Resum

Product Introduction

1. Definition: Resum is an enterprise-grade API and SDK developed by Viroke Technologies Inc. designed for live conversations. It functions as a speech-to-text pipeline that processes real-time audio streams from microphones, local files, or RTMP live streams, connecting live dialogue directly to named AI agents.
2. Core Value Proposition: Resum exists to automate post-conversation documentation and analysis by capturing live audio, generating a live SRT transcript during active sessions, and automatically compiling dedicated output files per named agent (such as structured summaries and notes) using LLMs like GPT-4o-mini upon session termination.

Main Features

1. Multi-Source Real-Time Audio Capture: Resum opens a secure speech socket and captures audio inputs via microphone (avfoundation), local recordings paced in real time using FFmpeg, or external RTMP live streams. FFmpeg normalizes the input stream into mono 16 kHz AAC chunks before routing them simultaneously to local storage and the speech-to-text pipeline.
2. Streaming Speech-to-Text and Live Transcription: Powered by AssemblyAI via WebSocket connections, Resum processes audio streams in ten-second chunks. Finalized conversational turns are appended sequentially to a transcript file (transcript.srt) with dedicated cue times while the session remains active.
3. Automated Multi-Agent Post-Processing: Upon session termination—triggered manually via Ctrl+E, expiration of the default 60-minute duration, or a stop-request.json file—Resum leverages OpenAI's GPT-4o-mini to generate structured files for each designated agent, such as notes.txt (excerpts, next steps, action items) and summary.txt (highlights, speakers, full summary), alongside a session.json index and a stateful control.json audit trail.

Problems Solved

1. Pain Point: Enterprise teams face severe friction converting live meetings, customer calls, and audio streams into structured documentation without manual transcription overhead or disconnected multi-step workflows. Resum eliminates this by fusing live speech-to-text transcription with autonomous, role-specific AI agents in a single API/SDK call.
2. Target Audience: Enterprise Software Engineers, DevOps Professionals, Backend Developers, and AI Integration Architects building real-time voice applications, automated meeting assistants, or enterprise compliance logging infrastructure.
3. Use Cases: Automated generation of meeting minutes and executive summaries for enterprise conferencing platforms; real-time transcription and compliance tagging for live customer support call centers; automated processing of audio feeds from RTMP streams for media intelligence and content indexing.

Unique Advantages

1. Strengths & Limitations (Pros & Cons):
   • Pros: Unified API and terminal runner handling both streaming capture and agent orchestration; zero-logging policy for sensitive environment variables (STT_KEY, OPENAI_API_KEY); guarantees finalized turn appending in real time; produces clean, standardized disk outputs (audio.aac, transcript.srt, notes.txt, summary.txt, session.json, control.json).
   • Cons: Offered exclusively to enterprise clients through direct sales without a self-serve tier or public pricing model; requires external API keys for AssemblyAI and OpenAI; limited to specific predefined agent tags and fixed file schemas out of the box.
2. Key Alternatives & Differentiation:
   • Alternative 1 (Custom AssemblyAI + OpenAI API Integration): Building an in-house pipeline requires manual handling of FFmpeg stream splitting, WebSocket state management for AssemblyAI, chunk buffering, and custom prompt orchestration for GPT-4o-mini. Resum encapsulates this entire lifecycle into a single TypeScript SDK and terminal runner.
   • Alternative 2 (Commercial Meeting Assistants like Otter.ai or Fireflies): These consumer and SaaS products offer end-user web apps but lack deep programmatic embeddability, headless server-side SDK control, and enterprise custom data-sink governance that Resum provides natively via API.

Frequently Asked Questions (FAQ)

1. What is Resum and who is it built for? Resum is an enterprise API and SDK created by Viroke Technologies Inc. designed for software engineering teams and enterprises that need to connect live conversational audio streams directly to named AI agents for real-time transcription and automated file generation.
2. What audio sources are supported by Resum? Resum supports live microphone capture (such as avfoundation devices), local audio recording files paced in real time via FFmpeg, and external RTMP live stream URLs.
3. How does Resum handle security and API keys? Resum keeps sensitive credentials secure within your local environment variables (STT_KEY for live sessions and OPENAI_API_KEY for post-session agent processing) and ensures that secret values are never written to execution logs.
4. How are agent files generated after a session ends? Once a session is terminated via keyboard command, duration limit, or stop request, Resum flushes the audio buffer, transmits a termination signal, and uses GPT-4o-mini to automatically write structured output files like notes.txt and summary.txt based on the finalized transcript.srt.
5. How can enterprises purchase or license Resum? Resum is not available via self-serve signups or public pricing tiers; licensing agreements, enterprise terms, and pricing are managed directly through custom consultations with Viroke Technologies Inc. sales.