maka-agent 0.2.0-dev.44.20260920 → 0.2.0-dev.46.20260922

package/node_modules/@maka/core/dist/deep-research.js

@@ -1,208 +0,0 @@
-/*
- * Licensed to the Apache Software Foundation (ASF) under one
- * or more contributor license agreements.  See the NOTICE file
- * distributed with this work for additional information
- * regarding copyright ownership.  The ASF licenses this file
- * to you under the Apache License, Version 2.0 (the
- * "License"); you may not use this file except in compliance
- * with the License.  You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing,
- * software distributed under the License is distributed on an
- * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
- * KIND, either express or implied.  See the License for the
- * specific language governing permissions and limitations
- * under the License.
- */
-import { SESSION_START_MODE_SPECS } from './session-start-mode.js';
-export const DEEP_RESEARCH_SESSION_NAME = SESSION_START_MODE_SPECS.deep_research.name;
-export const DEEP_RESEARCH_SESSION_LABEL = SESSION_START_MODE_SPECS.deep_research.labels[0];
-export const DEEP_RESEARCH_WORKFLOW_STEPS = [
-    {
-        title: '先定位入口',
-        body: '读目录、配置、启动链路和测试入口，建立项目地图。',
-    },
-    {
-        title: '再追数据流',
-        body: '沿关键模块、IPC、存储、权限和运行时边界追到真实实现。',
-    },
-    {
-        title: '然后对照参考',
-        body: '把可借鉴点拆成 borrow / diverge / risk / gate。',
-    },
-    {
-        title: '最后给可合入方案',
-        body: '输出文件清单、风险边界和验证命令，不在只读模式里动手改。',
-    },
-];
-export const DEEP_RESEARCH_REPORT_SECTIONS = [
-    {
-        title: '结论先行',
-        body: '用 3-5 条讲清楚真实现状、主要差距和优先建议。',
-    },
-    {
-        title: '源码证据',
-        body: '列出文件、函数、配置、测试和运行时路径，避免只给印象判断。',
-    },
-    {
-        title: '借鉴拆解',
-        body: '每个可借鉴点都写 borrow / diverge / risk / gate。',
-    },
-    {
-        title: '落地改进',
-        body: '给出按小步改进拆分的文件清单、边界和验证命令。',
-    },
-];
-export const DEEP_RESEARCH_SCOPE_OPTIONS = [
-    {
-        label: '快速',
-        body: '只扫入口、关键文件和最可能的数据流，适合已知范围的小问题。',
-    },
-    {
-        label: '标准',
-        body: '默认深度：梳理核心链路、相关测试和主要风险，再给落地建议。',
-    },
-    {
-        label: '深挖',
-        body: '跨模块、参考项目和边界条件多轮追踪；只在用户明确要求时使用。',
-    },
-];
-export const DEEP_RESEARCH_EVIDENCE_CHECKLIST = [
-    {
-        title: '项目入口',
-        body: '先看 README、package/config、启动脚本和目录分层，确认真实运行方式。',
-    },
-    {
-        title: '核心链路',
-        body: '追 UI 入口、IPC/服务、存储、运行时调用和错误处理，不只看表面组件。',
-    },
-    {
-        title: '边界条件',
-        body: '检查权限、隐身模式、token/路径暴露、失败重试和用户可见反馈。',
-    },
-    {
-        title: '验证证据',
-        body: '找对应测试、fixture、smoke 文档和可复现命令；缺口要明确标出来。',
-    },
-];
-export const DEEP_RESEARCH_PROGRESS_CHECKPOINTS = [
-    {
-        title: '先建清单',
-        body: '研究范围超过三个相互关联的点时，先列出可核验的检查项再开始追代码。',
-    },
-    {
-        title: '标当前项',
-        body: '推进时明确当前正在验证哪一项，拿到证据后再进入下一项。',
-    },
-    {
-        title: '记阻塞点',
-        body: '找不到源码、运行时或测试证据时标成 blocked，不用猜测补空白。',
-    },
-    {
-        title: '收敛方案',
-        body: '完成项必须汇总到 borrow / diverge / risk / gate 和可落地改进里。',
-    },
-];
-export const DEEP_RESEARCH_STARTER_PROMPTS = [
-    {
-        label: '研究一个参考项目',
-        prompt: '请只读研究这个项目：先梳理目录结构、核心模块、启动链路、数据流和测试入口，然后列出我们可以借鉴的功能设计、需要规避的风险，以及可落地到 Maka 的改进顺序。',
-    },
-    {
-        label: '完整读一遍参考项目',
-        prompt: '请按深挖范围只读研究这个参考项目：先建立目录和模块地图，再逐层读核心功能、运行时、存储、权限、UI、测试和文档；每个可借鉴点都按 borrow / diverge / risk / gate 输出，并给出 Maka 的落地改进顺序。',
-    },
-    {
-        label: '对比一个功能实现',
-        prompt: '请只读对比这个功能在参考项目和 Maka 里的实现差异：指出关键文件、运行时边界、UI 入口、持久化方式、测试覆盖，以及最小可合入的改进方案。',
-    },
-    {
-        label: '做一次安全边界审计',
-        prompt: '请只读审计这个功能的安全边界：权限、token/密钥流、IPC/renderer 暴露、文件路径、隐私模式、日志与 telemetry。输出 blocking 风险和对应 contract test。',
-    },
-];
-export function isDeepResearchSession(labels) {
-    return Array.isArray(labels) && labels.includes(DEEP_RESEARCH_SESSION_LABEL);
-}
-export const DEEP_RESEARCH_IMPLEMENTATION_PROMPT_MAX_CHARS = 12_000;
-export function buildDeepResearchImplementationPrompt(run) {
-    if (run.status !== 'completed' || !run.handoff || !run.reportArtifactId) {
-        throw new Error('Deep Research implementation handoff requires a completed run');
-    }
-    const lines = [
-        'This is a new implementation task created from a completed read-only Deep Research session.',
-        'The original research session remains read-only. Inspect the current code and present an implementation plan before changing project files.',
-        '',
-        `Research objective: ${run.objective}`,
-        `Source session: ${run.sessionId}`,
-        `Final report artifact: ${run.reportArtifactId}`,
-        `Handoff artifact: ${run.handoff.artifactId}`,
-        '',
-        'Implementation tasks:',
-        ...run.handoff.implementationTasks.map((item) => `- ${item}`),
-        '',
-        'Recommended issues:',
-        ...(run.handoff.recommendedIssues.length > 0
-            ? run.handoff.recommendedIssues.map((item) => `- ${item}`)
-            : ['- None specified.']),
-        '',
-        'Recommended pull requests:',
-        ...(run.handoff.recommendedPullRequests.length > 0
-            ? run.handoff.recommendedPullRequests.map((item) => `- ${item}`)
-            : ['- None specified.']),
-        '',
-        'Verification commands:',
-        ...run.handoff.verificationCommands.map((item) => `- ${item}`),
-    ];
-    const content = lines.join('\n');
-    const characters = Array.from(content);
-    if (characters.length <= DEEP_RESEARCH_IMPLEMENTATION_PROMPT_MAX_CHARS)
-        return content;
-    const marker = '\n[Handoff truncated to the safe composer limit.]';
-    return (characters
-        .slice(0, DEEP_RESEARCH_IMPLEMENTATION_PROMPT_MAX_CHARS - Array.from(marker).length)
-        .join('') + marker);
-}
-export function buildDeepResearchSystemPromptFragment() {
-    return [
-        'Deep research mode is active for this session.',
-        '',
-        'Mode contract:',
-        '- Inspect first. Prefer Read, Glob, Grep, and WebSearch.',
-        '- Do not write, edit, delete, move, or rename user project files; do not install, run migrations, start services, or send network requests unless the user explicitly leaves research mode.',
-        '- The deep_research_* tools are the one write exception: they only update Maka-owned research artifacts and an append-only workspace ledger, never the user project.',
-        '- If implementation is needed, produce a concrete plan with files, risks, and verification commands instead of modifying files.',
-        '- Keep findings source-grounded: name files, functions, configs, tests, and observed behavior.',
-        '- Summarize borrow / diverge / risk / gate when comparing a reference project to Maka.',
-        '',
-        'Durable workspace protocol:',
-        '- Call deep_research_start once with the concrete objective and scope level. After interruption or context compaction, call deep_research_status, then deep_research_read_artifact for the exact saved evidence needed to continue.',
-        '- Knowledge-base stage: archive each important raw source first with deep_research_save_artifact role=source, then save evidence notes that cite those source artifact ids.',
-        '- After each bounded local exploration or web-research substep, call deep_research_record_step with roots or query terms, ignored paths, a stopping condition, expected evidence, inspected files/symbols/URLs, worker run ids, persisted evidence ids, and any blocker.',
-        '- Keep the four durable checklist items current with deep_research_update_checklist. Completed items require evidence artifacts; blocked items require an explicit reason.',
-        '- Checkpoint every meaningful research round with deep_research_checkpoint, including open questions, next steps, related task ids, and the artifacts needed to resume.',
-        '- Report-writing stage: save an outline, then source-backed report_section artifacts for conclusion, source_evidence, borrow_diverge_risk_gate, implementation_recommendations, and verification. Mark each section completed only when it is ready.',
-        '- Save one final role=report artifact and one role=handoff artifact. The handoff must turn findings into implementation tasks, recommended issues and/or PRs, and verification commands without performing project writes.',
-        '- Call deep_research_complete only after every checklist item is completed or explicitly skipped, all five report sections are completed, and both report and handoff artifacts are persisted.',
-        '',
-        'Research workflow:',
-        ...DEEP_RESEARCH_WORKFLOW_STEPS.map((step) => `- ${step.title}: ${step.body}`),
-        '',
-        'Research scope budget:',
-        ...DEEP_RESEARCH_SCOPE_OPTIONS.map((option) => `- ${option.label}: ${option.body}`),
-        '- If the user does not specify a scope, use 标准. Use 深挖 only when the user explicitly asks for deep / exhaustive / full-project research.',
-        '',
-        'Evidence checklist:',
-        ...DEEP_RESEARCH_EVIDENCE_CHECKLIST.map((item) => `- ${item.title}: ${item.body}`),
-        '- If any checklist area cannot be verified from available files or runtime context, call that out explicitly instead of guessing.',
-        '',
-        'Progress checkpoints:',
-        ...DEEP_RESEARCH_PROGRESS_CHECKPOINTS.map((item) => `- ${item.title}: ${item.body}`),
-        '- Treat the checklist as a control loop for multi-step research, not as a hidden task system. Keep it visible in the answer or status update when the research spans multiple modules.',
-        '',
-        'Final report contract:',
-        ...DEEP_RESEARCH_REPORT_SECTIONS.map((section) => `- ${section.title}: ${section.body}`),
-    ].join('\n');
-}