Daily recap for June 12, 2026

Today we shipped eight new step-by-step tutorials for small-business AI, from automated regression testing for support bots to multi-agent order management for e-commerce.

New tutorials

Automated regression testing for SMB support bots

Small businesses deploying AI support bots need a reliable way to catch regressions before they reach customers. This tutorial walks through building a CI-friendly evaluation pipeline that replays golden conversations, scores responses with an LLM judge routed through Vercel AI Gateway, enforces quality gates, and exports traces to Langfuse. A failing gate halts CI with a non-zero exit code, so bad answers never ship.

Read the tutorial → · Download the code (zip)

Built with @reaatech/agent-eval-harness-cli, @reaatech/agent-eval-harness-suite, @reaatech/agent-eval-harness-gate, @reaatech/llm-cost-telemetry, @reaatech/llm-cache, @reaatech/structured-repair-core, on Vercel AI Gateway · 87 tests, 98.43% coverage.

RAG pipeline for insurance policy analysis

Insurance brokers waste hours manually searching policy PDFs to answer coverage questions. This tutorial shows how to build a retrieval-augmented generation service that lets brokers query documents with natural language. You'll upload PDFs, have them chunked and embedded via VoyageAI, stored in pgvector, and answered by a Databricks-hosted LLM—with semantic caching, context-window planning, and per-tenant cost tracking.

Read the tutorial → · Download the code (zip)

Built with @reaatech/agent-memory-retrieval, @reaatech/agent-memory-storage, @reaatech/llm-cache, @reaatech/context-window-planner, @reaatech/structured-repair-core, @reaatech/llm-cost-telemetry, on Databricks · 111 tests, 96.55% coverage.

Multi-agent order management for e-commerce

Small online retailers handle repetitive customer queries about orders, shipping, and returns. This tutorial builds a multi-agent mesh orchestrated by @reaatech/agent-mesh and powered by a vLLM-hosted model. Specialist agents classify intents, share conversation context, and maintain multi-turn sessions—no more lost handoffs or inconsistent responses.

Read the tutorial → · Download the code (zip)

Built with @reaatech/agent-mesh, @reaatech/agent-mesh-registry, @reaatech/agent-mesh-router, @reaatech/agent-handoff, @reaatech/session-continuity, @reaatech/llm-cost-telemetry, @reaatech/structured-repair-core, on vLLM · 165 tests, 99.42% coverage.

Auto-takeoff agent for GC bid prep

For construction estimators, manually measuring plan sets and spec docs to produce a bill of materials and subcontractor RFPs takes days. This agent automates that: it runs OCR on PDFs and images, uses an LLM to extract quantities and generate structured BOMs, groups line items by trade, and produces subcontractor RFPs—with budget enforcement and LLM caching built in.

Read the tutorial → · Download the code (zip)

Built with @reaatech/media-pipeline-mcp-doc-extraction, @reaatech/agent-mesh, @reaatech/llm-cache, @reaatech/agent-budget-engine, @reaatech/a2a-reference-persistence, @reaatech/agents-markdown, on agnostic · 127 tests, 96.48% coverage.

Field punch-list agent for superintendents

A superintendent on a jobsite captures defects via photo and voice memo, but manually transcribing them into project management tools causes delays. This tutorial builds a mobile-first agent that ingests audio and images, transcribes with Deepgram, extracts punch items with an LLM, runs guardrails, and syncs to the PM system via webhook—so items don't get lost.

Read the tutorial → · Download the code (zip)

Built with @reaatech/voice-agent-core, @reaatech/media-pipeline-mcp-deepgram, @reaatech/agent-memory, @reaatech/webhook-relay-webhooks, @reaatech/guardrail-chain, @reaatech/a2a-reference-auth, on agnostic · 135 tests, 95.53% coverage.

Anthropic lead intake for small law firms

Small law firms miss leads after hours and spend time manually qualifying inquiries. This tutorial builds a conversational intake agent powered by Claude that qualifies potential clients, captures case details, routes to the right practice area, and creates matters in Clio—all with cost tracking, idempotency, and session continuity via DynamoDB.

Read the tutorial → · Download the code (zip)

Built with @reaatech/confidence-router, @reaatech/confidence-router-core, @reaatech/agent-handoff, @reaatech/agent-handoff-routing, @reaatech/structured-repair-core, @reaatech/session-continuity, @reaatech/session-continuity-storage-dynamodb, @reaatech/llm-cost-telemetry, on Anthropic · 88 tests, 97.46% coverage.

AWS Bedrock knowledge agent for property managers

Property managers spend hours answering repetitive tenant questions about leases, payments, and maintenance. This agent uses retrieval-augmented generation over Buildium data and property documents, with semantic caching and multi-turn sessions. It runs on AWS Bedrock and sends answers via a chat API—so tenants get answers 24/7.

Read the tutorial → · Download the code (zip)

Built with @reaatech/agent-memory, @reaatech/agent-memory-embedding, @reaatech/agent-memory-storage, @reaatech/session-continuity, @reaatech/session-continuity-storage-dynamodb, @reaatech/llm-cost-telemetry, @reaatech/llm-cache, @reaatech/llm-cache-adapters-qdrant, @reaatech/structured-repair-core, on AWS Bedrock · 131 tests, 98.94% coverage.

Azure AI agent eval harness for support QA

Small businesses using Azure AI chatbots need consistent answer quality as prompts and models change. This tutorial builds an automated evaluation harness that ingests agent logs, scores them with an LLM-as-judge, tracks cost, enforces quality gates, and displays results in a dashboard—all on Azure OpenAI.

Read the tutorial → · Download the code (zip)

Built with @reaatech/agent-eval-harness-suite, @reaatech/agent-eval-harness-cost, @reaatech/agent-eval-harness-gate, @reaatech/agent-eval-harness-judge, on Azure AI · 42 tests, 100% coverage.

Browse all solutions →

Browse the full catalog at reaatech.com/products.