kevlar-4u 1.0.0

package/dist/tools/createPersonaWizardTool.js

@@ -0,0 +1,713 @@
+import * as path from "path";
+import * as fs from "fs";
+import { handleCreatePersona, generateIdFromDraft, getSubDirFromDraft, applyDedup, } from "./createPersonaTool.js";
+import { logger, getErrorInfo } from "../utils/observability.js";
+const AGE_RANGE_OPTIONS = [
+    { value: "18岁以下", label: "18岁以下" },
+    { value: "18-24岁", label: "18-24岁" },
+    { value: "25-30岁", label: "25-30岁" },
+    { value: "30-35岁", label: "30-35岁" },
+    { value: "35-40岁", label: "35-40岁" },
+    { value: "40岁以上", label: "40岁以上" },
+];
+export const createPersonaWizardToolDefinition = {
+    name: "create_persona_wizard",
+    description: "当用户说「创建/新建/自定义评论员/人设/角色」时，调用此工具。首次调用不带 sessionId，将 userMessage 设为用户的原话；工具会引导用户逐步完成年龄段、兴趣方向、性格特质、讲话语气、常用平台、与作者关系等信息收集。完全独立，不涉及内容评测流程。",
+    inputSchema: {
+        type: "object",
+        properties: {
+            sessionId: {
+                type: "string",
+                description: "人设创建向导的会话标识。首次调用请留空，工具会自动生成并返回一个 sessionId。后续调用必须传入此值以继续上一次的创建会话。",
+            },
+            userMessage: {
+                type: "string",
+                description: "用户在当前步骤的回复内容。首次调用时直接传入用户原话（例如「帮我创建一个时尚类评论员」），工具开始分步引导。后续步骤传入用户对工具提问的回复。",
+            },
+        },
+        required: ["userMessage"],
+    },
+};
+export const createPersonaWizardModule = {
+    definition: createPersonaWizardToolDefinition,
+    handler: (deps) => async (args) => {
+        if (!args)
+            throw new Error("向导需要提供参数");
+        const input = args;
+        if (deps.updateClientSamplingSupport()) {
+            input.samplingFn = deps.createMultiTurnSamplingFn();
+        }
+        return await handleCreatePersonaWizard(deps.skillsDir, deps.tmpDir, input);
+    },
+};
+export async function handleCreatePersonaWizard(skillsDir, tmpDir, input) {
+    const { userMessage, samplingFn } = input;
+    if (!userMessage || typeof userMessage !== "string") {
+        return {
+            content: [{ type: "text", text: "❌ 请提供当前步骤的用户回复。" }],
+            isError: true,
+        };
+    }
+    try {
+        await fs.promises.mkdir(tmpDir, { recursive: true });
+        const state = await loadOrCreateState(tmpDir, input.sessionId);
+        if (!input.sessionId) {
+            await saveState(tmpDir, state);
+            return toolResponse(state, [
+                "请选择这个角色的年龄段（回复编号或文字）：",
+                "",
+                "1. 18岁以下",
+                "2. 18-24岁",
+                "3. 25-30岁",
+                "4. 30-35岁",
+                "5. 35-40岁",
+                "6. 40岁以上",
+            ].join("\n"));
+        }
+        const result = await advanceWizard(skillsDir, tmpDir, state, userMessage, samplingFn);
+        return result;
+    }
+    catch (err) {
+        const info = getErrorInfo(err);
+        logger.error("Create persona wizard failed", { event: "wizard_error", error: info.code, message: info.message });
+        return {
+            content: [{ type: "text", text: `❌ 人设创建向导失败：${info.message}` }],
+            isError: true,
+        };
+    }
+}
+async function advanceWizard(skillsDir, tmpDir, state, userMessage, samplingFn) {
+    switch (state.step) {
+        case "ageRange": {
+            const resolved = resolveAgeRange(userMessage);
+            if (!resolved) {
+                return toolResponse(state, [
+                    "请从以下选项中选择（回复编号或文字）：",
+                    "",
+                    "1. 18岁以下",
+                    "2. 18-24岁",
+                    "3. 25-30岁",
+                    "4. 30-35岁",
+                    "5. 35-40岁",
+                    "6. 40岁以上",
+                ].join("\n"));
+            }
+            state.fields.ageRange = resolved;
+            state.step = "interests";
+            await saveState(tmpDir, state);
+            await saveDraft(tmpDir, state);
+            return toolResponse(state, [
+                `已记录年龄段：${state.fields.ageRange}`,
+                "",
+                "第二步：告诉我这个角色的日常兴趣与关注焦点？",
+            ].join("\n"));
+        }
+        case "interests": {
+            const extracted = await extractInterests(userMessage, samplingFn);
+            state.fields.interests = normalizeStringArray(extracted.value);
+            state.step = "traits";
+            await saveState(tmpDir, state);
+            await saveDraft(tmpDir, state);
+            return toolResponse(state, [
+                extracted.assistantMessage,
+                "",
+                "第三步：请描述这个角色的性格特质",
+                "自由描述即可，例如：容易跟风、对价格敏感、喜欢对比评测……",
+            ].join("\n"));
+        }
+        case "traits": {
+            const extracted = await extractTraits(userMessage, samplingFn);
+            state.fields.traits = normalizeStringArray(extracted.value);
+            state.step = "tone";
+            await saveState(tmpDir, state);
+            await saveDraft(tmpDir, state);
+            return toolResponse(state, [
+                extracted.assistantMessage,
+                "",
+                "第四步：请描述这个角色的讲话语气",
+                "自由描述即可，例如：毒舌犀利、温柔耐心、幽默风趣、一本正经……",
+            ].join("\n"));
+        }
+        case "tone": {
+            const toneExtracted = await extractTone(userMessage, samplingFn);
+            state.fields.tone = normalizeStringArray(toneExtracted.value);
+            state.step = "platform";
+            await saveState(tmpDir, state);
+            await saveDraft(tmpDir, state);
+            return toolResponse(state, [
+                toneExtracted.assistantMessage,
+                "",
+                "第五步：内容主要投放平台",
+                "如果多平台投放，建议创建更多针对某一平台的虚拟评论员。",
+            ].join("\n"));
+        }
+        case "platform": {
+            const raw = userMessage.trim();
+            if (/[和、/,]/.test(raw)) {
+                const first = raw.split(/[和、/,]/)[0].trim();
+                if (first) {
+                    state.fields.platform = first;
+                    state.fields.platformNote = `⚠️ 你输入了多个平台，将围绕「${first}」创建此评论员。如需覆盖其他平台，请另行创建。`;
+                    state.step = "authorRelation";
+                    await saveState(tmpDir, state);
+                    await saveDraft(tmpDir, state);
+                    return toolResponse(state, [
+                        `已记录常用平台：${state.fields.platform}`,
+                        state.fields.platformNote,
+                        "",
+                        "请选择这个角色与作者的关系（回复编号或文字）：",
+                        "",
+                        "1. 已关注（信任阈值较高，但期望值也更高）",
+                        "2. 未关注（信任阈值较低，更容易因细节问题流失注意力）",
+                    ].join("\n"));
+                }
+            }
+            state.fields.platform = raw;
+            state.step = "authorRelation";
+            await saveState(tmpDir, state);
+            await saveDraft(tmpDir, state);
+            return toolResponse(state, [
+                `已记录常用平台：${state.fields.platform}`,
+                "",
+                "请选择这个角色与作者的关系（回复编号或文字）：",
+                "",
+                "1. 已关注（信任阈值较高，但期望值也更高）",
+                "2. 未关注（信任阈值较低，更容易因细节问题流失注意力）",
+            ].join("\n"));
+        }
+        case "authorRelation": {
+            const resolved = resolveAuthorRelation(userMessage);
+            if (!resolved) {
+                return toolResponse(state, [
+                    "请从以下选项中选择（回复编号或文字）：",
+                    "",
+                    "1. 已关注",
+                    "2. 未关注",
+                ].join("\n"));
+            }
+            state.fields.authorRelation = resolved;
+            // Infer name, gender, culturalContext, stance, blindSpot before showing preview
+            const inferred = await inferFinalFields(state, skillsDir, samplingFn);
+            state.fields = { ...state.fields, ...inferred };
+            state.step = "finalConfirm";
+            await saveState(tmpDir, state);
+            await saveDraft(tmpDir, state);
+            return toolResponse(state, buildFinalConfirmationMessage(state, skillsDir));
+        }
+        case "finalConfirm":
+            if (isAffirmative(userMessage)) {
+                return completeWizard(skillsDir, tmpDir, state, samplingFn);
+            }
+            {
+                const modified = await applyFinalModification(state, userMessage, samplingFn);
+                if (!modified) {
+                    return toolResponse(state, "请说明要修改哪个字段：名字、性别、年龄段、兴趣方向、性格特质或常用平台。");
+                }
+                await saveState(tmpDir, state);
+                await saveDraft(tmpDir, state);
+                return toolResponse(state, [`已更新${fieldLabel(modified)}。`, "", buildFinalConfirmationMessage(state, skillsDir)].join("\n"));
+            }
+        case "completed":
+            return toolResponse(state, "这个人设创建流程已经完成。需要创建新角色时，请重新开始一个会话。");
+    }
+}
+async function applyFinalModification(state, userMessage, samplingFn) {
+    const field = detectModifiedField(userMessage);
+    if (!field)
+        return undefined;
+    const valueText = extractModificationValue(userMessage, field);
+    if (!valueText)
+        return undefined;
+    switch (field) {
+        case "name":
+            state.fields.personaName = valueText;
+            break;
+        case "gender":
+            state.fields.gender = valueText === "男" || valueText === "女" ? valueText : "未指定";
+            break;
+        case "ageRange":
+            state.fields.ageRange = valueText;
+            break;
+        case "interests": {
+            const extracted = await extractInterests(valueText, samplingFn);
+            state.fields.interests = normalizeStringArray(extracted.value);
+            break;
+        }
+        case "traits": {
+            const extracted = await extractTraits(valueText, samplingFn);
+            state.fields.traits = normalizeStringArray(extracted.value);
+            break;
+        }
+        case "tone": {
+            const extracted = await extractTone(valueText, samplingFn);
+            state.fields.tone = normalizeStringArray(extracted.value);
+            break;
+        }
+        case "platform":
+            state.fields.platform = valueText;
+            break;
+        case "authorRelation":
+            state.fields.authorRelation = valueText;
+            break;
+    }
+    state.step = "finalConfirm";
+    return field;
+}
+async function completeWizard(skillsDir, tmpDir, state, _samplingFn) {
+    state.step = "completed";
+    await saveState(tmpDir, state);
+    await saveDraft(tmpDir, state);
+    const createResult = await handleCreatePersona(skillsDir, tmpDir, {
+        name: state.fields.personaName || inferPersonaName(state),
+        sessionId: state.sessionId,
+        culturalContext: state.fields.culturalContext,
+        authorRelation: state.fields.authorRelation,
+        stance: state.fields.stance,
+        blindSpot: state.fields.blindSpot,
+        gender: state.fields.gender,
+    });
+    if (!createResult.isError) {
+        await cleanupState(tmpDir, state.sessionId);
+    }
+    return createResult;
+}
+async function extractInterests(userMessage, samplingFn) {
+    if (samplingFn) {
+        try {
+            const json = await runJsonExtraction(samplingFn, {
+                systemPrompt: "你是字段提炼器。请把用户对兴趣方向的自然语言描述提炼为最多 3 个中文短标签，并严格输出 JSON：{\"interests\":[\"标签\"],\"assistantMessage\":\"整理说明\"}。assistantMessage 可自由给出贴合场景的引导性举例辅助说明，再总结标签。不要要求用户确认。不要输出 markdown。",
+                userMessage,
+            });
+            const interests = normalizeStringArray(json.interests).slice(0, 3);
+            if (interests.length > 0) {
+                return {
+                    value: interests,
+                    assistantMessage: typeof json.assistantMessage === "string"
+                        ? sanitizeStepAssistantMessage(json.assistantMessage)
+                        : `我帮你总结为以下标签：${interests.join("、")}。`,
+                };
+            }
+        }
+        catch (err) {
+            const info = getErrorInfo(err);
+            logger.warn("Sampling extraction failed for interests, falling back to heuristic", {
+                event: "sampling_interests_fallback",
+                error: info.code,
+                message: info.message,
+            });
+        }
+    }
+    const interests = splitUserText(userMessage, 3);
+    return {
+        value: interests,
+        assistantMessage: `我帮你总结为以下标签：${interests.join("、")}。`,
+    };
+}
+async function extractTraits(userMessage, samplingFn) {
+    if (samplingFn) {
+        try {
+            const json = await runJsonExtraction(samplingFn, {
+                systemPrompt: "你是字段提炼器。请把用户对性格特质的自然语言描述提炼为最多 4 条「特质 → 行为描述」字符串，并严格输出 JSON：{\"traits\":[\"特质 → 因此当 X 时，会 Y\"],\"assistantMessage\":\"整理说明\"}。assistantMessage 只说明已总结的性格特质，不要要求用户确认。不要输出 markdown。",
+                userMessage,
+            });
+            const traits = normalizeStringArray(json.traits).slice(0, 4).map(normalizeTrait);
+            if (traits.length > 0) {
+                return {
+                    value: traits,
+                    assistantMessage: typeof json.assistantMessage === "string"
+                        ? sanitizeStepAssistantMessage(json.assistantMessage)
+                        : `我帮你总结为以下性格特质：\n${traits.map((t) => `- ${t}`).join("\n")}`,
+                };
+            }
+        }
+        catch (err) {
+            const info = getErrorInfo(err);
+            logger.warn("Sampling extraction failed for traits, falling back to heuristic", {
+                event: "sampling_traits_fallback",
+                error: info.code,
+                message: info.message,
+            });
+        }
+    }
+    const traits = splitUserText(userMessage, 4).map(normalizeTrait);
+    return {
+        value: traits,
+        assistantMessage: `我帮你总结为以下性格特质：\n${traits.map((t) => `- ${t}`).join("\n")}`,
+    };
+}
+async function extractTone(userMessage, samplingFn) {
+    if (samplingFn) {
+        try {
+            const json = await runJsonExtraction(samplingFn, {
+                systemPrompt: "你是字段提炼器。请把用户对讲话语气的自然语言描述提炼为最多 4 个中文短标签（每个标签是独立描述，简洁自然），并严格输出 JSON：{\"tone\":[\"标签\"],\"assistantMessage\":\"整理说明\"}。assistantMessage 只说明已总结的语气特点，不要要求用户确认。不要输出 markdown。",
+                userMessage,
+            });
+            const tone = normalizeStringArray(json.tone).slice(0, 4);
+            if (tone.length > 0) {
+                return {
+                    value: tone,
+                    assistantMessage: typeof json.assistantMessage === "string"
+                        ? sanitizeStepAssistantMessage(json.assistantMessage)
+                        : `我帮你总结为以下讲话特点：\n${tone.map((t) => `- ${t}`).join("\n")}`,
+                };
+            }
+        }
+        catch (err) {
+            const info = getErrorInfo(err);
+            logger.warn("Sampling extraction failed for tone, falling back to heuristic", {
+                event: "sampling_tone_fallback",
+                error: info.code,
+                message: info.message,
+            });
+        }
+    }
+    const tone = splitUserText(userMessage, 4);
+    return {
+        value: tone,
+        assistantMessage: `我帮你总结为以下讲话特点：\n${tone.map((t) => `- ${t}`).join("\n")}`,
+    };
+}
+async function inferFinalFields(state, skillsDir, samplingFn) {
+    const interest = state.fields.interests?.[0] || "内容";
+    const fallback = {
+        culturalContext: inferCulturalContext(state),
+        authorRelation: state.fields.authorRelation || "未关注",
+        stance: "默认质疑",
+        blindSpot: "无特定盲区",
+        personaName: inferPersonaName(state),
+        gender: "未指定",
+    };
+    const platformRefs = loadPlatformStanceBlindSpotRefs(state, skillsDir);
+    if (!samplingFn)
+        return fallback;
+    try {
+        const json = await runJsonExtraction(samplingFn, {
+            systemPrompt: [
+                "你是人设属性推断器。根据已确认字段推断以下字段，必须全部填写，不能为空：",
+                "- personaName：有创意、像真实互联网网名，不要带「评论员」后缀，也不要带平台名",
+                "- gender：男 / 女 / 未指定",
+                "- culturalContext",
+                "- stance：该角色看问题的基本立场（参考同平台已有角色的风格，但不要照搬）",
+                "- blindSpot：该角色因自身局限可能忽略的视角（参考同平台已有角色的风格，但不要照搬）",
+                platformRefs ? `\n同平台已有人设参考：\n${platformRefs}` : "",
+                `\n严格输出 JSON：{"personaName":"...","gender":"...","culturalContext":"...","stance":"...","blindSpot":"..."}`,
+                "authorRelation 已由用户明确选择，不要覆盖。不要输出 markdown。",
+            ].filter(Boolean).join("\n"),
+            userMessage: JSON.stringify(state.fields),
+        });
+        return {
+            personaName: typeof json.personaName === "string" && json.personaName.trim().length > 0
+                ? json.personaName.trim()
+                : fallback.personaName,
+            gender: typeof json.gender === "string" && (json.gender === "男" || json.gender === "女")
+                ? json.gender
+                : "未指定",
+            culturalContext: typeof json.culturalContext === "string" ? json.culturalContext : fallback.culturalContext,
+            authorRelation: fallback.authorRelation,
+            stance: typeof json.stance === "string" && json.stance.trim().length > 0 ? json.stance.trim() : fallback.stance,
+            blindSpot: typeof json.blindSpot === "string" && json.blindSpot.trim().length > 0 ? json.blindSpot.trim() : fallback.blindSpot,
+        };
+    }
+    catch {
+        return fallback;
+    }
+}
+function loadPlatformStanceBlindSpotRefs(state, skillsDir) {
+    const subDir = getSubDirFromDraft({ fields: state.fields });
+    if (!subDir)
+        return "";
+    const dir = path.join(skillsDir, subDir);
+    try {
+        if (!fs.statSync(dir).isDirectory())
+            return "";
+        const files = fs.readdirSync(dir).filter(f => f.endsWith(".md") && f !== "_template.md");
+        if (files.length === 0)
+            return "";
+        const refs = [];
+        for (const file of files) {
+            const content = fs.readFileSync(path.join(dir, file), "utf-8");
+            const nameMatch = content.match(/^name:\s*(.+)/m);
+            const stanceMatch = content.match(/^stance:\s*(.+)/m);
+            const blindSpotMatch = content.match(/^blindSpot:\s*(.+)/m);
+            const name = nameMatch ? nameMatch[1].trim() : file.replace(".md", "");
+            const stance = stanceMatch ? stanceMatch[1].trim() : "（未设置）";
+            const blindSpot = blindSpotMatch ? blindSpotMatch[1].trim() : "（未设置）";
+            refs.push(`- ${name}：立场="${stance}"，盲区="${blindSpot}"`);
+        }
+        return refs.join("\n");
+    }
+    catch {
+        return "";
+    }
+}
+async function runJsonExtraction(samplingFn, params) {
+    const response = await samplingFn({
+        systemPrompt: params.systemPrompt,
+        messages: [{ role: "user", content: params.userMessage }],
+        maxTokens: 1024,
+    });
+    const jsonText = stripCodeFence(response.content.trim());
+    return JSON.parse(jsonText);
+}
+function stripCodeFence(text) {
+    if (!text.startsWith("```"))
+        return text;
+    return text.replace(/^```json\s*/, "").replace(/^```\s*/, "").replace(/\s*```$/, "").trim();
+}
+async function loadOrCreateState(tmpDir, inputSessionId) {
+    if (inputSessionId && !/^[a-z0-9-]+$/.test(inputSessionId)) {
+        throw new Error("sessionId 格式不合法。");
+    }
+    const sessionId = inputSessionId || `wizard-create-${Math.random().toString(36).substring(2, 10)}`;
+    const statePath = getStatePath(tmpDir, sessionId);
+    if (inputSessionId && fs.existsSync(statePath)) {
+        const raw = await fs.promises.readFile(statePath, "utf-8");
+        return JSON.parse(raw);
+    }
+    return {
+        sessionId,
+        createdAt: Date.now(),
+        step: "ageRange",
+        fields: {},
+    };
+}
+async function saveState(tmpDir, state) {
+    await fs.promises.mkdir(tmpDir, { recursive: true });
+    const statePath = getStatePath(tmpDir, state.sessionId);
+    const tmpPath = statePath + ".tmp";
+    await fs.promises.writeFile(tmpPath, JSON.stringify(state, null, 2), "utf-8");
+    await fs.promises.rename(tmpPath, statePath);
+}
+async function saveDraft(tmpDir, state) {
+    const draft = {
+        sessionId: state.sessionId,
+        createdAt: state.createdAt,
+        step: stepNumber(state),
+        fields: state.fields,
+    };
+    const draftPath = getDraftPath(tmpDir, state.sessionId);
+    const tmpPath = draftPath + ".tmp";
+    await fs.promises.writeFile(tmpPath, JSON.stringify(draft, null, 2), "utf-8");
+    await fs.promises.rename(tmpPath, draftPath);
+}
+async function cleanupState(tmpDir, sessionId) {
+    for (const filePath of [getStatePath(tmpDir, sessionId), getDraftPath(tmpDir, sessionId)]) {
+        try {
+            if (fs.existsSync(filePath))
+                await fs.promises.unlink(filePath);
+        }
+        catch (err) {
+            const info = getErrorInfo(err);
+            logger.warn("Failed to clean wizard file", { event: "wizard_cleanup_error", path: filePath, error: info.code, message: info.message });
+        }
+    }
+}
+function getStatePath(tmpDir, sessionId) {
+    return path.join(tmpDir, `${sessionId}_wizard.json`);
+}
+function getDraftPath(tmpDir, sessionId) {
+    return path.join(tmpDir, `${sessionId}_draft.json`);
+}
+function toolResponse(state, assistantMessage) {
+    return {
+        content: [
+            {
+                type: "text",
+                text: [
+                    assistantMessage,
+                    "",
+                    "```kevlar-state",
+                    `sessionId: ${state.sessionId}`,
+                    "workflow: create_persona",
+                    `currentStep: ${state.step}`,
+                    `completedFields: ${Object.keys(state.fields).join(", ") || "none"}`,
+                    "```",
+                ]
+                    .filter(Boolean)
+                    .join("\n"),
+            },
+        ],
+    };
+}
+function buildFinalConfirmationMessage(state, skillsDir) {
+    const fields = state.fields;
+    // Build a preview of the persona file that will be created
+    const previewDraft = { fields };
+    const subDir = getSubDirFromDraft(previewDraft);
+    const baseId = generateIdFromDraft(previewDraft) ||
+        inferPersonaName(state).replace(/[^a-zA-Z0-9_]/g, "").toLowerCase() ||
+        `persona_${Math.random().toString(36).substring(2, 8)}`;
+    const id = applyDedup(skillsDir, baseId, subDir);
+    const personaName = fields.personaName || inferPersonaName(state);
+    const culturalContext = fields.culturalContext || "未提供";
+    const authorRelation = fields.authorRelation || "未关注";
+    const stance = fields.stance || "默认质疑";
+    const blindSpot = fields.blindSpot || "无特定盲区";
+    const gender = fields.gender || undefined;
+    const lines = [
+        "基本信息：",
+        `角色名：${personaName}`,
+        `ID：${id}`,
+        `年龄段：${fields.ageRange || ""}`,
+    ];
+    if (gender)
+        lines.push(`性别：${gender}`);
+    lines.push(`兴趣方向：${Array.isArray(fields.interests) ? fields.interests.join("、") : ""}`, `常用平台：${fields.platform || ""}`, `文化背景：${culturalContext}`, `立场：${stance}`, `盲区：${blindSpot}`, "性格特质：");
+    if (Array.isArray(fields.traits)) {
+        fields.traits.forEach((t) => lines.push(`  ${t}`));
+    }
+    lines.push("讲话语气：");
+    if (Array.isArray(fields.tone)) {
+        fields.tone.forEach((t) => lines.push(`  ${t}`));
+    }
+    lines.push(`与作者关系：${authorRelation}`);
+    if (fields.platformNote) {
+        lines.push("", fields.platformNote);
+    }
+    lines.push("", "如需修改，请直接说：名字改成... / 性别改成... / 年龄段改成... / 兴趣方向改成... / 性格特质改成... / 讲话语气改成... / 平台改成... / 关系改成...", "确认无误请回复：确认创建");
+    return lines.join("\n");
+}
+function detectModifiedField(input) {
+    if (/名字|名|名称|角色名|称呼/.test(input))
+        return "name";
+    if (/性别|男|女/.test(input))
+        return "gender";
+    if (/年龄|年龄段|\d+\s*[-到至]\s*\d+\s*岁|岁/.test(input))
+        return "ageRange";
+    if (/兴趣|方向|标签/.test(input))
+        return "interests";
+    if (/性格|特质|脾气|行为/.test(input))
+        return "traits";
+    if (/语气|讲话|说话|口吻/.test(input))
+        return "tone";
+    if (/平台|渠道|小红书|知乎|B站|公众号|微信|微博|Instagram|Reddit|YouTube|Twitter|X\b/i.test(input)) {
+        return "platform";
+    }
+    if (/关系|关注|作者/.test(input))
+        return "authorRelation";
+    return undefined;
+}
+function extractModificationValue(input, field) {
+    const trimmed = input.trim();
+    const explicitMatch = trimmed.match(/(?:改成|改为|修改为|换成|变成|调整为|设置为|[:：])\s*(.+)$/);
+    if (explicitMatch?.[1])
+        return explicitMatch[1].trim();
+    const fieldWords = {
+        name: /名字|名|名称|角色名|称呼/g,
+        gender: /性别/g,
+        ageRange: /年龄段?|岁数?/g,
+        interests: /兴趣方向|兴趣|方向|标签/g,
+        traits: /性格特质|性格|特质|脾气|行为/g,
+        tone: /讲话语气|语气|讲话|说话|口吻/g,
+        platform: /常用平台|平台|渠道/g,
+        authorRelation: /与作者的关系|关系|关注/g,
+    };
+    const cleaned = trimmed
+        .replace(fieldWords[field], "")
+        .replace(/^(请|帮我|把|将|给我|重新)?\s*/, "")
+        .replace(/(改一下|修改一下|调整一下|改|修改|调整|换)\s*/g, "")
+        .trim();
+    return cleaned || trimmed;
+}
+function fieldLabel(field) {
+    const labels = {
+        name: "名字",
+        gender: "性别",
+        ageRange: "年龄段",
+        interests: "兴趣方向",
+        traits: "性格特质",
+        tone: "讲话语气",
+        platform: "常用平台",
+        authorRelation: "与作者的关系",
+    };
+    return labels[field];
+}
+function sanitizeStepAssistantMessage(input) {
+    return input
+        .replace(/确认没问题吗？?如需调整请直接告诉我。?/g, "")
+        .replace(/确认没问题吗？?/g, "")
+        .replace(/如需调整请直接告诉我。?/g, "")
+        .trim();
+}
+function isAffirmative(input) {
+    const normalized = input.trim().toLowerCase();
+    // Short/ambiguous words require exact match to avoid false positives.
+    // e.g. "确认没问题" should match via containsMatchWords; "这是什么" should NOT match.
+    const exactMatchWords = ["是", "对", "好", "y"];
+    const containsMatchWords = ["确认", "可以", "没问题", "ok", "yes"];
+    return (exactMatchWords.some((w) => normalized === w) ||
+        containsMatchWords.some((w) => normalized.includes(w)));
+}
+function normalizeStringArray(value) {
+    if (!Array.isArray(value))
+        return [];
+    return value.map((item) => String(item).trim()).filter(Boolean);
+}
+function splitUserText(input, maxItems) {
+    const parts = input
+        .split(/[，,、；;\n]/)
+        .map((part) => part.trim())
+        .filter(Boolean);
+    return (parts.length > 0 ? parts : [input.trim()]).slice(0, maxItems);
+}
+function normalizeTrait(input) {
+    const text = input.trim();
+    if (text.includes("→"))
+        return text;
+    return `${text} → 因此在相关内容判断中会表现出这一倾向`;
+}
+function resolveAgeRange(input) {
+    const trimmed = input.trim();
+    const num = parseInt(trimmed, 10);
+    if (num >= 1 && num <= AGE_RANGE_OPTIONS.length) {
+        return AGE_RANGE_OPTIONS[num - 1].value;
+    }
+    const exact = AGE_RANGE_OPTIONS.find((o) => o.value === trimmed);
+    if (exact)
+        return exact.value;
+    const partial = AGE_RANGE_OPTIONS.find((o) => o.value.replace("岁", "") === trimmed);
+    if (partial)
+        return partial.value;
+    return null;
+}
+function resolveAuthorRelation(input) {
+    const trimmed = input.trim();
+    const num = parseInt(trimmed, 10);
+    if (num === 1)
+        return "已关注";
+    if (num === 2)
+        return "未关注";
+    if (/已关注|关注了|已关/.test(trimmed))
+        return "已关注";
+    if (/未关注|没关注|未关/.test(trimmed))
+        return "未关注";
+    return null;
+}
+function inferCulturalContext(state) {
+    const platform = state.fields.platform || "";
+    if (/小红书|微信|公众号|知乎|B站|抖音/.test(platform)) {
+        return "中国大陆互联网文化语境";
+    }
+    if (/Instagram|Reddit|YouTube|Twitter|X/.test(platform)) {
+        return "海外互联网文化语境";
+    }
+    return "未提供";
+}
+function inferPersonaName(state) {
+    const interest = state.fields.interests?.[0] || "内容";
+    return interest;
+}
+function stepNumber(state) {
+    const order = [
+        "ageRange",
+        "interests",
+        "traits",
+        "tone",
+        "platform",
+        "authorRelation",
+        "finalConfirm",
+        "completed",
+    ];
+    return Math.max(1, order.indexOf(state.step) + 1);
+}
+//# sourceMappingURL=createPersonaWizardTool.js.map