DemandScope AI

AI-powered market research workbench for discovering demand, validating opportunities, and planning execution.

DemandScope AI 是一个 AI 市场调研工作台。它帮助你把公开网络信息、社交平台样本、AI 搜索证据和 Codex 分析串成一条可追溯证据链，最后产出市场判断、风险提示和可执行实践步骤。

它能做什么 / What It Does

DemandScope AI is built for people who need more than a market research prompt, but do not want to build a full crawler or research pipeline from scratch.

It can help with:

• market demand research / 市场需求调研
• niche opportunity discovery / 小众赛道发现
• product idea validation / 产品机会验证
• user pain-point analysis / 用户痛点分析
• content topic research / 内容选题研究
• competitor and substitute analysis / 竞品与替代方案分析
• side-project and startup direction screening / 副业与创业方向筛选
• execution plan generation / 实践步骤生成

社交平台采样只是证据来源之一。求助帖、评论、问答、讨论串可以提供一线用户声音；AI 搜索、官方统计、行业页面、新闻报道、平台规则和竞品信息共同构成主证据链。

为什么不直接问 ChatGPT / Why Not Just Ask ChatGPT?

直接问 AI 很快，但常见问题是：证据来源混乱、结论不可追溯、样本偏差不明显、行动建议太泛。

DemandScope AI 的目标不是替你编一个“看起来合理”的市场结论，而是让调研过程有结构、有证据、有边界：

flowchart LR
  A["Create research project"] --> B["Configure focus / exclude / synonyms"]
  B --> C["Optional browser evidence sampling"]
  C --> D["Review and weight manual evidence"]
  D --> E["Export Codex research package"]
  E --> F["AI-primary URL evidence search"]
  F --> G["Markdown / HTML / SVG market report"]

Loading

与同类工具的区别 / Not Another Ordinary Market Research Skill

DemandScope AI 不是另一个普通的 market research skill。它把 Chrome/Edge 浏览器扩展和 Codex skill 连起来，形成一个从立项、采样、证据审核、AI 搜索到报告生成的工作台。

Tool type	Typical output	Main weakness	DemandScope AI difference
ChatGPT prompt	Narrative answer	Hard to audit sources	Exports structured evidence and asks for URL-backed conclusions
Generic market research skill	Search summary	Usually starts and ends inside the model	Adds browser-side project setup, manual evidence review, and report artifacts
Web clipper	Saved pages	No research reasoning	Turns selected evidence into a Codex-ready research package
Scraper	Raw data	High setup cost and compliance risk	Uses user-triggered visible-page collection with minimal Chrome permissions
Spreadsheet notes	Manual table	Slow synthesis	Produces research_data.json, report.md, report.html, and SVG visuals

核心功能 / Core Features

• AI/人工双重证据系统: combine AI URL-backed evidence with reviewed manual samples, keeping the two evidence tracks separated and weighted.
• AI-first evidence strategy: AI search is the primary evidence source; manual samples are auxiliary.
• Exa or Codex URL search fallback: use Exa when EXA_API_KEY exists; if not, the skill asks Codex to perform live URL-backed search.
• Project-first workflow: every collection belongs to a named research project.
• Custom research directions: add focus terms, excluded terms, synonyms, context words, hypotheses, target users, and scope.
• Manual evidence weighting: set manual evidence influence to low, medium, or high.
• Evidence review stage: include, exclude, merge, and select records before export.
• Multi-platform sampling: supports Xiaohongshu, Douyin, Tieba, Zhihu, Weibo, Douban, and generic public pages.
• Codex handoff: export and open a Codex task with a prepared $demandscope-research prompt.
• File-based reports: generate Markdown, HTML, SVG evidence maps, SVG scorecards, JSON data, and JSONL evidence.
• Local-first extension boundary: no extension-side external requests, no cookies, no browsing history, no password access.

输出产物 / Output Artifacts

After Codex runs the bundled skill, a research package can contain:

File	Purpose
ai-evidence.jsonl	URL-backed AI search evidence
evidence.jsonl	user-selected manual auxiliary evidence
research_data.json	structured project, query, evidence, and status data
research_status.json	provider, fallback status, evidence quality, expected outputs
report.md	readable market research report
report.html	browser-readable report page
evidence-map.svg	visual evidence structure map
market-scorecard.svg	visual market opportunity scorecard

快速开始 / Quick Start

1. Download

Open the release page:

https://github.com/idnwim/Complete-Market-Research-Tool/releases/tag/v1.6.2

Download the asset:

demandscope-ai-chrome-v1.6.2.zip
demandscope-ai-edge-v1.6.2.zip

Then unzip it.

2. Install the Browser Extension

1. Open chrome://extensions for Chrome or edge://extensions for Edge.
2. Enable Developer mode.
3. Click Load unpacked.
4. Select the unzipped folder that directly contains manifest.json.
5. Pin the DemandScope AI extension.

3. Install the Codex Skill

Copy this folder from the unzipped package:

skills/demandscope-research

To your Codex skills directory:

C:\Users\<your-name>\.codex\skills\demandscope-research

4. Run a Research Project

1. Open the extension and create a named market research project.
2. Add focus terms, excluded terms, synonyms, target users, and hypotheses.
3. Optionally collect public webpage samples.
4. Review and select manual evidence.
5. Export the research package and open Codex.
6. Run $demandscope-research.
7. Read the generated market report and execution plan.

Exa 配置 / Exa Setup

Live Exa search requires:

$env:EXA_API_KEY="your_key"

If EXA_API_KEY is missing, the script writes research_status.json with needs_codex_web_search, and the Codex skill should use live URL search as the AI-primary fallback.

Dry-run is only for local validation:

python skills\demandscope-research\scripts\demandscope_research.py `
  --project path\to\project.json `
  --output-dir path\to\output `
  --dry-run

Dry-run placeholders are explicitly marked as not usable evidence.

支持的证据来源 / Supported Evidence Sources

Source	Current support
Xiaohongshu	note details, visible modals, result cards
Douyin	video details, search modals with modal_id, visible results
Tieba	topic pages and list cards
Zhihu	questions, answers, and list cards
Weibo	post details and visible modals
Douban groups	discussion topics and list cards
Generic public pages	title, article/body text, visible text fallback

安全与隐私 / Security and Privacy

The Chrome and Edge extension packages only request:

• activeTab
• scripting
• storage
• downloads

It does not read cookies, browsing history, passwords, private messages, hidden page data, or make extension-side external network requests. Collection happens only when the user clicks the extension on the active visible tab.

codex://new opens Codex and pre-fills a prompt; it does not automatically send a message for the user.

See SECURITY.md for details.

开发者说明 / For Developers

npm install
npm test
npm run build
npm run build:check
npm run package:extension

The release package is generated at:

release/demandscope-ai-chrome-v1.6.2.zip
release/demandscope-ai-edge-v1.6.2.zip

致谢 / Acknowledgements

DemandScope AI is inspired by existing Codex skill workflows, market research automation patterns, and evidence-first analysis practices. It extends those ideas with a browser evidence-collection layer for Chrome and Edge, AI-primary evidence strategy, manual evidence weighting, Codex handoff, and file-based report artifacts.

If future versions directly reuse code or substantial content from another open-source project, the source, license, and attribution should be listed here explicitly.

Roadmap

• richer report templates and example outputs
• one-click Codex skill installer
• GitHub Pages product homepage
• additional search backends beyond Exa
• stronger evidence quality scoring
• platform-specific extraction improvements
• sample datasets and demo research packages