boki/poe2-bot
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

added test cases

Browse source
This commit is contained in:
Boki 2026-02-12 13:08:19 -05:00
parent c1892230b7
commit 93e2234c4e
9 changed files with 320 additions and 141 deletions
Download patch file Download diff file Expand all files Collapse all files

80
src/game/ScreenReader.ts
View file

@ -1,7 +1,7 @@
import { mkdir } from 'fs/promises';
import { join } from 'path';
import { logger } from '../util/logger.js';
import { OcrDaemon, type OcrResponse, type OcrEngine, type OcrPreprocess, type DiffOcrResponse, type TemplateMatchResult } from './OcrDaemon.js';
import { OcrDaemon, type OcrResponse, type OcrEngine, type OcrPreprocess, type DiffOcrParams, type DiffOcrResponse, type TemplateMatchResult } from './OcrDaemon.js';
import { GridReader, type GridLayout, type CellCoord } from './GridReader.js';
import type { Region } from '../types.js';
@ -9,11 +9,24 @@ function elapsed(start: number): string {
return `${(performance.now() - start).toFixed(0)}ms`;
}
export interface OcrSettings {
engine: OcrEngine;
screenPreprocess: OcrPreprocess;
tooltipPreprocess: OcrPreprocess;
tooltipParams: DiffOcrParams;
saveDebugImages: boolean;
}
export class ScreenReader {
private daemon = new OcrDaemon();
readonly grid = new GridReader(this.daemon);
debugOcrEngine: OcrEngine = 'tesseract';
debugPreprocess: OcrPreprocess = 'bgsub';
settings: OcrSettings = {
engine: 'easyocr',
screenPreprocess: 'none',
tooltipPreprocess: 'bgsub',
tooltipParams: {},
saveDebugImages: true,
};
// ── Screenshot capture ──────────────────────────────────────────────
@ -147,19 +160,23 @@ export class ScreenReader {
fuzzy: boolean = false,
): Promise<{ x: number; y: number } | null> {
const t = performance.now();
const result = await this.daemon.ocr();
const { engine, screenPreprocess } = this.settings;
const pp = screenPreprocess !== 'none' ? screenPreprocess : undefined;
const result = await this.daemon.ocr(undefined, engine, pp);
const pos = this.findWordInOcrResult(result, searchText, fuzzy);
if (pos) {
logger.info({ searchText, x: pos.x, y: pos.y, totalMs: elapsed(t) }, 'Found text on screen');
logger.info({ searchText, engine, x: pos.x, y: pos.y, totalMs: elapsed(t) }, 'Found text on screen');
} else {
logger.info({ searchText, totalMs: elapsed(t) }, 'Text not found on screen');
logger.info({ searchText, engine, totalMs: elapsed(t) }, 'Text not found on screen');
}
return pos;
}
async readFullScreen(): Promise<string> {
const result = await this.daemon.ocr();
const { engine, screenPreprocess } = this.settings;
const pp = screenPreprocess !== 'none' ? screenPreprocess : undefined;
const result = await this.daemon.ocr(undefined, engine, pp);
return result.text;
}
@ -170,7 +187,9 @@ export class ScreenReader {
searchText: string,
): Promise<{ x: number; y: number } | null> {
const t = performance.now();
const result = await this.daemon.ocr(region);
const { engine, screenPreprocess } = this.settings;
const pp = screenPreprocess !== 'none' ? screenPreprocess : undefined;
const result = await this.daemon.ocr(region, engine, pp);
const pos = this.findWordInOcrResult(result, searchText);
if (pos) {
@ -185,7 +204,9 @@ export class ScreenReader {
}
async readRegionText(region: Region): Promise<string> {
const result = await this.daemon.ocr(region);
const { engine, screenPreprocess } = this.settings;
const pp = screenPreprocess !== 'none' ? screenPreprocess : undefined;
const result = await this.daemon.ocr(region, engine, pp);
return result.text;
}
@ -201,7 +222,9 @@ export class ScreenReader {
}
async diffOcr(savePath?: string, region?: Region): Promise<DiffOcrResponse> {
return this.daemon.diffOcr(savePath, region);
const { engine, tooltipPreprocess, tooltipParams } = this.settings;
const pp = tooltipPreprocess !== 'none' ? tooltipPreprocess : undefined;
return this.daemon.diffOcr(savePath, region, engine, pp, tooltipParams);
}
// ── Template matching ──────────────────────────────────────────────
@ -238,43 +261,6 @@ export class ScreenReader {
logger.info({ path, region }, 'Region screenshot saved');
}
// ── Debug OCR (alternative engines) ─────────────────────────────────
async debugDiffOcr(savePath?: string, region?: Region): Promise<DiffOcrResponse> {
const t = performance.now();
const result = await this.daemon.diffOcr(savePath, region, this.debugOcrEngine, this.debugPreprocess);
logger.info({ engine: this.debugOcrEngine, preprocess: this.debugPreprocess, ms: elapsed(t) }, 'debugDiffOcr');
return result;
}
async debugOcr(region?: Region): Promise<OcrResponse> {
const t = performance.now();
const result = await this.daemon.ocr(region, this.debugOcrEngine, this.debugPreprocess);
logger.info({ engine: this.debugOcrEngine, preprocess: this.debugPreprocess, ms: elapsed(t) }, 'debugOcr');
return result;
}
async debugReadFullScreen(): Promise<string> {
const result = await this.daemon.ocr(undefined, this.debugOcrEngine, this.debugPreprocess);
return result.text;
}
async debugFindTextOnScreen(
searchText: string,
fuzzy: boolean = false,
): Promise<{ x: number; y: number } | null> {
const t = performance.now();
const result = await this.daemon.ocr(undefined, this.debugOcrEngine, this.debugPreprocess);
const pos = this.findWordInOcrResult(result, searchText, fuzzy);
if (pos) {
logger.info({ searchText, engine: this.debugOcrEngine, preprocess: this.debugPreprocess, x: pos.x, y: pos.y, totalMs: elapsed(t) }, 'debugFindText found');
} else {
logger.info({ searchText, engine: this.debugOcrEngine, preprocess: this.debugPreprocess, totalMs: elapsed(t) }, 'debugFindText not found');
}
return pos;
}
// ── Lifecycle ───────────────────────────────────────────────────────
async dispose(): Promise<void> {