Use cases — Tusk's Vault

🎭 Game Masters

• Mid-session recall without scrubbing through six Google Docs while four players wait.
• Worldbuilding consistency check — query your own setting like an outsider would, and surface contradictions before players do.
• Convert hallway-DM-conversations into canon — record the answer to "wait, is that elf the king's brother or his cousin?" once via Clarifications, and it stays remembered forever (even when future players paraphrase the question).
• Player-onboarding co-pilot — new player joining a 30-session campaign? Tell them to ask Tusk what they want to know about the world, instead of writing them a primer.

🛡️ Players

• Catch up after missing a session without bothering the DM at 1am.
• Verify what you actually agreed to ("did we tell the merchant we'd come back, or was that the other one?") with a citation, so debates stay short.
• Lookup your own character's backstory during a roleplay moment without breaking out of the game.

📜 Worldbuilders

• Stress-test new lore against existing canon — drop a draft document into Tusks-Lore/, ask the bot about it. Inconsistencies surface as contradictions because every claim is cited.
• Find what you forgot you wrote — a 200-page worldbuilding bible becomes searchable for the first time.

🎮 Streamers + actual-play creators

• Answer chat questions live without breaking immersion. "Who's the duke again?" gets a one-line answer in your campaign Discord — or in Foundry's chat bar, if that's where you run — while you stay in the scene.
• Episode recap generator — ask Tusk to summarise the last three sessions for your YouTube description.

The examples below are written as Discord mentions because that is the oldest surface. The same questions work verbatim in Foundry VTT's chat bar as /tusk who is Ser Alric Vane? (see Foundry VTT), and as an ask_lore call from any MCP client — one pipeline answers all three, so the persona, the corpus and the citations are identical whichever you use.

1. Drop-in lore reference

1. Drag a session journal into your lore folder — the sibling Tusks-Lore/ next to the Vault repo, or <repo>/Lore/ if you have not made one.
2. Ask @Tusk about it in Discord.
3. Done.

The bot walks Tusks-Lore/ recursively on every query, so updates are instant — no rebuild, no reindex. Sub-folders are fine; the file's relative path becomes its citation marker.

2. Resolving a lore gap

1. Someone asks something the bot doesn't know. Bot replies: "I am unsure about this detail. I have recorded this as a lore gap for the DM to clarify."
2. The DM opens the dashboard → Home → the campaign card's Lore Gaps tab.
3. Type the canonical answer.
4. The clarification is embedded locally (via MiniLM on CPU) and stored as clarifications.json + clarifications.embeddings.json inside your Tusks-Lore/ folder — so it follows your campaign and survives a clean reinstall of the app.
5. The next time anyone asks anything semantically similar — even if they phrase it differently — the answer surfaces automatically. You only teach Tusk a fact once.

A clarification outranks the knowledge base when the two disagree (rule 4 of the canonical rules block), which is what makes this the way to correct the archive rather than editing a document.

3. Tomes → Vault session loop (with the optional companion installed)

The full Tusk's Tomes + Vault workflow:

1. Play your session.
2. Record the audio (Craig multitrack, OBS, phone — whatever you have).
3. Transcribe with Tomes' optional Audio Transcription module (offline Whisper), or paste in a transcript you produced yourself.
4. Chronicle via Tomes' six-phase LLM pipeline — produces a polished narrative chronicle.
5. Save it into the shared folder. On the finished chronicle in Tomes, Save full .docx writes it to Tusks-Lore/Sessions/<campaign>/. One click, and nothing to import on the Vault side.
6. Query. From that moment on, "what happened in session 5?" cites the chronicle Tomes produced that morning — no rebuild, no reindex.

Each step is independent — you can do steps 3-5 by hand without Tomes and the rest still works.

4. Switching LLM mid-campaign

1. Hit Gemini's free-tier rate limit during a busy session?
2. Open the dashboard → Home → Active Provider picker → switch to another key, to Claude Code, or to Ollama.
3. Next message uses the new provider. No restart, no config file edit.

A rate limit is a statement about the key, not the model — the retry layer holds the whole provider for the backoff, so swapping model within the same provider will not escape a 429. Switch provider, or wait.

5. Reading lore from your Obsidian vault instead

1. Lore → Lore source in the dashboard.
2. Paste the vault's folder path and press Check. Vault reports what it found — notes, folders, frontmatter fields — so you can confirm it is the right one.
3. Press Use this vault.
4. For a large vault, press Build map: Vault reads every note once and reduces each to a line, so later questions carry the whole map plus only the notes they need.

Vault never writes to your vault. Switch back to the folder at any time — nothing is moved or changed. See Obsidian vault.

6. Asking from inside Foundry VTT

1. Dashboard → Home → Surfaces → switch Foundry VTT on.
2. In Foundry, enable the Tusk's Vault module, then Connect to Tusk's Vault from its settings. Both ends show the same six-digit code; check they match and click Allow. There is no token to copy.
3. Anyone the world's Who may ask setting admits types /tusk who runs the harbour? in the chat bar. The answer comes back cited, whispered to the asker and GMs by default.

Two things to decide before you let players use it.

• Who may ask is set per world, in Foundry. Vault holds a separate per-install ceiling (Answer players, not just the GM) that is off by default, so both ends must allow it.
• Vault answers from the whole corpus, with no notion of what you have revealed. A whisper hides an answer from the rest of the table, not from the player who asked.

Foundry VTT covers both properly.