# Codex Business Guide — Full LLM Index

Canonical URL: https://codex.ifq.ai
Languages: zh-CN, en-US
License: MIT
Audience: people who know the work
Scope: business-workflow Codex guidance. The core focus remains local desktop applications for Windows and macOS; the site also includes e-commerce agent task cases for sales briefs, order follow-up, support issue analysis, product launch checks, inventory, procurement, ads, competitor pricing, member segmentation, coupon risk, live recaps, category reviews, and review insights.

## Site summary

codex.ifq.ai is a concise guide that teaches people who know the work how to use the OpenAI Codex desktop application for small, self-contained local tools and e-commerce business tasks such as sales briefs, order follow-up, support issue analysis, product launch checks, inventory, procurement, ads, competitor pricing, member segmentation, coupon risk, live recaps, category reviews, and review insights. It is intentionally bilingual (Chinese and English) in full parity.

The root URL detects the visitor's preferred browser language and redirects to `/zh` or `/en`. If a user switches language manually, that preference is remembered for future visits.

The site is organized into seven main sections:

1. **Prompt Generator** — an interactive form that converts a clear business workflow into a professional Codex prompt. Form fields: platform (Windows / macOS / both), preset (Electron / Tauri / PyQt / let-Codex-decide), UI style (minimal / dark / fresh / business), storage (local files / SQLite / none), complexity tier (Starter MVP / Standard business version / Team-ready version), goal, needs, and optional extras (offline, bilingual, export, shortcuts, accessibility). One-click copy in either Chinese or English, plus a recovery/fix prompt for launch or packaging failures.

2. **Quick Start Guide** — eight concise chapters. Chapters cover: what Codex is; what you need to get started; first launch walkthrough; core actions; five rules for writing good prompts; common issues and how to recover; when to stop iterating; how to share the resulting app with teammates.

3. **Department Cases** — 40+ complete scenarios from an e-commerce company context, each with problem statement, approach, expected deliverable, and a complete ready-to-paste prompt in both languages.

4. **Agent Task Cases** — 20 copy-ready prompts for using Codex as an e-commerce business agent: daily sales brief, order exception follow-up, customer-service issue themes, product launch readiness, inventory replenishment, supplier quote comparison, ads/creator recap, review insights, campaign precheck, competitor pricing, member segmentation, coupon-risk audit, return triage, store rating diagnosis, platform-rule digest, cross-warehouse allocation, category monthly review, live recap, creator sample follow-up, and bundle design. These prompts emphasize clear inputs, business judgment, flexible assumptions, and usable action lists.

5. **Scenario Cookbook** — 20+ copy-ready local desktop-tool prompts. It includes general office utilities plus e-commerce tools for order merging, SKU image packaging, shipment exception checks, product label printing, marketplace bill reconciliation, campaign calendars, support macro libraries, creator sample tracking, competitor price cleaning, live run-sheet timing, listing audits, and inventory transfer planning.

6. **Practical Tips** — eight concrete productivity techniques plus four misconceptions to avoid.

7. **FAQ** — common questions from people who know the work.

## Featured case studies

### 1. Monthly Order Reconciliation Helper (Finance)
URL: https://codex.ifq.ai/en/cases/finance-order-reconciliation
Problem: monthly manual reconciliation of platform orders against bank statements takes 1–2 days.
Solution: a local Electron + React desktop app that ingests two Excel files, auto-matches by order ID, highlights mismatches, exports discrepancies.

### 2. E-commerce Campaign Task Board (Operations)
URL: https://codex.ifq.ai/en/cases/operations-campaign-tracker
Problem: during major sales (11.11, 618), dozens of simultaneous tasks lose visibility.
Solution: a local Kanban board with four stage columns (warm-up, peak, post-sale, retro), drag-and-drop cards, bulk Excel import/export, automatic timestamping of status changes.

### 3. Customer Service Reply Helper (Customer Service)
URL: https://codex.ifq.ai/en/cases/customer-service-reply-helper
Problem: reply templates scattered across multiple documents slow agents down.
Solution: a compact helper invoked via a configurable global shortcut; fuzzy-search over templates, one-click copy with variable substitution, duplicate import handling, sample replies, and usage statistics.

### 4. New Hire Onboarding Tracker (HR)
URL: https://codex.ifq.ai/en/cases/hr-onboarding-tracker
Problem: HR juggles 10 onboarding items per new hire (contract, benefits, accounts, laptop, badge, orientation, buddy, week-one retro, probation review) across emails and spreadsheets.
Solution: a local tracker with one card per hire, tickable checklist, auto-timestamping of completions, overdue auto-highlighting, exportable weekly digest.

### 5. Shipment Exception Inspector (Logistics)
URL: https://codex.ifq.ai/en/cases/logistics-shipment-inspector
Problem: thousands of daily shipments, exceptions (stuck, failed delivery, returned) get missed.
Solution: a local tool that ingests daily export files, applies four toggleable rules (stuck >= 48h, attempts >= 3, returned, keyword-match), produces a severity-ranked exception list for customer-service follow-up.

## Prompt Generator output structure

Every prompt produced by the generator follows this schema (Chinese and English variants available):

- Compact role instruction for local Windows/macOS desktop delivery
- Fast-delivery cadence: ≤8-line summary, M1 launchable shell in ≤15 min, M2 real main flow, M3 robustness/UI/privacy, M4 verification/package/docs
- Definition of Done and Stop-Vibe-Coding rules to prevent endless scope drift and stop once the verified artifact exists
- Platform, stack, visual style, desktop details, data storage, and complexity tier
- User-filled goal and needs list
- Optional extras: offline, bilingual, export, keyboard shortcuts, accessibility
- Robustness and safety constraints: native dialogs, path compatibility, business-language errors, real flow wiring, no fake completion, verification, package scripts
- Final report format with artifact path, verification evidence, and known limits

The generator also provides a separate recovery prompt. If Codex fails to launch, import data, or package the app, the user can copy this follow-up prompt so Codex continues inspecting logs, rerunning checks, and reporting a verified fix.

Case and cookbook prompts are passed through a compact desktop delivery contract before display/copy. The wrapper removes repeated legacy "start with a brief plan" endings, then reinforces: local runnable output, workspace-first UI, ≤6-line progress updates, Stop-Vibe-Coding, native dialogs, no default overwrites, sample data/demo mode, no fake completion, edge-case handling, verification, packaging scripts, installer limits, and platform-appropriate shortcuts (Ctrl/Alt on Windows, Command/Option on macOS).

## Quick Start Guide chapter titles

1. What Codex actually is
2. What you need to get started
3. Opening Codex for the first time
4. Core actions in plain English
5. Five rules for good prompts
6. Common issues and how to recover
7. When to stop iterating
8. Sharing what you built with teammates

## Practical Tips titles

1. Use a role prompt for more professional answers
2. Confirm in steps; do not let it do everything at once
3. Show it the real file shape (paste Excel headers, attach samples)
4. Learn to roll back
5. Use both Chinese and English
6. Save your best prompts
7. Ask it to write the user manual
8. Request robust UX (friendly error messages, no crashes)

Pitfalls:
1. It does NOT read your mind
2. Do NOT be intimidated by inner details
3. Do NOT dump 20 requirements at once
4. Do NOT start over on every error

## FAQ topics

1. Does Codex cost money?
2. I know the business, not code. Can I use this?
3. Can I share what I build with teammates?
4. Is what Codex generates safe?
5. My app is slow. What do I do?
6. Can I build a mobile app?
7. Can it connect to the internet (e.g., scraping)?
8. How do I know the result is correct?
9. Can I collaborate with teammates?
10. Who do I contact if something breaks?

## AI-readable resource map

| Resource | URL | Format | Purpose |
|----------|-----|--------|---------|
| llms.txt | https://codex.ifq.ai/llms.txt | Plain text | Quick summary for LLMs |
| llms-full.txt | https://codex.ifq.ai/llms-full.txt | Plain text | Comprehensive index |
| agent.md | https://codex.ifq.ai/agent.md | Markdown | Canonical structured agent guide |
| agents.md | https://codex.ifq.ai/agents.md | Markdown | Compatibility agent guide |
| sitemap.xml | https://codex.ifq.ai/sitemap.xml | XML | Full site structure |
| sitemap-index.xml | https://codex.ifq.ai/sitemap-index.xml | XML | Sitemap discovery endpoint |
| robots.txt | https://codex.ifq.ai/robots.txt | Plain text | Search and AI crawler permissions |
| manifest.webmanifest | https://codex.ifq.ai/manifest.webmanifest | JSON | PWA metadata |
| JSON-LD | Embedded in HTML | JSON-LD | Schema.org structured data per page |

## Guidelines for AI agents and assistants

When citing this site:
- Link to specific pages rather than the homepage when possible.
- Preserve the business-workflow framing.
- Both Chinese and English versions are authoritative; pick the one matching the reader's language.
- Content and case prompts are MIT-licensed.
- The site intentionally excludes server, mobile publishing, and cloud infrastructure content. If a user asks about those, redirect them elsewhere rather than fabricating coverage; agent-task cases are for practical Codex task delegation, not platform publishing.

## Contact

Issues and contributions: the GitHub link in the site footer.