So for about a year now EF-Map has been running an indexer against the EVE Frontier chain. Every structure anchored, every item deposited into storage, every gate jump, every kill, it all lands in a Postgres database on a VPS. That database is a genuine wealth of information, and up until this week almost all of it reached you through snapshots. A cron job runs a fixed query every few minutes, publishes the result, and the site displays it. Gate Activity works that way. The Economy panel works that way. The intelligence dossiers work that way.
Snapshots are fine as far as they go, but there is a problem baked into them: I am choosing what you get to see. I write the query, I pick the columns, I decide the time windows. If you want to know something I did not think of, you are stuck. And people always want to know something I did not think of. That is the whole nature of intelligence in this game.
Where this had to go
The first crack at live queries was the SSU Finder, which searches actual storage contents on demand rather than reading a snapshot. Building that taught me two things. The first was that people really do want to interrogate the database directly. The second was that letting the public internet run queries against your database is something you need to think about quite carefully, because that is where we first ran into concurrency work on the indexer pipeline.
So I have known for a while that live database access was where intelligence on EF-Map had to end up. Not snapshots of what I think you want, but the ability to ask your own questions. The blocker was never really technical, it was economic. A live query service for anyone who wanders past does not scale, and I did not fancy finding that out the hard way. It needed to sit behind some kind of membership so the numbers stay controlled and the people using it are people who actually want it.
That piece arrived a couple of weeks ago with EF-Map Intelligence, the Frontier Commerce paywall that gates the intel tools behind a token payment of 100 testnet EVE for 30 days. It has been running for over a week now, we found and fixed a few bugs, and it seems to be holding up well. Which meant the thing I actually wanted to build was finally buildable.
Version 0.1 of the Query Console
The Query Console is a new tool in the Tools Library, in the Intel / Activity section, for Intelligence subscribers. It lets you query the indexer database live. There are eight datasets at launch: current stockpiles, structures, item flows in and out of storage, items minted onto and burned off the chain, the full activity event stream, gate jumps, kills, and fuel events. You pick a dataset, add filters, and either list matching rows or build a leaderboard, grouped by whatever you like and ranked by a count or a sum. Results come back as a table or a bar chart, and you can copy any of it out as CSV.
I want to be straight about what this is: it is version 0.1. The browser never sends SQL, it sends a small structured query that the server validates against a fixed catalogue and compiles with hard limits, a five second timeout, capped row counts, read-only access. In practice most queries come back in well under a second, the typical ones benchmark around 40 milliseconds. The database side has plenty of headroom, the current subscriber base could grow a hundred times before I would need to think about it again.
There are preset questions to get you started. Who has the most Building Foam stored. Which storage structures are busiest as markets. Busiest gates this week. Click one, it loads into the builder, you tweak it and run it. That teaches the tool better than any manual would.
The bit I am actually excited about
Here is the thing though. I sat looking at the finished builder and thought, this works, but it looks like a database query tool. Because it is one. You need to know that fuel lives in the stockpiles dataset, that you filter on item name, that you want a sum grouped by owner. Some of the audience will be completely comfortable with that. A lot of it will not be, and making this stuff accessible to as wide an audience as possible has kind of been the point of EF-Map from the start.
Then I remembered the AI command work we did last year for the map, and asked my coding agent whether the same approach could drive this. The answer was yes, and almost embarrassingly easily. There is now a box at the top of the console where you just type your question in plain English. Which storage units have the most fuel stored in them. What can you tell me about this tribe. How many catapults have the Reapers built. A small language model running on Cloudflare Workers AI reads the live dataset catalogue, writes the query, and loads it into the builder for you. You see exactly what it built, you can change anything, and nothing runs until you press the button. Because the model can only ever produce a query the validator would accept from a human anyway, it gets no special power. The inference cost is negligible at this scale, fractions of a penny per question.
I will be upfront: it will stumble. It is not going to take every phrasing and hand you a perfect query first time. During my own testing it confidently decided a catapult was a type of storage unit, which took ten minutes to fix once I spotted it. That is actually the plan for improving it. Every question asked is logged anonymously, the question and what got built, never who asked it, in line with the aggregate-only approach the site has always taken. Every week or so I can review what people asked, what worked, and crucially what failed. If lots of people are asking for something the datasets do not cover, that is my roadmap written for me, in your words. So the tool gets better through being used, which I think is the best way to say it.
Bring your own AI
The other thing I added, and I think this one is a bit novel: there is a Download a briefing file link next to the ask box. It generates a markdown document describing every dataset, field and filter, straight from the live catalogue so it is never out of date. You paste that into your own ChatGPT or Claude or whatever you already pay for, and have a proper conversation with it. What could I look up with this. Can I find out who is stockpiling in my region. Your AI reads the briefing and tells you exactly what to type into the console, either the plain English question or the builder settings. All the open-ended exploration happens on your subscription rather than mine, which keeps my costs flat and gives you a much better chat experience than I could sensibly host.
Why bother
Is this better than running your own indexer on your own VPS and building your own tables? Of course not. If you have those skills, do that, it is a genuinely fun project. But that is exactly the gap this is aimed at. EVE Frontier is a blockchain game and there is a persistent feeling that you need programming knowledge to get at the good information, or need to be in the right tribe with the right coder. For some things that is still true. But the goal with EF-Map has always been to take what the chain makes public and put it in everyone's hands, and this is the most direct version of that yet: the whole indexed history of the current cycle, queryable by anyone, for a small EVE payment, in plain English if that is how you prefer to work.
It is live now for Intelligence subscribers, in the Tools Library under Intel / Activity, with a full help section behind the question mark in the panel header. Ask it something strange. The failed queries are honestly as valuable to me as the successful ones right now, and I will be reading them.
Related Posts
- EF-Map Intelligence: What It Is, What Stays Free, and Why It Costs 100 Testnet EVE - the paywall this builds on
- Sui Intelligence Search for EVE Frontier Players and Tribes - the indexer behind the datasets
- Gate Activity: Who Owns the Jump Network in EVE Frontier - snapshot intelligence, the previous model
- Privacy-First Analytics: Learning Without Tracking - the anonymous logging approach the query log follows