Resum

Overview

Resum is an enterprise API and SDK developed by Viroke Technologies Inc that connects live conversations to named AI agents. The software ingests audio in real time to generate incremental transcripts during an active session, followed by structured text outputs for each designated agent once the session concludes. Audio capture and language model processing execute locally using the customer's own cloud credentials rather than routing through hosted intermediary servers.

Key Features

• Multi-source audio capture — Ingests live audio feeds from local microphones, pre-recorded audio files paced with ffmpeg, or incoming RTMP stream URLs.
• Real-time speech transcription — Converts audio into mono 16 kHz AAC format and streams ten-second chunks via WebSocket directly to AssemblyAI.
• Finalized turn logging — Appends completed conversation turns to a local SubRip Subtitle file while the live session is still in progress.
• Automated agent file generation — Submits the finalized transcript to OpenAI's gpt-4o-mini upon session completion to generate structured summaries, action items, or tagged text files.
• Configurable session controls — Terminates sessions through manual keyboard interrupts, preset duration limits, or file-based stop triggers while flushing remaining audio buffers.
• Client-managed credential architecture — Operates directly within the user's infrastructure using customer-provided API keys without logging secrets or transmitting audio to Viroke Technologies.

Who It's For

Resum is built exclusively for enterprise engineering teams and software developers integrating real-time audio transcription and automated post-session summarization into their systems. It targets organizations that require programmatic audio handling across standard devices or live streams without relying on multi-tenant SaaS platforms. It is particularly relevant for enterprises needing strict control over API credential usage and local data storage.

Notable Strengths

A primary technical strength is Resum's direct-to-provider architecture, which keeps speech-to-text and language model API keys inside the customer's execution environment rather than proxying audio through vendor servers. The SDK also provides a deterministic session lifecycle, saving identical AAC audio bytes and timestamped subtitle files alongside post-session analytical text files upon termination.