$1");
- return out;
-}
-
-function markdownTableCells(row) {
- const cells = [""];
- const text = row.replace(/^\||\|$/g, "");
- for (let i = 0; i < text.length; i++) {
- if (text[i] === "\\" && i + 1 < text.length) {
- cells[cells.length - 1] += text[i] + text[++i];
- } else if (text[i] === "|") cells.push("");
- else cells[cells.length - 1] += text[i];
- }
- return cells.map(cell => cell.trim());
-}
-
-function markdownToHtml(md) {
- const lines = md.replace(/\r\n/g, "\n").split("\n");
- let html = "";
- let i = 0;
- let inList = null; // 'ul' | 'ol' | null
-
- function closeList() {
- if (inList) { html += ``; inList = null; }
- }
-
- while (i < lines.length) {
- const line = lines[i];
-
- if (/^\s*$/.test(line)) { closeList(); i++; continue; }
-
- const heading = line.match(/^(#{1,4})\s+(.*)$/);
- if (heading) {
- closeList();
- const level = heading[1].length;
- html += `${inlineMd(quoteLines.join(" "))}`; - continue; - } - - if (/^\s*\|.*\|\s*$/.test(line)) { - closeList(); - const rows = []; - while (i < lines.length && /^\s*\|.*\|\s*$/.test(lines[i])) { - rows.push(lines[i].trim()); - i++; - } - if (rows.length >= 2 && /^\|?\s*:?-{2,}:?\s*(\|\s*:?-{2,}:?\s*)+\|?$/.test(rows[1])) { - const headCells = markdownTableCells(rows[0]); - html += "
| ${inlineMd(c)} | `).join("") + "
|---|
| ${inlineMd(c)} | `).join("") + "
${inlineMd(r)}
`; }); - } - continue; - } - - const bullet = line.match(/^\s*[-*]\s+(.*)$/); - if (bullet) { - if (inList !== "ul") { closeList(); html += "${inlineMd(paraLines.join(" "))}
`; - } - closeList(); - return html; -} - // -------------------------------------------------------------- bootstrap -- // Guarded so this file can also be loaded under Node (no `document`) to unit // test the pure functions above (markdownToHtml, extractLastVisibleAiText, diff --git a/jobscout-web/core.js b/jobscout-web/core.js new file mode 100644 index 00000000..7698fb9d --- /dev/null +++ b/jobscout-web/core.js @@ -0,0 +1,565 @@ +// Pure display, evidence and stream helpers. No DOM or application state. +(function (root, factory) { + const value = factory(); + if (typeof module !== "undefined") module.exports = value; + else root.JobScoutCore = value; +})(globalThis, function () { + function resolveDeploymentConfig(config = {}, location = {}) { + const localSplit = + location.protocol === "http:" && + location.port === "5500" && + ["localhost", "127.0.0.1", "[::1]"].includes(location.hostname); + const localHost = location.hostname || "localhost"; + const gatewayBase = + config.gatewayBase ?? (localSplit ? `http://${localHost}:8001` : ""); + const capabilityCenterUrl = + config.capabilityCenterUrl ?? + (localSplit + ? `http://${localHost}:3000/workspace/capabilities` + : "/workspace/capabilities"); + function validateUrl(value, allowPath) { + if (typeof value !== "string") throw new Error("部署地址必须是字符串"); + if (allowPath && /^\/(?!\/)/.test(value) && !value.includes("\\")) + return value; + if (!value && !allowPath) return ""; + const parsed = new URL(value); + if ( + !["http:", "https:"].includes(parsed.protocol) || + parsed.username || + parsed.password || + (!allowPath && + (parsed.pathname !== "/" || parsed.search || parsed.hash)) + ) { + throw new Error( + "部署地址必须是有效的 HTTP(S) 地址，且不能包含账号密码", + ); + } + return allowPath ? parsed.href : parsed.origin; + } + return { + gatewayBase: validateUrl(gatewayBase, false), + capabilityCenterUrl: validateUrl(capabilityCenterUrl, true), + }; + } + + const TERMINAL_TRACKER_STATUSES = new Set(["Offer", "未通过", "流程终止"]); + const STOPPED_TRACKER_STATUSES = new Set(["未通过", "流程终止"]); + + function isTerminalTrackerStatus(status) { + return TERMINAL_TRACKER_STATUSES.has(status); + } + + function trackerDisplayRow(row) { + const source = row?.source_summary; + if (!source || source.conflict || source.source !== "email") return row; + return { + ...row, + portal_status: row.status, + portal_evidence: row.evidence, + status: source.status, + stage: row.stage_manual ? row.stage : source.status, + raw_status: source.status, + evidence: source.evidence, + checked_at: source.received_at, + changed_at: source.received_at, + check_result: "成功", + confidence: 1, + changed: false, + }; + } + + function trackerRowPresentation(row) { + const terminal = + Boolean(row?.terminal) || isTerminalTrackerStatus(row?.status); + const needsReview = + Boolean(row?.checked_at) && Number(row?.confidence) < 0.7; + const changed = Boolean(row?.changed && row?.previous_status); + const stopped = + STOPPED_TRACKER_STATUSES.has(row?.status) || + STOPPED_TRACKER_STATUSES.has(row?.stage); + const unknown = + needsReview || + !row?.status || + row.status === "未知" || + (row.check_result && row.check_result !== "成功"); + return { + terminal, + canRefresh: !terminal, + needsReview, + changeText: changed ? `${row.previous_status} → ${row.status}` : "", + tone: changed ? "changed" : needsReview ? "review" : "normal", + statusTone: stopped ? "stopped" : unknown ? "unknown" : "active", + }; + } + + function trackerSummary(rows) { + const summary = { total: rows.length, active: 0, review: 0, offers: 0 }; + for (const stored of rows) { + const row = trackerDisplayRow(stored); + const presentation = trackerRowPresentation(row); + const review = + row.source_summary?.conflict || + presentation.needsReview || + !row.status || + row.status === "未知" || + (row.check_result && row.check_result !== "成功"); + if (review) summary.review += 1; + else if (row.status === "Offer") summary.offers += 1; + else if (!presentation.terminal) summary.active += 1; + } + return summary; + } + + function trackerStageWaitText(row, now = new Date()) { + if (!row?.status || row.status === "未知") return "阶段待确认"; + if (row.terminal || isTerminalTrackerStatus(row.status)) + return "流程已结束"; + if (row.stage_manual) return "手动环节，起点待确认"; + if (!row.changed_at) return "等待首次检查"; + const started = new Date(row.changed_at); + const elapsed = now.getTime() - started.getTime(); + if (!Number.isFinite(elapsed) || elapsed < 0) return "起点时间待确认"; + const days = Math.floor(elapsed / 86400000); + return days < 1 + ? "不足 1 天（自首次识别）" + : `至少 ${days} 天（自首次识别）`; + } + + function trackerStageLabel(row) { + row = trackerDisplayRow(row); + return row?.stage || row?.status || "未分类"; + } + + function trackerStageFilterOptions(rows, stages = []) { + const counts = new Map(); + for (const row of rows) { + const label = trackerStageLabel(row); + counts.set(label, (counts.get(label) || 0) + 1); + } + const ordered = [...new Set(stages)].filter((label) => counts.has(label)); + for (const label of counts.keys()) { + if (!ordered.includes(label)) ordered.push(label); + } + return ordered.map((label) => ({ label, count: counts.get(label) })); + } + + function filterTrackerRows(rows, stage) { + return stage === null + ? rows + : rows.filter((row) => trackerStageLabel(row) === stage); + } + + function parseSseFrames(buffer) { + const normalized = String(buffer || "").replace(/\r\n/g, "\n"); + const frames = normalized.split("\n\n"); + const remainder = frames.pop() || ""; + const events = []; + for (const frame of frames) { + const data = frame + .split("\n") + .filter((line) => line.startsWith("data:")) + .map((line) => line.slice(5).trimStart()) + .join("\n"); + if (data) events.push(JSON.parse(data)); + } + return { events, remainder }; + } + + // ---------------------------------------------------------------- helpers -- + + function jobScoutRunContext(mode, contextRef = null) { + return { + jobscout_mode: mode === "match" ? "base_match" : "interview_prep", + jobscout_evidence_version: 2, + ...(mode === "match" && typeof contextRef === "string" && contextRef + ? { jobscout_base_context_ref: contextRef } + : {}), + }; + } + + /** Canonical display text for both values streams and persisted history. + * Backend-filtered content also feeds renderReport's print/download actions. + * Old reports cannot be retroactively verified without their original tools. */ + function guardedMessageText(message, mode = "prep") { + if (message?.additional_kwargs?.hide_from_ui) return ""; + if (message?.type === "tool" && message.name !== "ask_clarification") + return ""; + if (message?.tool_calls?.length) return ""; + const text = contentToText(message?.content).trim(); + if ( + !text || + (message?.type === "tool" && message.name === "ask_clarification") + ) + return text; + const stamp = message?.additional_kwargs?.jobscout_evidence; + const route = message?.additional_kwargs?.jobscout_route; + if ( + route?.version === 3 && + typeof route.run_id === "string" && + route.run_id && + typeof route.in_scope === "boolean" && + Array.isArray(route.missing_fields) && + (!route.in_scope || route.missing_fields.length > 0) + ) + return text; + if ( + stamp?.version === 2 && + stamp.mode === (mode === "match" ? "base_match" : "interview_prep") && + typeof stamp.run_id === "string" && + stamp.run_id && + Number.isInteger(stamp.accepted_count) && + stamp.accepted_count >= 0 && + Number.isInteger(stamp.removed_count) && + stamp.removed_count >= 0 + ) + return text; + return "这份内容尚无本次运行的结构化证据与来源校验记录，暂不展示或导出。旧报告需要重新生成；新报告请先启用 JobScout 来源校验中间件。"; + } + + function contentToText(content) { + if (typeof content === "string") return content; + if (Array.isArray(content)) { + return content + .map((part) => (typeof part === "string" ? part : part?.text || "")) + .filter(Boolean) + .join("\n"); + } + return ""; + } + + function extractLastVisibleAiText(messages, mode = "prep") { + // `ask_clarification` is `return_direct=True`: LangGraph ends the run with + // its ToolMessage (type "tool") as the final state, and the model's own + // preceding AIMessage carries the tool *call*, not the question text the + // user is meant to see. Only checking type "ai" here silently swallowed + // every clarification question — a real bug found via a live user report, + // not code review (Gateway logs proved the server side worked fine; the + // Gateway's actual response just wasn't the shape this function assumed). + for (let i = messages.length - 1; i >= 0; i--) { + const m = messages[i]; + if (m?.type === "human") break; + if (m?.additional_kwargs?.hide_from_ui) continue; + if (m?.type === "ai" || m?.type === "tool") { + const text = guardedMessageText(m, mode); + if (text) return text; + } + } + return ""; + } + + /** A real report has these as standalone section-title lines. The preferred + * shape is a Markdown heading, but live models may emit a numbered outline + * (`1) 公司速览`, `2) 岗位拆解（...）`) while still returning the complete + * document. A prose mention must NOT count, or the UI renders print/download + * buttons for an offer to create a report rather than the report itself. */ + function looksLikeReport(text) { + const sectionPrefix = "(?:#{1,3}\\s+|\\d+\\s*[)）.、]\\s*)?"; + const sectionSuffix = "(?:[（(][^\\r\\n]*[）)])?\\s*$"; + const hitCount = (markers) => + markers.filter((m) => + new RegExp(`^${sectionPrefix}${m}${sectionSuffix}`, "m").test(text), + ).length; + return ( + hitCount(["公司速览", "岗位拆解", "面试题预测"]) >= 2 || + hitCount(["候选人画像", "推荐岗位", "匹配依据", "风险与数据边界"]) >= 3 + ); + } + + /** Keep the JobScout document contract stable when a model appends generic + * coaching chapters of its own. This is deliberately conservative: only + * known top-level drift is removed, so numbered questions and preparation + * steps inside an allowed section are never mistaken for new chapters. */ + function sanitizeJobScoutReportMarkdown(markdown) { + const source = String(markdown || ""); + const matchingMode = + source.includes("简历 × 飞书岗位匹配报告") || + ["候选人画像", "推荐岗位", "匹配依据", "风险与数据边界"].filter((s) => + source.includes(s), + ).length >= 3; + const allowedSections = matchingMode + ? ["候选人画像", "推荐岗位", "匹配依据", "风险与数据边界"] + : [ + "公司速览", + "岗位拆解", + "面试题预测", + "差距分析", + "证据边界与后续建议", + ]; + const unwantedSections = [ + "使用说明", + "一周上岸计划", + "上岸计划", + "冲刺计划", + "话术模板", + "专项准备资料", + "清单与打卡", + "执行凭据", + "工具收据", + "sources", + ]; + const startsWithAny = (value, candidates) => { + const normalized = value.trim().toLowerCase(); + return candidates.some((candidate) => + normalized.startsWith(candidate.toLowerCase()), + ); + }; + + let keep = true; + const keptLines = []; + for (const line of source.split(/\r?\n/)) { + const trimmed = line.trim(); + const numbered = trimmed.match(/^\d+\s*[)）.、]\s*(.+)$/); + const levelTwo = trimmed.match(/^##(?!#)\s+(.+)$/); + const bareStop = /^(?:执行凭据(?:（工具收据）)?|工具收据|sources)$/i.test( + trimmed, + ); + const candidate = + numbered?.[1] || levelTwo?.[1] || (bareStop ? trimmed : ""); + + if (candidate) { + if (startsWithAny(candidate, allowedSections)) { + keep = true; + } else if ( + levelTwo || + bareStop || + startsWithAny(candidate, unwantedSections) + ) { + keep = false; + } + } + + if (keep) keptLines.push(line); + } + + return keptLines.join("\n").trimEnd() + "\n"; + } + + /** Normalize common model formatting drift for the rendered/printed view. + * The sanitized Markdown is also used for download. */ + function normalizeReportMarkdownForRender(markdown) { + let seenNonEmpty = false; + return markdown + .split(/\r?\n/) + .map((line) => { + const trimmed = line.trim(); + if (!trimmed) return line; + + if (!seenNonEmpty) { + seenNonEmpty = true; + if ( + !trimmed.startsWith("#") && + (trimmed.includes("面试准备包") || trimmed.includes("岗位匹配报告")) + ) { + return `# ${trimmed}`; + } + } + + const numberedSection = trimmed.match( + /^\d+\s*[)）.、]\s*(公司速览|岗位拆解|面试题预测|差距分析|候选人画像|推荐岗位|匹配依据|风险与数据边界)(.*)$/, + ); + if (numberedSection) { + return `## ${numberedSection[1]}${numberedSection[2]}`; + } + + if ( + !trimmed.startsWith("#") && + /^(?:业务与产品|近期动态|融资\s*[\/／]\s*规模(?:（.*）)?|公司技术栈核对(?:（.*）)?|技术\s*[\/／]\s*岗位题|行为题|证据清单(?:（.*）)?)$/.test( + trimmed, + ) + ) { + return `### ${trimmed}`; + } + + return line; + }) + .join("\n"); + } + + function escapeHtml(s) { + return s.replace(/&/g, "&").replace(//g, ">"); + } + + function inlineMd(s) { + // Code-rendered evidence escapes Markdown punctuation. Preserve those + // characters as text before interpreting actual formatting and links. + let out = escapeHtml(s) + .replace( + /\\([\\`*_{}\[\]()#+.!|])/g, + (_, char) => "&#" + char.charCodeAt(0) + ";", + ) + .replace(/\\(<|>)/g, "$1"); + out = out.replace( + /\[([^\]]+)\]\(<(https?:\/\/[^\s]+?)>\)/g, + '$1', + ); + out = out.replace( + /\[([^\]]+)\]\((https?:\/\/[^\s)]+)\)/g, + '$1', + ); + out = out.replace(/\*\*([^*]+)\*\*/g, "$1"); + out = out.replace(/`([^`]+)`/g, "$1");
+ return out;
+ }
+
+ function markdownTableCells(row) {
+ const cells = [""];
+ const text = row.replace(/^\||\|$/g, "");
+ for (let i = 0; i < text.length; i++) {
+ if (text[i] === "\\" && i + 1 < text.length) {
+ cells[cells.length - 1] += text[i] + text[++i];
+ } else if (text[i] === "|") cells.push("");
+ else cells[cells.length - 1] += text[i];
+ }
+ return cells.map((cell) => cell.trim());
+ }
+
+ function markdownToHtml(md) {
+ const lines = md.replace(/\r\n/g, "\n").split("\n");
+ let html = "";
+ let i = 0;
+ let inList = null; // 'ul' | 'ol' | null
+
+ function closeList() {
+ if (inList) {
+ html += ``;
+ inList = null;
+ }
+ }
+
+ while (i < lines.length) {
+ const line = lines[i];
+
+ if (/^\s*$/.test(line)) {
+ closeList();
+ i++;
+ continue;
+ }
+
+ const heading = line.match(/^(#{1,4})\s+(.*)$/);
+ if (heading) {
+ closeList();
+ const level = heading[1].length;
+ html += `${inlineMd(quoteLines.join(" "))}`; + continue; + } + + if (/^\s*\|.*\|\s*$/.test(line)) { + closeList(); + const rows = []; + while (i < lines.length && /^\s*\|.*\|\s*$/.test(lines[i])) { + rows.push(lines[i].trim()); + i++; + } + if ( + rows.length >= 2 && + /^\|?\s*:?-{2,}:?\s*(\|\s*:?-{2,}:?\s*)+\|?$/.test(rows[1]) + ) { + const headCells = markdownTableCells(rows[0]); + html += + "
| ${inlineMd(c)} | `).join("") + + "
|---|
| ${inlineMd(c)} | `).join("") + + "
${inlineMd(r)}
`; + }); + } + continue; + } + + const bullet = line.match(/^\s*[-*]\s+(.*)$/); + if (bullet) { + if (inList !== "ul") { + closeList(); + html += "${inlineMd(paraLines.join(" "))}
`; + } + closeList(); + return html; + } + + return { + resolveDeploymentConfig, + isTerminalTrackerStatus, + trackerDisplayRow, + trackerRowPresentation, + trackerSummary, + trackerStageWaitText, + trackerStageLabel, + trackerStageFilterOptions, + filterTrackerRows, + parseSseFrames, + jobScoutRunContext, + guardedMessageText, + contentToText, + extractLastVisibleAiText, + looksLikeReport, + sanitizeJobScoutReportMarkdown, + normalizeReportMarkdownForRender, + escapeHtml, + inlineMd, + markdownTableCells, + markdownToHtml, + }; +}); diff --git a/jobscout-web/index.html b/jobscout-web/index.html index 0eb91752..5456d090 100644 --- a/jobscout-web/index.html +++ b/jobscout-web/index.html @@ -367,6 +367,11 @@