AI setup prompts

Use these prompts in Cursor, Copilot, Windsurf, ChatGPT, or Claude so an agent can integrate Amlexia without prior context. Prompts include all four SDKs, auto cost, alerts, environments, and verification.

Human-readable full guide

For step-by-step setup (not AI), start with Complete setup and Using the dashboard.

Security

Never put AMLEXIA_SDK_KEY in the browser, NEXT_PUBLIC_*, mobile apps, or git. Get your key from app.amlexia.com → Project → Settings (am_...).

Before you start

Step	Action
1	Sign up at app.amlexia.com/signup
2	Create organization and project
3	Copy SDK key from Settings
4	Set `AMLEXIA_SDK_KEY` in server environment only
5	Set `AMLEXIA_ENVIRONMENT=production` (or `staging`)
6	Confirm events at Overview or Live (~60s)

Resource	URL
Docs	https://docs.amlexia.com
Ingest	https://ingest.amlexia.com
Dashboard API	https://api.amlexia.com (Clerk JWT — not for SDK)
Node	`npm install @amlexiahq/node`
Python	`pip install amlexia`
Go	`go get github.com/amlexiahq/amlexia-go`
Ruby	`gem install amlexia`
Plans	Startup free = 10k events/month

Master prompt (recommended)

Paste the block below into your agent. Fill section 12 with your repo paths.

Expand — full master prompt

text

You are a senior engineer integrating Amlexia observability. You have NO prior context — follow ONLY this spec.

═══════════════════════════════════════════════════════════════
1. WHAT AMLEXIA IS
═══════════════════════════════════════════════════════════════
Operational intelligence (lightweight APM) for backends:
- Inbound HTTP (Express, FastAPI, Next.js, Go, Ruby, …)
- Outbound AI (OpenAI, Anthropic, Gemini, Groq), payments (Stripe, Razorpay, PayPal), auth (Clerk), messaging, infra
- Traces, live stream, alerts (email/Slack/webhook), insights, AI summary, operations/SLO view
- Environment filter: production / staging / development on dashboard when events include environment tag

Data flow:
  Server ──SDK──► https://ingest.amlexia.com/v1/events
  Team   ──Clerk──► https://app.amlexia.com (NEVER put SDK key in browser)

Amlexia does NOT run my app logic or call vendors for me.

═══════════════════════════════════════════════════════════════
2. KEYS & ENV (server only)
═══════════════════════════════════════════════════════════════
AMLEXIA_SDK_KEY=am_...                    # required
AMLEXIA_INGEST_URL=https://ingest.amlexia.com
AMLEXIA_ENVIRONMENT=production            # recommended for env filter
AMLEXIA_RELEASE=1.0.0                     # optional
AMLEXIA_SERVICE_NAME=api                  # optional (Django default: api)

NEVER: SDK key in frontend, NEXT_PUBLIC_*, mobile, committed secrets.

═══════════════════════════════════════════════════════════════
3. INSTALL (exact names)
═══════════════════════════════════════════════════════════════
Node:   npm install @amlexiahq/node       # org is @amlexiahq NOT @amlexia
Python: pip install amlexia
        pip install "amlexia[fastapi]"     # optional extras
Go:     go get github.com/amlexiahq/amlexia-go
Ruby:   gem install amlexia
Bun:    bun add @amlexiahq/node
Deno:   import from npm:@amlexiahq/node

═══════════════════════════════════════════════════════════════
4. CLIENT PATTERNS (all languages)
═══════════════════════════════════════════════════════════════
- Buffered batching, flush on interval, shutdown() on SIGTERM
- sampleRate / sample_rate 0–1 to reduce volume
- diagnostic: true → log flush errors to stderr
- Health: CLI `npx amlexia health` or `amlexia health` (Go/Ruby too)
- wrapFetch / HTTP wrapper → auto-track outbound calls
- OpenAI/Anthropic helpers → pass usage from vendor response
- On 402: monthly plan limit (Startup = 10k events/mo) — do not spin-retry forever

Node:
  import { AmlexiaClient, wrapFetch, trackOpenAICompletion } from '@amlexiahq/node';
  const client = AmlexiaClient.fromEnv();
  globalThis.fetch = wrapFetch(client, fetch);
  await client.shutdown();

Python:
  from amlexia import AmlexiaClient
  client = AmlexiaClient.from_env()

Go:
  client, _ := amlexia.NewFromEnv()
  defer client.Shutdown()

Ruby:
  client = Amlexia::Client.from_env

Cloudflare Worker: ctx.waitUntil(client.flush()) after track()

═══════════════════════════════════════════════════════════════
5. FRAMEWORKS
═══════════════════════════════════════════════════════════════
Express:  AmlexiaMiddleware from @amlexiahq/node/express
Fastify:  amlexiaPlugin from @amlexiahq/node/fastify
Hono:     amlexiaHonoMiddleware from @amlexiahq/node/hono
Next.js:  withAmlexia from @amlexiahq/node/next — Route Handlers ONLY (server)
FastAPI:  AmlexiaMiddleware from amlexia.fastapi_integration
Flask:    amlexia_track from amlexia.flask_integration
Django:   AmlexiaMiddleware in MIDDLEWARE

═══════════════════════════════════════════════════════════════
6. COST (reported + estimated)
═══════════════════════════════════════════════════════════════
Ingest enriches cost_usd when you send model + provider + tokens WITHOUT cost_usd:
  cost_source = "estimated" (public model price table)
If you send cost_usd > 0:
  cost_source = "reported"

OpenAI after completion — tokens required for estimate:
  await client.track({
    endpoint: 'POST /v1/chat/completions',
    method: 'POST',
    statusCode: 200,
    latencyMs,
    provider: 'openai',
    modelName: res.model,
    tokensInput: usage.prompt_tokens,
    tokensOutput: usage.completion_tokens,
    // optional explicit: costUsd: yourNumber,
  });

Or use trackOpenAICompletion(client, { model, statusCode, latencyMs, usage }).

Stripe:
  provider: 'stripe',
  costUsd: chargeAmountUsd,
  metadata: { payment_intent_id: pi.id },  // NO card numbers

Inbound webhooks: isWebhook: true, provider: 'stripe'

Providers: openai, anthropic, gemini, groq, togetherai, replicate, stripe, razorpay, paypal, twilio, resend, sendgrid, clerk, supabase, firebase, aws

═══════════════════════════════════════════════════════════════
7. EVENT FIELDS
═══════════════════════════════════════════════════════════════
Required: endpoint, method, statusCode/status_code, latencyMs/latency_ms
Optional: provider, modelName/model_name, tokens_*, costUsd/cost_usd, environment, release, trace_id, span_id, session_id, user_id, metadata (no secrets), isWebhook/is_webhook
Headers: x-session-id, x-user-id

═══════════════════════════════════════════════════════════════
8. DASHBOARD (human — not in code)
═══════════════════════════════════════════════════════════════
User configures at app.amlexia.com:
- Alerts: latency, error_rate, cost, anomaly, provider_outage, retry_storm, cost_explosion, token_spike, webhook_degradation
- Channels: email | slack | webhook (HTTPS POST JSON)
- Saved views, env compare, usage meter (10k/mo free)
No SDK code for alerts.

═══════════════════════════════════════════════════════════════
9. VERIFY
═══════════════════════════════════════════════════════════════
curl -X POST https://ingest.amlexia.com/v1/events \
  -H "Content-Type: application/json" \
  -d '{"sdk_key":"am_...","events":[{"endpoint":"GET /health","method":"GET","status_code":200,"latency_ms":1,"timestamp":'"$(date +%s)"'","environment":"production"}]}'
Expect: accepted / queued

npx amlexia health

═══════════════════════════════════════════════════════════════
10. TASKS FOR YOU
═══════════════════════════════════════════════════════════════
1. Pick SDK for my stack (Node/Python/Go/Ruby)
2. Wire framework middleware OR manual track()
3. Track LLM calls (helpers or manual tokens)
4. Track Stripe/payments with costUsd
5. Mark inbound webhooks isWebhook: true
6. Set AMLEXIA_ENVIRONMENT per deploy
7. shutdown() / flush on serverless
8. .env.example with placeholders only
9. README section: Amlexia + health check command

═══════════════════════════════════════════════════════════════
11. MY PROJECT (fill in)
═══════════════════════════════════════════════════════════════
Framework:
Entry files:
OpenAI/Anthropic location:
Stripe/webhooks location:
Deploy target (Vercel, Fly, Workers, K8s):

Implement. List every file changed. Do not commit real SDK keys.

Quick prompts by stack

Node / TypeScript (Express)

text

Integrate @amlexiahq/node into this Express app.
Use AmlexiaMiddleware, AmlexiaClient.fromEnv(), AMLEXIA_ENVIRONMENT, wrapFetch for outbound OpenAI if present.
Track Stripe webhooks with isWebhook: true. shutdown on SIGTERM.
Run npx amlexia health when done. List files changed.

Python (FastAPI)

text

Integrate amlexia into this FastAPI app.
Use AmlexiaMiddleware from amlexia.fastapi_integration, AmlexiaClient.from_env().
Track OpenAI with tokens for cost estimate. AMLEXIA_ENVIRONMENT=production in .env.example.
amlexia health when done.

Go

text

Integrate github.com/amlexiahq/amlexia-go.
NewFromEnv(), defer Shutdown(), Track() on main HTTP handlers and OpenAI client calls.
Set Environment pointer on events. go run .../cmd/amlexia health when done.

Ruby

text

Integrate gem amlexia. Client.from_env, track routes, Amlexia::OpenAI.track_openai_completion for LLM.
amlexia health when done.

Cloudflare Workers

text

Use @amlexiahq/node in this Worker. AmlexiaClient with small maxBatchSize, track in fetch handler, ctx.waitUntil(client.flush()).
Never expose AMLEXIA_SDK_KEY to client bundle — use env secret binding only.

OpenAI / LLM only

text

Add Amlexia tracking for every OpenAI/Anthropic call in this repo.
Use SDK helpers (trackOpenAICompletion / track_anthropic_message) or manual track with modelName, tokensInput, tokensOutput, provider.
Do not log API keys. Verify with amlexia health.

Stripe / payments

text

Track all Stripe API calls with provider: stripe and costUsd from charge amount.
Inbound POST /webhooks/stripe handlers: isWebhook: true, provider: stripe.

Production audit

text

Audit this repo for Amlexia production readiness:
- SDK key only server-side
- shutdown/flush on all processes
- AMLEXIA_ENVIRONMENT set per environment
- LLM events include model + tokens
- No secrets in metadata
- amlexia health passes
List gaps and fix them.

Quick reference

Install

bash

npm install @amlexiahq/node
pip install amlexia
go get github.com/amlexiahq/amlexia-go
gem install amlexia

Cost rules

Topic	Rule
Estimated cost	Send model + provider + tokens; omit `cost_usd`
Reported cost	Send `costUsd` / `cost_usd` > 0
Dashboard	Shows reported vs estimated on Overview
Stripe / payments	You send `cost_usd` from your fee logic

Framework cheat sheet

Stack	Pattern
Express	`AmlexiaMiddleware`
Fastify	`amlexiaPlugin`
Hono	`amlexiaHonoMiddleware`
Next.js	`withAmlexia` (server routes only)
FastAPI	`AmlexiaMiddleware`
Flask	`@amlexia_track`
Django	`AmlexiaMiddleware`

Verify

bash

npx amlexia health
curl -s https://ingest.amlexia.com/health

After the agent finishes

Set real AMLEXIA_SDK_KEY in deployment secrets.
Deploy and hit a route or LLM call.
Open app.amlexia.com → Overview, Live, Providers.
Check usage meter (10k/mo on free tier).
Create alerts (email, Slack, or webhook).
Use environment filter with AMLEXIA_ENVIRONMENT.

Support: support@amlexia.com

AI setup prompts ​

Before you start ​

Master prompt (recommended) ​

Quick prompts by stack ​

Node / TypeScript (Express) ​

Python (FastAPI) ​

Go ​

Ruby ​

Cloudflare Workers ​

OpenAI / LLM only ​

Stripe / payments ​

Production audit ​

Quick reference ​

Install ​

Cost rules ​

Framework cheat sheet ​

Verify ​

After the agent finishes ​

AI setup prompts

Before you start

Master prompt (recommended)

Quick prompts by stack

Node / TypeScript (Express)

Python (FastAPI)

Go

Ruby

Cloudflare Workers

OpenAI / LLM only

Stripe / payments

Production audit

Quick reference

Install

Cost rules

Framework cheat sheet

Verify

After the agent finishes