| license: mit | |
| pretty_name: OpenFront.io RL GameRecords | |
| tags: | |
| - reinforcement-learning | |
| - games | |
| - openfront | |
| # OpenFront.io RL game replays | |
| Sparse-turn **GameRecord v0.0.2** JSON for every finished RL episode | |
| (training, watch, showcase). Same format as the OpenFront client archive / | |
| Node bridge (`version`, `gitCommit`, sparse `turns` with intents). | |
| Each parquet row is one game. The `record_json` column is the full Node- | |
| compatible GameRecord string (~100–200 KB for long games). When a watch / | |
| showcase episode recorded policy decisions, `thinking_json` holds a compact | |
| top-3 action trace (typically a few KB) derived from the MODEL overlay | |
| `.debug.json` sidecar. Metadata columns capture policy / engine / stage. | |
| Generated by [djmango/openfront-ai](https://github.com/djmango/openfront-ai). | |
| Replay with `datagen/replay.ts` or the client archive API / `ofshowcase archive`. | |