Tusk's Tomes — the companion chronicler

Vault is only as good as what you feed it, and the hardest campaign material to produce is a record of what actually happened at the table. Notes get thinner every week; nobody writes up session fourteen.

Tomes is the other half: a locally-hosted chronicler that turns a session recording into a written chronicle, saved straight into the folder Vault already reads. Do that and the loop closes — you play, and by the morning the session is a thing your players can ask questions about, cited, in Discord.

That is the whole pitch for the pair. What Tomes is, what it costs, what it needs and how to install it are all explained far better on its own site:

Tusk's Tomes →

Same convictions as Vault: local-first, MIT, no account and no telemetry. Both work perfectly well alone — Vault is happy with lore you wrote by hand, and Tomes is happy writing chronicles wherever you point it. The pairing is a convenience, not a dependency.

Both projects read and write the same on-disk folder. Save a chronicle there from Tomes and Vault needs no import, no sync and no reindex to answer from it.

parent-dir/
├── Tusks-Tomes/        (the chronicler — optional)
├── Tusks-Vault/        (this repo — the answerer)
└── Tusks-Lore/         ← shared knowledge directory
    ├── tusks-lore.json (Tomes' consolidated glossary + speaker map)
    └── Sessions/
        └── <campaign>/
            ├── Session-01-2026-05-19-full.docx
            └── Session-01-2026-05-19-condensed.docx

• Vault walks Tusks-Lore/ recursively on every question — files at any depth surface as lore. There is no cache: the folder is re-read each time, so a chronicle saved a minute ago is already citable. .docx, .pdf, .md, .txt, .html and a few more are all ingested; see INGESTIBLE_EXTENSIONS in loader.ts.
• Tomes writes its chronicle files into Tusks-Lore/Sessions/<campaign>/ when you press Save full .docx (or Save condensed .docx) on a finished chronicle. That click is the only manual step in the loop.
• A few filenames are skipped from generic ingestion, matched on the basename at any depth: tusks-lore.json, glossary.json, speakers.json, the README, dotfiles, _-prefixed files, and Vault's own state files (clarifications.json, lore_gaps.json, tusks-vault.log). Tomes' JSON is structured metadata rather than prose; Vault's own state is worse — left unfiltered the bot reads its list of unanswered questions back as campaign canon and cites it.
• You can use either project standalone. Vault is happy with hand-uploaded lore and no Tomes. Tomes is happy writing chronicles wherever you point it; the pairing is an opt-in convenience, not a hard dependency.

Vault auto-detects the sibling folder via the resolution order documented in How it's built: TUSKS_VAULT_LORE_PATH in the environment, then a path set in settings.json, then a sibling ../Tusks-Lore/, then repo-local Lore/ as the fallback. The resolution happens once at boot, so moving the folder needs a restart.

   Play  →  Record  →  [ Tomes ]  →  Tusks-Lore/  →  [ Vault ]  →  Cited answer
                       transcribes     the shared      reads the      "@Tusk what
                       and writes      folder both     folder on      happened in
                       the chronicle   projects use    every question  session 5?"

Saving the chronicle is the only manual step, and it is one click on the Tomes side. Nothing is imported, synced or reindexed afterwards: Vault re-reads the shared folder on every question, so a chronicle saved a minute ago is already citable, in Discord, in Foundry, or from an MCP client.

Play on Saturday, save on Sunday, and by the time someone asks "wait, what did we actually agree with the harbourmaster?" the answer is there with the chronicle cited next to it.

How Tomes gets from a recording to a chronicle — transcription, speaker attribution, the writing pipeline, what it can run offline — is covered on its own site.