aiex-cli 0.0.5-beta.5 → 0.0.6-beta.1

package/README.md

@@ -202,17 +202,6 @@ aiex completion fish | source
 
 <br>
 
-## 📄 Large Document Processing
-
-When processing very large documents (exceeding `40,000` characters), `aiex` runs an optimized **Pipeline Mode** to handle context window limits and control API costs:
-
-- **Token-Aware AST Splitting**: Parses structural Markdown elements (headings, paragraphs, lists) using an AST-based parser (`marked.lexer`) and splits them using precise token counters (`js-tiktoken`). Active heading hierarchies are tracked and prepended to each chunk as context. Tables and code blocks are kept intact (atomic blocks) to avoid syntax corruption.
-- **Concurrency Limiting**: To respect strict model rate limits, chunk extractions are processed in parallel with a strict concurrency limit (capped at 2 concurrent requests).
-- **Pre-filtering**: Integrates hybrid search-based pre-filtering to score and select only the most relevant document chunks based on schema queries, preventing unnecessary token usage on unrelated sections.
-- **Recursive Merging**: The final extracted JSON objects from each chunk are recursively merged, concatenating lists and deduplicating primitive fields.
-
-<br>
-
 ## 🔧 AI Configuration
 
 aiex works with any OpenAI-compatible API provider. Configure in the Web UI (AI Settings panel):

package/dist/cli.mjs

@@ -1,4 +1,4 @@
-import { A as doctorDiagnosticsTableRows, C as createConfig, D as package_default, E as name, O as version, S as AIConfigSchema, T as description, _ as DEFAULT_MINERU_API_CONFIG, a as parseJsonSchema, b as PLACEHOLDER_SCHEMA, c as recognizeImageText, d as t, f as getDefaultAIConfig, g as DEFAULT_MARKITDOWN_CONFIG, h as DEFAULT_MARKER_CONFIG, i as JsonSchemaDefinitionSchema, j as formatDoctorDiagnosticsJson, l as shouldUseImageOcrFallback, m as writeAIConfig, n as createMigrationConfig, o as toSnakeCase, p as readAIConfig, s as generateDrizzleSchema, t as collectDoctorDiagnostics, u as initI18n, v as DEFAULT_MINERU_CONFIG, w as seedConfig, x as PLACEHOLDER_TEXT, y as DEFAULT_PROMPT_CONFIG } from "./doctor-collector-NTNBFeBw.mjs";
+import { A as doctorDiagnosticsTableRows, C as createConfig, D as package_default, E as name, O as version, S as AIConfigSchema, T as description, _ as DEFAULT_MINERU_API_CONFIG, a as parseJsonSchema, b as PLACEHOLDER_SCHEMA, c as recognizeImageText, d as t, f as getDefaultAIConfig, g as DEFAULT_MARKITDOWN_CONFIG, h as DEFAULT_MARKER_CONFIG, i as JsonSchemaDefinitionSchema, j as formatDoctorDiagnosticsJson, l as shouldUseImageOcrFallback, m as writeAIConfig, n as createMigrationConfig, o as toSnakeCase, p as readAIConfig, s as generateDrizzleSchema, t as collectDoctorDiagnostics, u as initI18n, v as DEFAULT_MINERU_CONFIG, w as seedConfig, x as PLACEHOLDER_TEXT, y as DEFAULT_PROMPT_CONFIG } from "./doctor-collector-hWEvJ4lw.mjs";
 import { createRequire } from "node:module";
 import fs from "node:fs/promises";
 import os from "node:os";
@@ -17,7 +17,6 @@ import Database from "better-sqlite3";
 import pc from "picocolors";
 import { Buffer } from "node:buffer";
 import * as XLSX from "xlsx";
-import { getEncoding } from "js-tiktoken";
 import { createOpenAICompatible } from "@ai-sdk/openai-compatible";
 import { APICallError, Output, generateText, jsonSchema } from "ai";
 import pRetry from "p-retry";
@@ -25,7 +24,6 @@ import mime from "mime";
 import { jsonrepair } from "jsonrepair";
 import { LangfuseSpanProcessor } from "@langfuse/otel";
 import { NodeTracerProvider } from "@opentelemetry/sdk-trace-node";
-import { marked } from "marked";
 import crypto from "node:crypto";
 import { Client, extractNotionId } from "@notionhq/client";
 import { execa } from "execa";
@@ -13130,7 +13128,7 @@ function propertyToExtractionSchema(property) {
 	}
 	return { type: nullableType(property.type) };
 }
-function isRecord$1(value) {
+function isRecord(value) {
 	return typeof value === "object" && value !== null && !Array.isArray(value);
 }
 function schemaToExtractionOutputSchema(schema) {
@@ -13168,7 +13166,7 @@ function validatePropertyValue(path$1, property, value, issues) {
 			}
 			return;
 		case "object":
-			if (!isRecord$1(value)) {
+			if (!isRecord(value)) {
 				issues.push(`${path$1}: expected object or null`);
 				return;
 			}
@@ -13191,7 +13189,7 @@ function validateProperties(basePath, properties, data, issues) {
 	}
 }
 function validateExtractedData(schema, data) {
-	if (!isRecord$1(data)) return {
+	if (!isRecord(data)) return {
 		success: false,
 		error: "Extracted data must be a JSON object."
 	};
@@ -13514,220 +13512,6 @@ function insertExtractedData(db, schema, data) {
 	}
 }
 
-//#endregion
-//#region src/core/ai-extraction/json-merger.ts
-function isRecord(value) {
-	return typeof value === "object" && value !== null && !Array.isArray(value);
-}
-function mergePropertyValue(property, values) {
-	const nonNullValues = values.filter((v) => v !== null && v !== void 0);
-	if (nonNullValues.length === 0) return null;
-	if (property.type === "array") {
-		const concatenated = [];
-		for (const val of nonNullValues) if (Array.isArray(val)) concatenated.push(...val);
-		return concatenated;
-	}
-	if (property.type === "object") {
-		const childProperties = property.properties;
-		if (!childProperties) {
-			const mergedObj$1 = {};
-			for (const val of nonNullValues) if (isRecord(val)) Object.assign(mergedObj$1, val);
-			return mergedObj$1;
-		}
-		const mergedObj = {};
-		for (const [propName, propDef] of Object.entries(childProperties)) mergedObj[propName] = mergePropertyValue(propDef, nonNullValues.map((v) => isRecord(v) ? v[propName] : void 0));
-		return mergedObj;
-	}
-	const bestValue = nonNullValues.find((v) => {
-		if (typeof v === "string") return v.trim() !== "";
-		return true;
-	});
-	return bestValue !== void 0 ? bestValue : null;
-}
-/**
-* Merges structured extraction outputs from multiple document chunks
-* according to the schema properties.
-*/
-function mergeExtractionResults(schema, results) {
-	if (results.length === 0) return {};
-	if (results.length === 1) return results[0];
-	const merged = {};
-	for (const [propName, propDef] of Object.entries(schema.properties)) {
-		if (propDef.primary && propDef.autoIncrement) continue;
-		merged[propName] = mergePropertyValue(propDef, results.map((r) => r[propName]));
-	}
-	return merged;
-}
-
-//#endregion
-//#region src/core/ai-extraction/text-splitter.ts
-const encoding$1 = getEncoding("cl100k_base");
-function countTokens(text$1) {
-	return encoding$1.encode(text$1).length;
-}
-function formatHeadingContext(headings) {
-	const active = headings.filter(Boolean);
-	if (active.length === 0) return "";
-	return `> **[Context]** Belong to: ${active.join(" > ")}\n\n`;
-}
-function getMetadata(headings) {
-	return {
-		h1: headings[0] || void 0,
-		h2: headings[1] || void 0,
-		h3: headings[2] || void 0,
-		h4: headings[3] || void 0
-	};
-}
-/**
-* Splits text recursively using a list of separators.
-* Preserves the separators when re-joining.
-*/
-function splitTextRecursively(text$1, maxTokens, separators = [
-	"\n\n",
-	"\n",
-	"。",
-	". ",
-	" "
-]) {
-	if (countTokens(text$1) <= maxTokens) return [text$1];
-	if (separators.length === 0) {
-		const chunks = [];
-		let current = "";
-		for (const char of text$1) if (countTokens(current + char) > maxTokens) {
-			chunks.push(current);
-			current = char;
-		} else current += char;
-		if (current) chunks.push(current);
-		return chunks;
-	}
-	const separator = separators[0];
-	const nextSeparators = separators.slice(1);
-	const parts = text$1.split(separator);
-	const result = [];
-	let currentChunk = [];
-	let currentChunkTokens = 0;
-	for (let i = 0; i < parts.length; i++) {
-		const part = parts[i];
-		const itemText = part + (i < parts.length - 1 ? separator : "");
-		const partTokens = countTokens(itemText);
-		if (partTokens > maxTokens) {
-			if (currentChunk.length > 0) {
-				result.push(currentChunk.join(""));
-				currentChunk = [];
-				currentChunkTokens = 0;
-			}
-			const subParts = splitTextRecursively(part, maxTokens, nextSeparators);
-			for (let j = 0; j < subParts.length; j++) {
-				const finalSub = subParts[j] + (j === subParts.length - 1 && i < parts.length - 1 ? separator : "");
-				result.push(finalSub);
-			}
-		} else if (currentChunkTokens + partTokens > maxTokens) {
-			result.push(currentChunk.join(""));
-			currentChunk = [itemText];
-			currentChunkTokens = partTokens;
-		} else {
-			currentChunk.push(itemText);
-			currentChunkTokens += partTokens;
-		}
-	}
-	if (currentChunk.length > 0) result.push(currentChunk.join(""));
-	return result;
-}
-/**
-* Splits a Markdown document into chunks based on heading contexts, AST block parsing, and token limits.
-* Protects tables, list items, and code blocks from being broken.
-*/
-function splitMarkdown(text$1, maxTokens = 8e3, overlapTokens = 1e3) {
-	const tokens = marked.lexer(text$1);
-	const chunks = [];
-	let currentHeadings = [];
-	let currentChunkList = [];
-	let accumulatedTokens = 0;
-	const flushCurrentChunk = (isHeadingChange = false) => {
-		if (currentChunkList.length === 0) return;
-		const pageContent = currentChunkList.map((item) => item.text).join("");
-		const firstHeadings = currentChunkList[0].headings;
-		chunks.push({
-			pageContent,
-			metadata: getMetadata(firstHeadings)
-		});
-		if (isHeadingChange || overlapTokens <= 0) {
-			currentChunkList = [];
-			accumulatedTokens = 0;
-		} else {
-			const overlapItems = [];
-			let currentOverlapTokens = 0;
-			for (let i = currentChunkList.length - 1; i >= 0; i--) {
-				const item = currentChunkList[i];
-				const itemTokens = countTokens(item.text);
-				if (currentOverlapTokens + itemTokens > overlapTokens && overlapItems.length > 0) break;
-				overlapItems.unshift(item);
-				currentOverlapTokens += itemTokens;
-			}
-			currentChunkList = [...overlapItems];
-			accumulatedTokens = currentOverlapTokens;
-		}
-	};
-	for (const token of tokens) {
-		if (token.type === "space") {
-			if (currentChunkList.length > 0) {
-				currentChunkList[currentChunkList.length - 1].text += token.raw;
-				accumulatedTokens += countTokens(token.raw);
-			}
-			continue;
-		}
-		if (token.type === "heading") {
-			flushCurrentChunk(true);
-			const depth = token.depth;
-			const title = token.text.trim();
-			currentHeadings = currentHeadings.slice(0, depth - 1);
-			currentHeadings[depth - 1] = title;
-		}
-		const rawText = token.raw;
-		if (token.type === "list" && countTokens(rawText) > maxTokens) for (const item of token.items) processTextBlock(item.raw, currentHeadings);
-		else {
-			const isAtomic = token.type === "table" || token.type === "code";
-			processTextBlock(rawText, currentHeadings, isAtomic);
-		}
-	}
-	flushCurrentChunk(true);
-	return chunks;
-	function processTextBlock(blockText, headings, isAtomic = false) {
-		const blockTokens = countTokens(blockText);
-		const contextTokens = countTokens(formatHeadingContext(headings));
-		const safetyBuffer = Math.min(100, Math.max(2, Math.floor(maxTokens * .1)));
-		const budgetLimit = Math.max(5, maxTokens - contextTokens - safetyBuffer);
-		if (blockTokens > budgetLimit) if (isAtomic) {
-			flushCurrentChunk(false);
-			currentChunkList.push({
-				text: blockText,
-				headings: [...headings]
-			});
-			accumulatedTokens = blockTokens;
-			flushCurrentChunk(false);
-		} else {
-			flushCurrentChunk(false);
-			const subBlocks = splitTextRecursively(blockText, budgetLimit);
-			for (const sub of subBlocks) {
-				currentChunkList.push({
-					text: sub,
-					headings: [...headings]
-				});
-				accumulatedTokens += countTokens(sub);
-				if (accumulatedTokens > budgetLimit) flushCurrentChunk(false);
-			}
-		}
-		else {
-			if (accumulatedTokens + blockTokens + contextTokens > maxTokens && currentChunkList.length > 0) flushCurrentChunk(false);
-			currentChunkList.push({
-				text: blockText,
-				headings: [...headings]
-			});
-			accumulatedTokens += blockTokens;
-		}
-	}
-}
-
 //#endregion
 //#region src/core/extraction-audit.ts
 const AUDIT_ID_RE = /^[\w.-]+$/;
@@ -14669,44 +14453,7 @@ async function runBatchExtraction(aiexDir, config, aiConfig, schemaName, dir, gl
 
 //#endregion
 //#region src/core/extract-runner.ts
-const encoding = getEncoding("cl100k_base");
 const JSON_EXT_RE$1 = /\.json$/;
-async function limitConcurrency(concurrency, items, fn) {
-	const results = Array.from({ length: items.length });
-	let nextIndex = 0;
-	async function worker() {
-		while (nextIndex < items.length) {
-			const currentIndex = nextIndex++;
-			results[currentIndex] = await fn(items[currentIndex], currentIndex);
-		}
-	}
-	const workers = Array.from({ length: Math.min(concurrency, items.length) }, worker);
-	await Promise.all(workers);
-	return results;
-}
-function getSchemaKeywords(schema) {
-	const keywords = /* @__PURE__ */ new Set();
-	function walk(properties) {
-		if (!properties) return;
-		for (const [name$1, prop] of Object.entries(properties)) {
-			keywords.add(name$1.toLowerCase());
-			const parts = name$1.replace(/([a-z0-9])([A-Z])/g, "$1 $2").split(/[\s._:/\\-]+/g);
-			for (const part of parts) if (part.length > 1) keywords.add(part.toLowerCase());
-			if (prop && typeof prop === "object") {
-				const p = prop;
-				if (typeof p.title === "string") keywords.add(p.title.toLowerCase());
-				if (typeof p.description === "string") {
-					const descParts = p.description.toLowerCase().match(/[\p{L}\p{N}_-]+/gu) ?? [];
-					for (const d of descParts) if (d.length > 2) keywords.add(d);
-				}
-				if (p.type === "object") walk(p.properties);
-				if (p.type === "array" && p.items?.type === "object") walk(p.items.properties);
-			}
-		}
-	}
-	walk(schema.properties);
-	return Array.from(keywords);
-}
 async function ensureDatabaseReady(dbPath, schema) {
 	try {
 		await fs.access(dbPath);
@@ -14778,153 +14525,7 @@ async function extractSingle(aiexDir, config, aiConfig, schemaName, text$1, file
 	}
 	const s = spinner();
 	if (!options?.quiet) s.start(filePath ? t("command.extract.file.extractedFrom", { file: path.basename(filePath) }) : t("command.extract.file.extracting"));
-	const maxTokens = aiConfig.extraction?.maxTokens ?? 8e3;
-	const overlapTokens = aiConfig.extraction?.overlapSize ?? 1e3;
-	let result;
-	const totalTokens = text$1 ? encoding.encode(text$1).length : 0;
-	if (text$1 && totalTokens > maxTokens) {
-		if (!options?.quiet) consola.info(t("command.extract.file.chunking", {
-			length: totalTokens,
-			limit: maxTokens
-		}));
-		const finalDocs = splitMarkdown(text$1, maxTokens, overlapTokens);
-		if (!options?.quiet) consola.info(t("command.extract.file.chunksCount", { count: finalDocs.length }));
-		let processedDocs = finalDocs;
-		if (!!aiConfig.extraction?.preFiltering && finalDocs.length > 1) {
-			const preFilteringLimit = aiConfig.extraction?.preFilteringLimit ?? 5;
-			const keywords = getSchemaKeywords(schemaLoad.schema);
-			const scoredChunks = finalDocs.map((doc, idx) => {
-				if (idx === 0) return {
-					index: idx,
-					score: Number.POSITIVE_INFINITY
-				};
-				let score = 0;
-				const docTextLower = doc.pageContent.toLowerCase();
-				for (const kw of keywords) {
-					let pos = docTextLower.indexOf(kw);
-					while (pos !== -1) {
-						score++;
-						pos = docTextLower.indexOf(kw, pos + kw.length);
-					}
-				}
-				return {
-					index: idx,
-					score
-				};
-			}).slice(1).sort((a, b) => b.score - a.score);
-			const selectedIndices = new Set([0]);
-			let keptCount = 0;
-			for (const sc of scoredChunks) if (sc.score > 0 && keptCount < preFilteringLimit) {
-				selectedIndices.add(sc.index);
-				keptCount++;
-			}
-			processedDocs = finalDocs.filter((_, idx) => selectedIndices.has(idx));
-			if (!options?.quiet) consola.info(t("command.extract.file.preFiltering", {
-				original: finalDocs.length,
-				filtered: processedDocs.length
-			}));
-		}
-		const chunkResults = [];
-		const accumulatedTokens = {
-			prompt: 0,
-			completion: 0,
-			total: 0
-		};
-		let success = true;
-		let errorMsg = "";
-		const extractionTasks = processedDocs.map((doc, i) => {
-			return async () => {
-				if (!success) return;
-				const headings = [];
-				if (doc.metadata) {
-					if (doc.metadata.h1) headings.push(doc.metadata.h1);
-					if (doc.metadata.h2) headings.push(doc.metadata.h2);
-					if (doc.metadata.h3) headings.push(doc.metadata.h3);
-					if (doc.metadata.h4) headings.push(doc.metadata.h4);
-				}
-				let chunkText = doc.pageContent;
-				if (headings.length > 0) chunkText = `> **[Context]** Belong to: ${headings.join(" > ")}\n\n${chunkText}`;
-				const chunkResult = await extractStructuredData({
-					config: aiConfig,
-					schema: schemaLoad.schema,
-					text: chunkText,
-					aiexDir,
-					modelOverride,
-					onRetry(info) {
-						if (!options?.quiet) s.message(t("command.extract.file.extractRetryChunk", {
-							current: i + 1,
-							total: processedDocs.length,
-							code: info.statusCode,
-							delay: info.delayMs / 1e3,
-							attempt: info.attempt,
-							max: info.maxRetries
-						}));
-					}
-				});
-				if (!chunkResult.success) {
-					success = false;
-					errorMsg = chunkResult.error || t("common.unknownError");
-					if (!options?.quiet) {
-						s.stop(t("command.extract.file.extractFailChunk", { current: i + 1 }));
-						consola.error(errorMsg);
-					}
-					return;
-				}
-				if (chunkResult.data) chunkResults.push(chunkResult.data);
-				if (chunkResult.tokensUsed) {
-					accumulatedTokens.prompt += chunkResult.tokensUsed.prompt ?? 0;
-					accumulatedTokens.completion += chunkResult.tokensUsed.completion ?? 0;
-					accumulatedTokens.total += chunkResult.tokensUsed.total ?? 0;
-				}
-			};
-		});
-		const concurrency = Math.min(aiConfig.extraction?.concurrency ?? 2, 2);
-		if (!options?.quiet && processedDocs.length > 0) s.message(t("command.extract.file.extractingChunk", {
-			current: 1,
-			total: processedDocs.length
-		}));
-		try {
-			await limitConcurrency(concurrency, extractionTasks, async (task, idx) => {
-				if (!options?.quiet && success) s.message(t("command.extract.file.extractingChunk", {
-					current: idx + 1,
-					total: processedDocs.length
-				}));
-				await task();
-			});
-		} catch (e) {
-			success = false;
-			errorMsg = e instanceof Error ? e.message : String(e);
-		}
-		if (!success) return {
-			success: false,
-			error: errorMsg
-		};
-		const mergedData = mergeExtractionResults(schemaLoad.schema, chunkResults);
-		const validation = validateExtractedData(schemaLoad.schema, mergedData);
-		if (!validation.success) {
-			const valError = validation.error || "Merged data validation failed";
-			if (!options?.quiet) {
-				s.stop(t("command.extract.file.validationFail"));
-				consola.error(valError);
-			}
-			return {
-				success: false,
-				error: valError
-			};
-		}
-		const outputDir = path.resolve(aiexDir, aiConfig.extraction?.outputDir?.replace(".aiex/", "") ?? "extracted");
-		await fs.mkdir(outputDir, { recursive: true });
-		const timestamp = (/* @__PURE__ */ new Date()).toISOString().replace(/[:.]/g, "-");
-		const outputFileName = `${schemaLoad.schema.table.name}-${timestamp}.json`;
-		const finalMergedOutputPath = path.join(outputDir, outputFileName);
-		await fs.writeFile(finalMergedOutputPath, JSON.stringify(mergedData, null, 2));
-		result = {
-			success: true,
-			data: mergedData,
-			tokensUsed: accumulatedTokens,
-			outputPath: finalMergedOutputPath
-		};
-	} else result = await extractStructuredData({
+	const result = await extractStructuredData({
 		config: aiConfig,
 		schema: schemaLoad.schema,
 		text: text$1 ?? "",
@@ -14952,11 +14553,6 @@ async function extractSingle(aiexDir, config, aiConfig, schemaName, text$1, file
 	}
 	if (!options?.quiet) s.stop(t("command.extract.file.extractComplete"));
 	if (result.outputPath && !options?.quiet) consola.success(t("command.extract.file.resultSaved", { path: pc.cyan(result.outputPath) }));
-	if (result.evidenceSummary && !options?.quiet) {
-		const summary = result.evidenceSummary;
-		const issueText = summary.issueCount > 0 ? pc.yellow(String(summary.issueCount)) : pc.green("0");
-		consola.info(pc.gray(`Evidence coverage: ${summary.evidenceCount}/${summary.fieldCount} fields, found ${summary.foundCount}, inferred ${summary.inferredCount}, missing ${summary.missingCount}, issues ${issueText}`));
-	}
 	if (result.tokensUsed && !options?.quiet) consola.info(pc.gray(t("command.extract.file.tokenUsage", {
 		prompt: result.tokensUsed.prompt,
 		completion: result.tokensUsed.completion,
@@ -14985,7 +14581,6 @@ async function extractSingle(aiexDir, config, aiConfig, schemaName, text$1, file
 						outputPath: result.outputPath,
 						data: result.data,
 						tablesInserted: insertResult.tablesInserted,
-						evidenceSummary: result.evidenceSummary,
 						tokensUsed: result.tokensUsed
 					};
 				} else {
@@ -15012,7 +14607,6 @@ async function extractSingle(aiexDir, config, aiConfig, schemaName, text$1, file
 		success: true,
 		outputPath: result.outputPath,
 		data: result.data,
-		evidenceSummary: result.evidenceSummary,
 		tokensUsed: result.tokensUsed
 	};
 }
@@ -15125,7 +14719,6 @@ async function runAuditedExtraction(options) {
 				outputName: updated.outputName,
 				tablesInserted: updated.tablesInserted,
 				notionPages: updated.notionPages,
-				evidenceSummary: r.evidenceSummary,
 				tokensUsed: updated.tokensUsed,
 				auditId: updated.id,
 				fileHash
@@ -16253,7 +15846,6 @@ function aiRoutes(config) {
 //#endregion
 //#region src/core/data-service.ts
 const FILE_REGEX = /\.json$/;
-const EVIDENCE_FILE_SUFFIX = ".evidence.json";
 const EXTRACTION_TIMESTAMP_RE = /-\d{4}-\d{2}-\d{2}T/;
 const INTERNAL_ROWID_COLUMN = "__aiex_rowid";
 const TIMESTAMP_CLEANUP = /(\d{2})-(\d{2})-(\d{2})/;
@@ -16269,24 +15861,6 @@ function getAuditNotionStatus(record) {
 	if (record.status === "failed") return "failed";
 	return "not_synced";
 }
-async function readEvidenceSummary(extractedDir, outputName) {
-	const evidencePath = path.join(extractedDir, outputName.replace(FILE_REGEX, EVIDENCE_FILE_SUFFIX));
-	try {
-		const coverage = (await readFile(evidencePath))?.coverage;
-		if (!coverage || typeof coverage !== "object") return void 0;
-		return {
-			path: evidencePath,
-			fieldCount: Number(coverage.fieldCount) || 0,
-			evidenceCount: Number(coverage.evidenceCount) || 0,
-			foundCount: Number(coverage.foundCount) || 0,
-			missingCount: Number(coverage.missingCount) || 0,
-			inferredCount: Number(coverage.inferredCount) || 0,
-			issueCount: Number(coverage.issueCount) || 0
-		};
-	} catch {
-		return;
-	}
-}
 async function getRowExtractionActions(aiexDir, tableName) {
 	const actions = /* @__PURE__ */ new Map();
 	const auditRecords = await listExtractionAuditRecords(aiexDir);
@@ -16314,7 +15888,7 @@ async function listExtractions(config) {
 	const aiexDir = path.dirname(config.schemaPath);
 	const extractedDir = path.join(aiexDir, "extracted");
 	await fs.mkdir(extractedDir, { recursive: true });
-	const jsonFiles = (await fs.readdir(extractedDir)).filter((f) => f.endsWith(".json") && !f.endsWith(".prompt.md") && !f.endsWith(EVIDENCE_FILE_SUFFIX));
+	const jsonFiles = (await fs.readdir(extractedDir)).filter((f) => f.endsWith(".json") && !f.endsWith(".prompt.md"));
 	const auditRecords = await listExtractionAuditRecords(aiexDir);
 	const auditByOutputName = new Map(auditRecords.map((record) => [record.outputName, record]));
 	const records = [];
@@ -16333,7 +15907,6 @@ async function listExtractions(config) {
 				timestamp,
 				fileSize: stat.size,
 				modifiedAt: stat.mtime.toISOString(),
-				evidenceSummary: await readEvidenceSummary(extractedDir, file),
 				notionStatus: notionPages ? "synced" : audit?.status === "failed" ? "failed" : "not_synced",
 				notionPages,
 				notionError: !notionPages && audit?.status === "failed" ? audit.error : void 0
@@ -16513,7 +16086,6 @@ async function retryNotionSync(config, fileName) {
 
 //#endregion
 //#region src/server/routes/data.ts
-const JSON_FILE_SUFFIX_RE = /\.json$/;
 const tableParamSchema = z.object({ name: z.string().regex(/^[a-z][a-z0-9_]*$/) });
 const extractionFileParamSchema = z.object({ name: z.string().regex(/^[\w.-]+\.json$/).refine((name$1) => name$1 === path.basename(name$1) && !name$1.includes("..")) });
 const tableQuerySchema = z.object({
@@ -16566,22 +16138,10 @@ function dataRoutes(config) {
 		const filePath = path.join(extractedDir, name$1);
 		try {
 			const content = await fs.readFile(filePath, "utf-8");
-			const evidencePath = path.join(extractedDir, name$1.replace(JSON_FILE_SUFFIX_RE, ".evidence.json"));
-			let evidenceSummary;
-			try {
-				const evidence = JSON.parse(await fs.readFile(evidencePath, "utf-8"));
-				evidenceSummary = evidence?.coverage ? {
-					...evidence.coverage,
-					path: evidencePath
-				} : void 0;
-			} catch {
-				evidenceSummary = void 0;
-			}
 			return c.json({
 				success: true,
 				content,
-				name: name$1,
-				evidenceSummary
+				name: name$1
 			});
 		} catch {
 			return c.json({ error: t("server.extractionNotFound") }, 404);
@@ -16725,7 +16285,6 @@ function extractRoutes(config) {
 				outputName: result.outputName,
 				tablesInserted: result.tablesInserted,
 				notionPages: result.notionPages,
-				evidenceSummary: result.evidenceSummary,
 				tokensUsed: result.tokensUsed,
 				auditId: result.auditId
 			}, 200);
@@ -16793,7 +16352,6 @@ function extractRoutes(config) {
 			outputName: result.outputName,
 			tablesInserted: result.tablesInserted,
 			notionPages: result.notionPages,
-			evidenceSummary: result.evidenceSummary,
 			tokensUsed: result.tokensUsed,
 			auditId: result.auditId
 		}, 200);

package/dist/{doctor-collector-NTNBFeBw.mjs → doctor-collector-hWEvJ4lw.mjs}

@@ -74,7 +74,7 @@ function doctorDiagnosticsTableRows(d) {
 //#endregion
 //#region package.json
 var name = "aiex-cli";
-var version = "0.0.5-beta.5";
+var version = "0.0.6-beta.1";
 var description = "JSON Schema → SQLite with AI-powered data extraction";
 var package_default = {
 	name,
@@ -158,11 +158,9 @@ var package_default = {
 		"hono": "catalog:",
 		"i18next": "catalog:",
 		"i18next-fs-backend": "catalog:",
-		"js-tiktoken": "catalog:",
 		"jsonfile": "catalog:",
 		"jsonrepair": "catalog:",
 		"kysely": "catalog:",
-		"marked": "catalog:",
 		"mime": "catalog:",
 		"open": "catalog:",
 		"p-retry": "catalog:",
@@ -230,15 +228,7 @@ const PromptConfigSchema = z.object({
 	systemTemplate: z.string().min(1),
 	userTemplate: z.string().min(1)
 });
-const ExtractionConfigSchema = z.object({
-	outputDir: z.string().min(1),
-	mode: z.enum(["pipeline"]).default("pipeline").optional(),
-	concurrency: z.number().int().min(1).optional(),
-	maxTokens: z.number().int().positive().default(8e3).optional(),
-	overlapSize: z.number().int().nonnegative().optional(),
-	preFiltering: z.boolean().optional(),
-	preFilteringLimit: z.number().int().min(1).optional()
-});
+const ExtractionConfigSchema = z.object({ outputDir: z.string().min(1) });
 const ImageOcrConfigSchema = z.object({
 	ocrFallback: z.enum([
 		"auto",
@@ -345,10 +335,7 @@ Extraction requirements:
 	userTemplate: `Please extract data from the following text:
 {text}`
 };
-const DEFAULT_EXTRACTION_CONFIG = {
-	outputDir: ".aiex/extracted",
-	mode: "pipeline"
-};
+const DEFAULT_EXTRACTION_CONFIG = { outputDir: ".aiex/extracted" };
 const DEFAULT_IMAGE_OCR_CONFIG = {
 	ocrFallback: "auto",
 	ocrLanguages: "en-US, zh-Hans",
@@ -580,13 +567,6 @@ const en = {
 				extractFail: "Extraction failed",
 				extractComplete: "Extraction complete",
 				extractRetry: "API responded with {{code}}, retrying in {{delay}}s ({{attempt}}/{{max}})",
-				chunking: "Input text ({{length}} tokens) exceeds limit ({{limit}} tokens). Splitting into chunks...",
-				chunksCount: "Split into {{count}} chunk(s).",
-				preFiltering: "Hybrid pre-filtering: selected {{filtered}} out of {{original}} chunks based on schema relevance.",
-				extractingChunk: "Extracting chunk {{current}}/{{total}}...",
-				extractRetryChunk: "Chunk {{current}}/{{total}} API responded with {{code}}, retrying in {{delay}}s ({{attempt}}/{{max}})",
-				extractFailChunk: "Extraction failed for chunk {{current}}/{{total}}",
-				validationFail: "Merged data validation failed",
 				resultSaved: "Result saved: {{path}}",
 				tokenUsage: "Token usage: prompt={{prompt}}, completion={{completion}}, total={{total}}",
 				insertingDb: "Inserting into database...",
@@ -976,7 +956,7 @@ async function initI18n(lng) {
 			fallbackLng: "en",
 			resources: {
 				"en": { translation: en },
-				"zh-CN": { translation: await import("./zh-CN-Ca-Dv775.mjs").then((m) => m.zhCN) }
+				"zh-CN": { translation: await import("./zh-CN-Qcn0DHFh.mjs").then((m) => m.zhCN) }
 			},
 			interpolation: { escapeValue: false },
 			returnNull: false