Daily recap for June 24, 2026

You can now try self-hosted voice agents that book jobs into ServiceTitan, observability stacks for your vLLM servers, resume scoring agents, and five other production AI recipes. Each tutorial comes with a downloadable zip and a step-by-step walkthrough.

New tutorials

vLLM Observability Suite for SMB AI Operations

A prebuilt observability stack that automatically instruments every LLM call to your vLLM endpoint, exports traces to Langfuse, and shows token usage and cost in a dashboard. If you run vLLM for AI inference and want to know where your spend is going, this tutorial gives you OpenTelemetry setup, a SQLite aggregation pipeline, and a Next.js dashboard in a few hundred lines of TypeScript.

Read the tutorial → · Download the code (zip)

Built with @reaatech/otel-genai-semconv-core, @reaatech/otel-genai-semconv-openai, @reaatech/otel-genai-semconv-exporters, @reaatech/llm-cost-telemetry-calculator, @reaatech/llm-cost-telemetry-observability, on vLLM · 61 tests / 100% coverage.

vLLM Voice Agent for After-Hours Small Business Support

Answers after-hours calls with a voice agent that runs entirely on your own vLLM server, so no customer data leaves your infrastructure. It books appointments and handles FAQs over Twilio phone lines, using Deepgram for speech-to-text and Cartesia for speech synthesis. Built with an intent router and Redis-backed sessions, the agent keeps conversation context across call segments.

Read the tutorial → · Download the code (zip)

Built with @reaatech/voice-agent-core, @reaatech/voice-agent-telephony, @reaatech/voice-agent-stt, @reaatech/voice-agent-tts, @reaatech/session-continuity, on vLLM · 101 tests / 97.44% coverage.

Automated Resume Scoring Agent for Boutique Recruiters

For boutique recruiters who spend hours grading resumes, this agent scores up to 200 resumes in seconds using an LLM-powered rubric. You upload PDFs or DOCX files, define weighted criteria, and get back structured scores and an independent quality evaluation. It caches previous evaluations to save costs and tracks token spend per scoring run.

Read the tutorial → · Download the code (zip)

Built with @reaatech/agent-eval-harness-golden, @reaatech/agent-eval-harness-judge, @reaatech/llm-cache, @reaatech/agent-budget-spend-tracker, @reaatech/llm-cost-telemetry, @reaatech/agent-memory, on agnostic · 84 tests / 98.96% coverage.

Interview Coordination Agent for Small Recruiting Teams

Eliminates the email ping-pong of scheduling panel interviews by automatically negotiating availability across a candidate and up to five panelists, then sending Google Calendar invites with Meet links. It checks free/busy data, routes scheduling tasks to specialized sub-agents, and even generates a runbook with alert procedures. Perfect for recruiting coordinators handling dozens of interviews a week.

Read the tutorial → · Download the code (zip)

Built with @reaatech/agent-handoff, @reaatech/agent-handoff-routing, @reaatech/agent-mesh, @reaatech/agent-memory, @reaatech/agent-runbook, @reaatech/agent-runbook-agent, on agnostic · 119 tests / 94.07% coverage.

Cohere Document Pipeline for BigCommerce SMB Order Processing

For small e-commerce merchants on BigCommerce, this pipeline takes emailed purchase orders (PDFs, images) and creates orders directly in your store. A Fastify server extracts text, classifies it into structured order data using Cohere’s Command A model, repairs any malformed output, and pushes it to BigCommerce — all while caching product lookups and tracking LLM costs in a dashboard.

Read the tutorial → · Download the code (zip)

Built with @reaatech/media-pipeline-mcp-doc-extraction, @reaatech/structured-repair-core, @reaatech/llm-cache, @reaatech/llm-cost-telemetry, on Cohere · 113 tests / 95.86% coverage.

Databricks Code Sandbox for NetSuite SMB Financial Modeling

Turns plain-English questions like “show me Q3 revenue by region” into safe Python scripts run against your NetSuite data. A Hono-hosted API spins up an E2B sandbox, generates code via Databricks LLM, repairs syntax issues, and enforces per-query budget caps. No spreadsheets, no manual data exports.

Read the tutorial → · Download the code (zip)

Built with @reaatech/agent-budget-engine, @reaatech/structured-repair-core, @reaatech/confidence-router, @reaatech/llm-cost-telemetry, on Databricks · 83 tests / 98.87% coverage.

Mistral AI Voice Agent for ServiceTitan Field Service Scheduling

Field service companies can stop losing after-hours jobs to voicemail. This voice agent answers calls, converses using Mistral AI, extracts job details, and books them directly into ServiceTitan’s REST API. It runs on a Next.js app with Deepgram speech-to-text and LLM caching to keep costs predictable.

Read the tutorial → · Download the code (zip)

Built with @reaatech/voice-agent-core, @reaatech/voice-agent-telephony, @reaatech/voice-agent-stt, @reaatech/voice-agent-tts, @reaatech/llm-cache, @reaatech/llm-cost-telemetry, on Mistral · 80 tests / 99.15% coverage.

Anthropic Salesforce Contract Extraction for SMB Sales

For sales teams tired of retyping contract details, this pipeline pulls PDFs and images from Salesforce, processes them through OCR or text extraction, and uses Anthropic’s Claude to output clean structured fields like contract value and effective dates. A repair engine fixes any malformed JSON, and a budget engine caps cost per document.

Read the tutorial → · Download the code (zip)

Built with @reaatech/media-pipeline-mcp-core, @reaatech/session-continuity, @reaatech/media-pipeline-mcp-doc-extraction, @reaatech/structured-repair-core, @reaatech/agent-budget-engine, on Anthropic · 76 tests / 100% coverage.

Browse all solutions →

Browse the full tutorial catalog at reaatech.com/solutions.