Daily recap for May 27, 2026

Today we shipped three new step-by-step tutorials for small-business AI, plus three monorepo ideas covering agent evaluation and streaming data pipelines. Here's what you can try right now.

New tutorials

Ollama agent mesh for small business workflow automation

This tutorial shows how to set up an Express server that uses a mesh of local LLM specialists to automate email triage, CRM updates, and report generation—all without sending data outside your network. It covers intent classification, confidence-based routing, session continuity, and agent handoffs, so you can orchestrate complex workflows on your own hardware. Perfect for small businesses that handle sensitive client data and want to avoid cloud AI services.

Read the tutorial → Download the code (zip)

Ollama local models, Express, 139 tests, 97.19% coverage.

AWS Bedrock document pipeline for AppFolio lease data extraction

This pipeline automates lease data extraction from PDFs and scanned images into AppFolio. It uses unpdf or Tesseract OCR for text extraction, AWS Bedrock (Claude) for structured data parsing, and a JSON repair layer to ensure clean records. A firewall prevents unauthorized API calls, and a budget engine keeps costs predictable. Ideal for property managers tired of manual data entry.

Read the tutorial → Download the code (zip)

AWS Bedrock, Express, 97 tests, 100% coverage.

OpenAI voice agent for OpenTable reservation management

Build a voice agent that lets diners manage OpenTable reservations by phone. It uses Twilio for calls, Deepgram for speech-to-text, OpenAI for intent recognition, and Cartesia for text-to-speech, all orchestrated through REAA's voice-agent-core. The agent handles reservation creation, modification, and cancellation via the OpenTable API, with session continuity for natural multi-turn conversations. Restaurants can finally handle after-hours bookings without staff.

Read the tutorial → Download the code (zip)

OpenAI, Fastify, 127 tests, 98.96% coverage.

Synthetic user simulation for agent evaluation

An early-stage monorepo idea that provides a framework for simulating realistic user interactions to stress-test AI agents across end-to-end scenarios. Use it to evaluate agent performance before putting them into production.

Read the tutorial →

Agent eval archetype, built with TypeScript, pnpm, vitest, zod, CLI.

Voice agent testing and quality assurance toolkit

A testing toolkit for voice agents, covering speech-to-text accuracy, latency benchmarks, and scenario-based call simulations. Useful for ensuring quality in production voice systems.

Read the tutorial →

Voice agent archetype, built with TypeScript, pnpm, vitest, zod, CLI.

Real-time streaming data pipelines for AI agents

A blueprint for building streaming data pipelines that feed real-time events and context to AI agents with exactly-once semantics. Designed for high-throughput agent workflows.

Read the tutorial →

Data pipelines archetype, built with TypeScript, pnpm, vitest, zod, CLI.

Browse all tutorials at reaatech.com/solutions.