntax highlighting at /source

Prerequisites

Node.js 20+
Python 3.10+
API keys: Coralogix (RUM + OTEL), Google Gemini

Setup

1. Install dependencies

npm install
cd backend && python -m venv .venv && .venv/bin/pip install -r requirements.txt

2. Configure environment

Copy the env template and add your keys (backend also reads from project root .env):

cp .env.template .env

Edit .env with your values:

Variable	Description
`OTEL_EXPORTER_OTLP_ENDPOINT`	OTLP gRPC address for the Python API and Node build (default `http://localhost:4317`); targets your OpenTelemetry Collector.
`CORALOGIX_PRIVATE_KEY`	Send-Your-Data key for otel-collector-config.yaml so the collector exports to Coralogix (not used by the app process for OTLP).
`CORALOGIX_GUARDRAILS_KEY`	Optional. Separate Guardrails API key for `/api/chat` checks — do not reuse the OTLP ingestion key (OTLP keys do not work on the Guardrails HTTP API).
`GUARDRAILS_ENABLED`	Optional. `true` to enforce blocking on policy hits; `false` for observe-only (default).
`VITE_CORALOGIX_RUM_KEY`	Coralogix RUM key (browser monitoring)
`GEMINI_API_KEY`	Google AI Studio API key
`VITE_CHAT_API_URL`	Backend API URL (default: `http://localhost:8000`)

3. Run

Terminal 1 – Backend

npm run api

Terminal 2 – Frontend

npm run dev

Chat: http://localhost:5173
Source viewer: http://localhost:5173/source

Scripts

Script	Description
`npm run dev`	Start Svelte dev server (port 5173)
`npm run api`	Start Python FastAPI backend (port 8000)
`npm run build`	Build frontend for production
`npm run start`	Run production frontend (port 3000)
`npm run test:rum`	Run RUM test harness once (`BASE_URL` default localhost:5173)
`npm run test:rum:both`	Run harness against `BASE_URL_TEAM_A` then `BASE_URL_TEAM_B` (dual RUM stacks; see `.env.template`)
`npm run test:rum:loop`	Run RUM harness every 5 minutes

Project Structure

├── backend/           # Python FastAPI + Gemini
│   ├── main.py        # Chat API, OTLP traces/logs + GenAI instrumentation → collector
│   └── requirements.txt
├── src/
│   ├── lib/
│   │   ├── coralogix-rum.ts   # RUM init, session replay, user context
│   │   └── app.css
│   └── routes/
│       ├── +layout.svelte     # RUM init on mount
│       ├── +page.svelte       # Chat UI
│       └── source/           # Source code viewer
├── tests/
│   ├── rum-harness.mjs       # Puppeteer RUM tests
│   ├── run-rum-both.mjs       # Sequential runs for two stacks (dual RUM keys)
│   └── rum-harness-loop.mjs  # Loop harness every 5 min
└── instrumentation.cjs       # OpenTelemetry (production)

Coralogix AI Center — Evaluator Setup

The test harness sends duck, egg, and chocolate questions to generate data for the following Policy Catalog evaluators. Configure these in the Coralogix UI under AI Center → Policy Catalog:

Evaluator type	Value	Effect
Restricted topic	`duck`	Flags duck-related conversations as off-topic
Allowed topic	`egg`	Egg-related content passes the topic check
Competitor	`chocolate`	Flags chocolate references as a competitor mention

The harness (tests/rum-harness.mjs) sends questions covering all three topics plus the guardrail probes (PII, prompt injection, toxicity) in every run. After ~5 minutes the AI Center will show evaluator hits for duck conversations (restricted) and chocolate mentions (competitor).

User Context (RUM)

Call setCoralogixUserContext() after login to associate RUM sessions with users:

import { setCoralogixUserContext } from '$lib/coralogix-rum';

setCoralogixUserContext({
  userId: 'user-123',
  userName: 'Jane Doe',
  userEmail: '[email protected]'
});

License

MIT

Top categories