AI/LLMs I built a free web tool to automate 5e prep (Visual Encounter Graph, Cursed Items, Shops). Looking for veteran feedback!

[Cmcccombs01]

A Heartfelt Thank You to This Community

I wanted to post an update here because the blunt feedback on this thread on EN World has fundamentally shaped DM Co-Pilot into what it is today. You pushed me to move away from text generation and tackle the actual "Bridge Problem"—transforming AI from a passive assistant into an active workflow tool for DMs.

Today, we are launching DM Co-Pilot on Product Hunt! I'm here to ask for your vote and, more importantly, your comment. I want the Product Hunt community to see how this tool was built on veteran feedback. What problem are we solving together, and how has this tool helped you? Your comment would mean the world to me.

Vote for us on Product Hunt here

 

[Cmcccombs01]

The Beacons are Lit: We are LIVE on Product Hunt!

I just want to take a second to thank the veterans in this thread (especially @The Firebird). Your blunt, honest feedback over the last few days completely reshaped this app. Because of you, we ripped out the "JSON walls," built the "Human-First" Markdown stat-block editor, and perfected the v7.5 VTT Bridge.

You guys helped me build a tool made by DMs, for DMs.

Today, I'm trying to show the rest of the tech world what we've built. We just went live on Product Hunt. If this tool has helped your weekend prep, it would mean the absolute world to me if you could take 30 seconds to drop an upvote or a comment on our page to help us climb the ranks.

Let's show them the power of the TTRPG community! https://www.producthunt.com/products/dm-co-pilot?utm_source=other&utm_medium=social

 

[Cmcccombs01]

Quick Infrastructure Update: The Monolith is Dead. Long Live the Microservice!

Hey everyone! The reception to the v7.5 "Human-First" refactor (and the new Monster Lab) has been unbelievable. Because of your support, we are officially holding the #1 Global App spot on Viberank, and we hit an all-time high of 450+ concurrent DMs online today!

However, that massive traffic spike introduced a new boss fight: my single monolithic server started to choke under the weight of all the AI vector database queries.

To make sure none of your campaigns crash mid-session, I just spent the last two hours completely decoupling the app into a dual-server Microservice Architecture:

• Server 1 (The Cloud): This handles your UI, so the web app stays lightning fast.
• Server 2 (The Oracle): A dedicated backend that does all the heavy lifting for the AI LLMs.
• The Glue: I deployed a distributed Redis Edge-Cache to link their brains together.

What this means for you: The app should feel infinitely faster, and the JSON payloads for your VTT exports are now fully stabilized.

Now that the plumbing is bulletproof, I am heading back into the codebase to refine the AI Bestiary prompts. (Fun fact: The old AI just tried to tell me a 3rd-level Fireball does 1d6 per caster level... we are fixing that 3.5e hallucination right now!) >Thanks again for the incredible feedback! Let me know how the speed feels today.

 

[The Firebird]

This is a masterclass in constructive criticism—thank you for being blunt. You're 100% right. I’ve been so deep in the server-side 'plumbing' (making the JSON schemas unbreakable for Foundry v12) that I let the 'interior design' get far too technical for actual table use.

Comparison to donjon is the ultimate wake-up call. I’m pausing the roadmap today to prioritize a v7.1 'Human-First' Refactor:

1. Intuitive Labels: Ditching 'Forge' and 'Factory' jargon. If it builds a monster, it’ll be called a Monster Lab.
2. The Verification Loop: No more 'JSON Walls.' The app will now show a clean, formatted Markdown stat block for you to review and edit in-browser before you ever hit the export button.
3. Sidebar Cleanup: Grouping tools by 'Prep' vs. 'Live Play' so you aren't hunting through 36 modules mid-session.

I'd value your eyes on the v7.1 UI when I push it later today. Thanks for helping me move the needle from 'Engineering Project' to 'DM Tool'.

Looking good. I like that the monster lab has a better name and is now editable. It still could be more readable from the human perspective. Is there a way to generate a nice table that the user can see? They could either edit the table directly or edit the text which affects the table.

Then, I had a look at the "Dragon's Hoard". Right now it looks like a LLM call. I think you'd have more user-friendly results if you referenced an actual ruleset--perhaps SRD available--with a python backend for rolling on those tables to generate an actual hoard.

 

[Cmcccombs01]

@The Firebird You are basically writing my roadmap at this point, and I cannot thank you enough for the brutal honesty. You are 100% right on both fronts.

1. The Live-Preview Stat Block: The Markdown text box is a good band-aid, but a split-pane editor where you type on the left and see a beautiful, D&D Beyond-style table update in real-time on the right is the ultimate goal. That is officially locked in for the next UI update.

2. The Python SRD Hoard: This is brilliant. Ripping out the AI prompt entirely for the Dragon's Hoard and replacing it with a hardcoded Python script that actually rolls virtual d100s on the official SRD loot tables will make it infinitely more reliable and instantly familiar to veteran DMs.

I just pushed the v7.5 microservice architecture live to handle the traffic spike from Viberank today, so I am taking a mandatory dev-cooldown to let the servers stabilize. But the second I open VSCode back up for v7.6, those two features are at the absolute top of the backlog.

Keep the feedback coming. It is fundamentally changing how this app works for the better.

 

[Cmcccombs01]

Known Issue: The Oracle is having 3.5e Flashbacks

Full transparency for everyone testing the Real-Time Rules Lawyer today: The microservice bridge is flawless, but the AI model itself is currently acting like a senile wizard. It keeps hallucinating 3.5e rules (like telling me a Fireball does 1d6 per caster level).

I am actively rebuilding its brain to enforce strict 5e RAW (Rules as Written). Until then, treat the Lawyer as a highly opinionated NPC—always double-check its math! Fix coming in the next patch.