Major Updates for Proton!!!

🎲 Major Update — RPGX Proton v2.1.0 & AI Assistant v3.1.0 Are Live!

Hey everyone — big drop today. Both RPGX Proton and the RPGX AI Assistant have received major updates, and I'm really excited about this one.

This has been a significant amount of work under the hood, and I think you're going to feel the difference immediately. Let me break down what's new.

---

🧠 The AI Actually Knows Your World Now!! This is the headline feature of this update, and honestly, it was a bug fix as much as anything else.

A security change in Proton 2.1.0 introduced a token authentication system to protect your local RAG server. The problem was that the AI Assistant module wasn't sending that token — so every single query was being silently rejected and falling back to a generic Ollama response with zero knowledge of your world.

If your AI has felt "dumb" lately and keeps saying your custom characters don't exist — this is why, and it's fixed.

After updating, there's a one-time setup step to connect everything back up (details below). Once that's done, your AI will be back to knowing your world's lore, your characters, your session history — all of it. And, better than before.

---

🔒 Security Overhaul (The Boring-But-Important Stuff)

A supporter sent me a thorough security audit of Proton's local server, and honestly, it was really solid work. Thank you — Jakub.

The short version: Proton's RAG server was accessible to anything on your local network, with no authentication. That meant in theory another device on the same WiFi could read your campaign notes, wipe your knowledge base, or inject fake responses into your Foundry chat. Not great.

Here's what's been fixed:

- The server now only listens on your own machine. Nothing on your network can reach it unless you explicitly open it.

- Every request now requires a secret token that's unique to your Proton install. Your Foundry module needs this token to talk to Proton — it's a one-time copy/paste during setup.

- CORS is now restricted to only your configured Foundry URL. Random websites can't reach your local server from your browser.

- Several smaller hardening improvements across the board. For most of you, this is invisible — everything works the same, just safely.

---

🚀 RAG Engine — Completely Rebuilt

Even beyond the bug fix, the retrieval engine that powers world-aware answers has been significantly upgraded. This is what actually finds the right information in your ingested journals and hands it to the AI.

What changed:

- Smarter chunking. The new chunker respects actual paragraph boundaries instead of blindly slicing text every 400 characters. Your lore is stored in meaningful pieces now, not arbitrary cuts mid-sentence. Dense content like character sheets is handled with a special fallback that keeps it accessible.

- Keyword search + semantic search, combined. The old system only used AI embeddings to find relevant chunks. The new system also runs a keyword search pass (the same algorithm that powers Google and Elasticsearch) and merges the two results. This is huge for proper nouns — character names, place names, spell names — which embeddings alone often miss.

- AI-powered reranking. Before the final answer is generated, the top candidate chunks are scored by a second AI pass that judges which ones actually answer the question. Irrelevant context gets filtered out before it can confuse the main model.

- Session journal boost. Queries about recent events ("what happened last session?", "who did we just find out about?") now correctly prioritize your session recap journals. Before this, character sheets were always outranking session notes even when the session note had the actual answer.

Bottom line: Queries about your world should be noticeably more accurate, more specific, and less likely to hallucinate.

> ⚠️ Re-ingest your world after updating. Your existing ingested data was

> built with the old chunker. A fresh ingest will get you all the improvements.

> Open Proton → Worlds, wipe your world, and re-ingest from the Foundry module.

> It only takes a few minutes.

---

✨ Quality of Life

A bunch of smaller improvements that have been on the list for a while:

- Ollama starts automatically when Proton launches — no more clicking "Start Ollama" every time

- RAG Server Token is now displayed in Proton Settings with a copy button so you can easily grab it for the module setup

- Collapsible world sections in the Worlds page — no more scrolling through 500 entries to get to another world

- Copy button on the server log for easy debugging and bug reports

- Quit Proton button on the Dashboard — clean shutdown without hunting for the system tray

- Report a Bug button added to the About page

---

🎯 AI Assistant — AI Model Tier System

The RPGX AI Assistant module has also been updated to v3.1.0 with a new

AI Model Tier dropdown that now matches the Quest Log — consistent across all RPGX modules:

| Tier | Model | RAM |

| Lite | qwen2.5:3b | 4–6 GB |

| Standard ⭐ | qwen2.5:7b | 8 GB+ |

| Performance | qwen2.5:14b | 16 GB+ |

| Ultra | qwen3:30b | 32 GB+ |

Standard (7b) is now the default — a better starting point for most systems.

The setup guide in the module has been fully rewritten to reflect all the current changes, including a dedicated RPGX Proton setup section.

---

⚙️ Setup Steps After Updating

This is the only manual step required — takes about 60 seconds.

1. Update RPGX Proton to v2.1.0

2. Update RPGX AI Assistant to v3.1.0 in Foundry

3. Launch Proton and go to Settings

4. Copy the RAG Server Token shown there

5. In Foundry, open Settings → RPGX AI Assistant

6. Paste the token into the RAG Server Token field and save

7. Wipe and re-ingest your world from the Knowledge Base panel

Forge / remote Foundry users: Also make sure your Foundry URL in

Proton Settings is set to your actual server address, not localhost.

---

## 🔜 What's Coming Next

- PDF ingestion — ingest PDF files directly into your knowledge base

- Image-to-text ingestion — drop in artwork, maps, or handouts and the AI will be able to reference what's in them

---

As always, thank you for the support — this stuff genuinely wouldn't happen without you. If you run into any issues with the update, hit the Report a Bug button in either app, or drop a message here on Patreon.

Happy adventuring. 🗡️

— Ashton / RPGX Studios