How Nouns DAO made hundreds of hours of voice and chat searchable with a RAG assistant

The challenge

As the DAO grew, its knowledge outpaced its memory. Governance decisions were discussed across voice calls, chat channels, and proposal documents, with no central record connecting them. That created three recurring problems:

• Voice conversations were effectively write-only. Once a call ended, the reasoning behind a decision was locked inside a recording nobody would re-listen to.
• Context lived in different places. Answering "what did we decide about X, and why?" meant searching chat history, proposal text, and people's memories separately.
• New and part-time members had no practical way to catch up without asking questions the group had already answered.

Roko AI set out to build a governance knowledge hub for the DAO and brought in Eventum to engineer it. A dedicated Eventum engineer owned the design and build of the pipeline and chatbot.

The solution

Eventum built an ingestion-to-answer pipeline that turns the DAO's raw activity into a queryable knowledge base:

• Voice capture and transcription: meetings are recorded and transcribed, then summarized so each session produces both a full transcript and a concise record of what was discussed and decided.
• Auto-indexing into a central knowledge base: transcripts, summaries, and chat data are automatically chunked, embedded, and indexed as they are created. No one has to file anything; the knowledge base stays current on its own.
• RAG-powered chatbot in Discord: members ask questions in natural language where they already work. The bot retrieves the relevant transcript and chat passages and answers from them, so responses are grounded in what was actually said rather than the model's guesswork.
• Web app for search and uploads: a companion app lets members search the knowledge base directly and upload additional audio for transcription, summarization, and indexing, so contributions from outside the standard meetings still make it into the record.

The result

The DAO went from scattered, unsearchable history to a single knowledge base that answers questions conversationally. Members ask the bot instead of scrubbing recordings or scrolling channels; the reasoning behind past decisions is retrievable in seconds and onboarding a new contributor no longer depends on someone else's memory.

Why it mattered

"Chat with your data" only works if the data pipeline behind it is trustworthy. The visible part of this system is a chatbot; the engineering was in everything upstream: reliable transcription, summarization that preserves decisions, automatic indexing that keeps the knowledge base current without human effort, and retrieval that grounds every answer in a real source. That is the difference between a demo bot and a system a community actually governs with.

Conclusion

Eventum engineered the pipeline that turned Nouns DAO's voice calls and chat history into living, queryable knowledge: automatic transcription and summarization, a self-updating indexed knowledge base, and a RAG chatbot that meets members inside Discord. Scattered conversations became institutional memory.