AI Session Archive: Build a Searchable Record of Every Game Session

Why does a TTRPG group need a session archive?

By session twenty, your group has logged eighty hours of play. Half of it is referenced again in passing: "remember when the party met that smuggler in the harbor?" Without an archive, those references are lost. A searchable session archive turns hours of play into a record you can query. The party meets the smuggler in session three; he reappears in session twenty-one. Campaign Memory can keep those appearances together when the GM confirms they are the same person, and it can keep them apart when they are not.

What does the AI archive capture?

For each session Tabletop Arc captures:

• Audio — original recording, retained at your visibility setting (GM-only, members, or public).
• Transcript — speaker-separated, time-coded, segment-level.
• Scenes — semantic clusters of segments with summaries, participants, and locations.
• Entity mentions — every NPC, location, faction, item, quest, or event referenced.
• Lore observations — facts about each entity, with confidence and evidence segments.
• Campaign Memory — the facts, entities, and identity decisions this session contributed to the persistent campaign record.
• Recap — dual-track GM-private and player-safe outputs.

Every layer is searchable, exportable, and linked to the canon ledger.

How is search different from a normal CMS?

Search runs over structured, typed data. You can ask:

• "Every scene where the Iron Crown faction was mentioned."
• "All NPCs introduced in episodes 4–8 with motivation 'revenge'."
• "Every transcript segment where the Lich first names the players."

These are queries against the canon ledger, not full-text matches against unstructured prose. That is the payoff of treating the archive as a memory layer rather than a folder of files.