The rift at eo7

Train your own AI bot inside the game!

You only need neat-python, can be installed by writing 'pip install neat-python' inside command prompt

This game is a work in progress, would love to get feedback on the game.. I am making it with the help of Codex/ChatGPT

## Races

Humans have higher range and higher armor, but lower raw damage. Their ranged soldier is the Ranger, and their ranged hero is Sable.

Aliens have higher damage and faster pressure, but shorter range and lighter armor. Their ranged soldier is the Spitter, and their ranged hero is Nihl Sporeseer.

## Current Gameplay Loop

You start with a command building, workers, three soldiers, and a small stockpile. Every Aether node now contains more than 4000 resources, and your resource information appears in the top bar. Harvest Aether, train race units, build Houses for unit capacity, build a Barracks and Armory, unlock tech upgrades, recruit heroes, and push across the map to destroy the enemy Command.

The enemy now has its own workers, gathers resources, expands with extra Commands when nearby resources dry up, builds Houses, Barracks, Armories, and Towers based on difficulty, trains soldiers and heroes, researches upgrades after its Armory is ready, uses hero abilities, and attacks your nearby units and buildings instead of only rushing the Command.

Heroes on both sides gain experience from killing enemy units. Hero max level is 100, with health, damage, armor, and ranged reach improving as they level.

Player campaign heroes now begin each campaign map pre-trained. Their level and XP are stored on the active user profile, so a hero that reaches level 8 on map 2 keeps that progress on map 3 and later maps.

Heroes also have Energy, separate from HP. Energy regenerates over time and is spent when you click hero abilities in the command panel above the minimap. The panel only appears for the selected player hero, and each hero now has their own three visible abilities. If a hero dies, the match ends immediately for that side.

The game also has procedural pixel-art sprites, lightweight sound effects, fog of war, a click-to-pan minimap, hover tooltips, green/red building placement previews, obstacle-aware movement around buildings, save/load support, and scenario JSON support. Workers can build extra Command buildings, Houses, Churches, and defensive towers. Churches train healers, who automatically restore wounded nearby allies. Workers deposit resources at the nearest Command.

Unit capacity starts from Command buildings, with each Command adding 10 space. Each House adds 5 more space, up to a maximum cap of 200. Tower upgrades take time, show visible progress over the tower, and complete into the next tower stage when the timer finishes.

`ARM` in the selected-unit panel means armor. Higher armor reduces incoming damage from enemy units.

## Users, Saves, And Rewards

Use `User / Profile` to create or switch users. Each user keeps their own campaign unlocks, completed maps, high scores, gold, hero inventory, and saved-game list. Campaign progress saves automatically after each completed map; save slots are created with `F5` and loaded from `Saved Games` or `F9`.

Player score rises from gathering resources, training units, building structures, researching upgrades, upgrading towers, using hero abilities, leveling heroes, and killing enemies. Longer games apply a small time penalty, but score only reaches 0 if a side has no living units or buildings.

Winning gives at least 10 gold. Losing still gives at least 1 gold. Setting a new high score for the current map gives 100 bonus gold and user XP. User level caps at 1000, and every 10 user levels creates a reward pick from 10 hero items. Reward items go into the user's hero inventory and improve hero armor, attack, or ranged reach when equipped. Head slots take helmets, hand slots take weapons or bracelets, and feet slots take boots. The inventory starts at 30 slots; buy +5 more slots from the profile screen with gold.

## Campaign Maps

There are 12 campaign maps. Winning the current map unlocks the next one; you cannot progress to the next map before victory. Before each map begins, a scrolling story briefing appears.

- Map 1: Eos-7 Landing. The enemy begins with a passive fortified ridge-city, two upgraded towers on high ground, all four heroes, and a first-contact dialogue when your units approach.

- Map 2: Glass Dunes

- Map 3: The Broken Relay

- Map 4: Crimson Sink

- Map 5: Spine Valley

- Map 6: Aether Orchard

- Map 7: Silent Harbor

- Map 8: The Root Network

- Map 9: Obsidian Gate

- Map 10: Storm Crown

- Map 11: The Heart Wound

- Map 12: Rift Core

## AI Training Lab

The main menu includes `Train AI (NEAT)`. It opens a watchable self-play training lab where genomes evolve economy, building, hero, upgrade, and aggression priorities against one another. The lab starts paused so generations do not rush by; click the Population field, type a number from 2 to 120, and press Enter to rebuild the trainer before running. Use `Step Gen` for one generation or `Run Training` to keep training at a slower visible pace.

Use `Watch Match` from that screen to watch the current population matchup play an actual RTS match on the game map. If no generation is queued, Watch Match starts a full-population generation and shows the first scheduled genome pair. Leaving the viewer with `Esc` keeps that same match alive headlessly in the training lab, and `Resume Match` returns to it without restarting. When a match ends, the trainer records both genomes' fitness and immediately starts the next scheduled pair whether you are watching or not. During the match, `S` cycles 1x, 2x, and fast speed, `Space` pauses, and `R` restarts the current matchup.

When you are on the training screen and not watching, the current matchup continues headlessly instead of blocking the UI. The generation number only advances after every scheduled population pairing has finished. The training screen and live match overlay both show recent genome fitness values so you can see whether the population is actually changing. Use `Save AI` or press `S` in the training lab to export the current best agent into its own folder under `saves/neat_agents`; the folder name includes generation, genome number, and fitness.

If `neat-python` is installed, the lab uses a real NEAT population and config from `neat_selfplay_config.txt`. If the dependency is missing, it falls back to the built-in visual trainer so the screen remains usable.

## Story Direction

The suggested premise is **The Rift at Eos-7**. Humans land on Eos-7 to mine Aether crystals and save a failing colony. The Aliens are the planet's buried intelligence waking up because the mining operation is hurting a living world.

Both sides should feel playable and understandable: Humans are desperate colonists trying to survive, while Aliens are defending their home from an empire that does not understand what it has found.