Blog
-
How to Handle Stripe Webhooks Reliably on Serverless
Verify the stripe-signature on the raw body, acknowledge fast, dedupe on the event id, and fulfill asynchronously — the four moves that make serverless Stripe webhooks reliable.
Aleksandr Kubarskii · · 12 min read -
Secrets and environment variables for serverless functions
Why serverless secrets belong in per-function config injected as env vars — off your code, prompts, and client bundles — plus the auth-vs-config boundary and an honest scope: a config store, not a secrets manager.
Aleksandr Kubarskii · · 13 min read -
Three Steps to Create Your Endpoint with Inquir Compute
Aleksandr Kubarskii · · 5 min read -
Building a REST API on serverless functions
A pragmatic guide to shipping a serverless REST API on gateway-backed functions: map resources to routes, enforce API keys at the API gateway, standardize your request and response shape, add cursor pagination and a single error contract, and know exactly where the limits are.
Aleksandr Kubarskii · · 12 min read -
Serverless observability: traces, logs, and execution history
Ephemeral containers can't be SSH'd into. See how per-invocation execution history — traces, structured logs, retry counts, and 30-day retention — lets you debug a failed webhook, alert on failure rates, and answer 'what happened to request X?' as a lookup.
Aleksandr Kubarskii · · 10 min read -
Idempotency: making webhooks and background jobs safe to retry
At-least-once delivery means the same webhook will arrive more than once, so duplicates are inevitable. Here is the idempotency-key pattern, natural upserts, and where Inquir's startNew dedup, retries, and dead-letter queue help — and where you must be idempotent yourself.
Aleksandr Kubarskii · · 13 min read -
Serverless ML inference in Python: ONNX on CPU, load once
Serve a trained model behind an HTTP route without running a model server. The load-once-at-module-scope pattern, ONNX and scikit-learn attached as layers, memory and timeout budgets, warm pools, and the honest CPU-only limits.
Aleksandr Kubarskii · · 13 min read -
How to Handle GitHub Webhooks on Serverless
Verify x-hub-signature-256 with a timing-safe HMAC (or the turnkey webhookMode: 'github' route), route by x-github-event, dedupe by x-github-delivery, ack fast, then run CI/CD, bots, and notifications in async background jobs.
Aleksandr Kubarskii · · 12 min read -
Multi-step serverless pipelines: retries, branching, and human approval
Model long, multi-step work as a serverless pipeline of small steps — each with its own timeout, step retries, and run record — with branching, fan-out/fan-in, human approval gates that pause and resume, and cron triggers.
Aleksandr Kubarskii · · 13 min read -
A Cloudflare Workers Alternative: When You Need Containers, Not Isolates
Cloudflare Workers runs V8 isolates at the edge; Inquir runs full Node, Python, and Go containers. Where the isolate ceiling bites, what containers change, and how to choose.
Aleksandr Kubarskii · · 12 min read -
API Key Authentication for Serverless Functions
Why API key authentication belongs at the gateway—before your code runs—instead of scattered across every serverless handler: route-level keys vs bearer tokens, rotation without a redeploy, per-tool keys for AI-agent endpoints, and where authorization still lives.
Aleksandr Kubarskii · · 13 min read -
An AWS Lambda Alternative: Container-Backed Serverless
An honest look at what pushes teams off AWS Lambda (packaging, runtime ceiling, five services per endpoint), what a container-backed alternative changes, what it does NOT (cold starts, timeouts), and when Lambda is still right.
Aleksandr Kubarskii · · 5 min read -
Node.js vs Python vs Go: Choosing a Serverless Runtime
Node.js 22, Python 3.12, and Go 1.22 run container-backed behind one gateway. Here are the differences that actually matter when you choose — ecosystem, streaming, cold starts, and Go's compile-to-plugin caveat.
Aleksandr Kubarskii · · 6 min read -
Streaming Serverless Responses with Server-Sent Events (SSE)
Streaming is the single biggest lever on perceived latency. Here's how to stream responses from serverless functions with SSE — async generators, LLM token streaming, disconnect handling, and heartbeats — and when not to.
Aleksandr Kubarskii · · 8 min read -
A Postgres-Backed Job Queue You Don't Have to Run
Move slow work off the request path without standing up Redis and a worker fleet. A durable, Postgres-backed job queue with retries, backoff, a dead-letter path, and visibility-timeout reaping — built in.
Aleksandr Kubarskii · · 4 min read -
Verify Webhook Signatures at the Gateway
Signature verification is the first line of defense for any webhook — and the step teams most often get subtly wrong. Here's how to move it off your handler and onto the gateway for GitHub and Stripe.
Aleksandr Kubarskii · · 5 min read -
Cold Starts, Warm Starts, and Hot Containers Explained
Understand cold starts, warm starts, and hot containers in serverless platforms, and learn how they affect latency, cost, and backend architecture.
Aleksandr Kubarskii · · 4 min read -
Edge Functions vs Container-Backed Functions
Edge functions and container-backed functions solve different problems. Learn when to use edge isolates and when full containers are a better backend runtime.
Aleksandr Kubarskii · · 3 min read -
Vercel Functions Alternative for Backend Jobs and Webhooks
Vercel is excellent for frontend apps, but backend jobs, webhooks, cron tasks, and AI agent tools often need a different runtime.
Aleksandr Kubarskii · · 4 min read -
Serverless Background Jobs: Patterns, Pitfalls, and Examples
A practical guide to serverless background jobs: when to use them, common patterns, pitfalls, async APIs, retries, logs, and examples.
Aleksandr Kubarskii · · 4 min read
