diff --git a/docs/03-use-cases/01-general/06-data-analysis-agent.md b/docs/03-use-cases/01-general/06-data-analysis-agent.md index 5aa1c9d..09e14a6 100644 --- a/docs/03-use-cases/01-general/06-data-analysis-agent.md +++ b/docs/03-use-cases/01-general/06-data-analysis-agent.md @@ -169,5 +169,5 @@ keywords: [数据分析, 报告生成, 可视化, Excel, 财务报告] | 格式一致性 | 易出错 | 高 | 高 | :::tip 使用建议 -何洁这类周期性报告最适合先固化模板和指标口径。模板稳定后，每月主要检查异常解释和管理建议，不需要反复调整版式。 +这类周期性报告最适合先固化模板和指标口径。模板稳定后，每月主要检查异常解释和管理建议，不需要反复调整版式。 ::: diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/01-document-writing.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/01-document-writing.md index 7c9a597..464380f 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/01-document-writing.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/01-document-writing.md @@ -1,20 +1,26 @@ --- title: Document Writing -description: Use the AI Copywriter agent to efficiently complete various writing tasks from outline to final draft, including project reports and proposal documents. +description: Use the AI Copywriter agent to organize ideas, generate outlines, draft project reports and proposal documents. keywords: [Document Writing, AI Copywriter, Reports, Writing Assistant, Content Creation] --- # Document Writing -## Scenario Description +## Real Workflow -You need to write a project report but don't know where to start facing a blank page. Or you already have ideas, but organizing them into a well-structured, professionally expressed document takes a lot of time. +This scenario doesn't work by "letting AI write a report from scratch"—it requires the user to already have a batch of scattered but real materials that need to be organized into deliverable documents. -The AI Copywriter agent can help you from scratch: first organize ideas to generate an outline, then gradually expand it into a complete draft, and finally make precise modifications based on your feedback. The entire process is like collaborating with an experienced writing partner. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | Project review, client report, quarterly summary, proposal review requiring formal materials | +| Existing Materials | Project background, key data, meeting minutes, chat conclusions, existing PPTs or spreadsheets | +| Pain Point | Information scattered, structure undecided, tone and audience unclear, repeated format changes time-consuming | +| DesireCore Intervention | AI Copywriter first organizes outline, then generates drafts by chapter; data or charts can be delegated to Data Analyst when needed | +| Acceptance Result | Person in charge receives a reviewable document draft, then supplements facts, sets conclusions, adjusts tone | ## Recommended Agent -**AI Copywriter** -- A knowledge-based agent focused on content creation and copywriting, excelling in brand copy, marketing planning, and multi-style adaptation. +**AI Copywriter** — Suitable for reports, proposals, announcements, marketing copy and other text tasks; generates drafts according to your specified audience, tone and template. ## Complete Conversation Example @@ -70,17 +76,25 @@ AI Copywriter: Adjusted to a more formal business style and added quantitative m ## Key Steps -1. **Clarify Requirements** -- Tell the agent the document type, purpose, and target audience -2. **Generate Outline** -- The agent generates a structured outline based on your information -3. **Write Draft** -- After confirming the outline, expand each chapter into complete content -4. **Review and Revise** -- You provide revision suggestions, and the agent precisely adjusts style, wording, and data +1. **Define Deliverable Audience** — First clarify who the document is for: client, boss, review committee, archiving system; different audiences mean different structure and tone +2. **Organize Fact Package** — List project scope, deliverables, timeline, budget, key data, outstanding issues and lessons learned +3. **Review Structure Before Writing** — Confirm chapters cover background, completed work, result comparison, issues and next steps before expanding each section +4. **Complete Evidence Chain** — For each conclusion, add data, screenshots, links or sources to avoid just having nice words +5. **Generate Versions by Use Case** — The same materials can generate formal reports, summary briefs, email versions or archive versions ## Final Deliverable -A well-structured, professionally expressed project summary report with quantitative data support and clear logical flow, ready for presentation or archiving. +A complete, well-structured document draft with traceable factual basis, chapter logic and reusable versions. You can continue having the AI Copywriter supplement data, adjust tone, compress length or create presentation versions. + +## DesireCore Capabilities Used + +- **Super Document**: View AI modifications in Diff format, accept, reject or continue editing each change individually +- **Natural Language Teaching**: Teach report templates, word prohibitions, writing styles to the AI Copywriter for future reuse +- **Multi-Agent Collaboration**: Data, charts, legal clauses etc. can be delegated to corresponding agents first, then integrated by AI Copywriter :::tip Best Practices -- Provide sufficient background information (data, timelines, key stakeholders) in the first communication for more accurate content generation -- Make good use of "Super Document Mode" which can coordinate multiple agents in one document, such as having the Data Analyst provide chart data, then having the AI Copywriter integrate it into the report -- If you're not satisfied with a paragraph, point out specific issues (e.g., "tone too casual", "lacks data support") rather than vaguely saying "make it better" -::: +- Provide background information, data, timeline and key people in your first communication; minimize having the agent fill in assumptions +- When dissatisfied with a paragraph, point out specific issues like "tone too casual", "lacks data support", "this section doesn't match facts" +- For charts, data conclusions or contract clauses, let corresponding agents process them first in Super Document mode, then have AI Copywriter integrate into the report +- Project summary documents should not just list "what was done"—also include whether scope was completed, timeline/budget deviations, outstanding issues and lessons learned +::: \ No newline at end of file diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/02-data-analysis.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/02-data-analysis.md index 2792e3c..802641b 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/02-data-analysis.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/02-data-analysis.md @@ -1,24 +1,30 @@ --- title: Data Analysis -description: Use the Data Analyst agent to quickly analyze sales data, user behavior, etc., and obtain visual reports and trend insights. +description: Use the Data Analyst agent to clean spreadsheets, calculate metrics, generate charts and organize analysis conclusions. keywords: [Data Analysis, Reports, Visualization, Trend Prediction, Data Analyst] --- # Data Analysis -## Scenario Description +## Real Workflow -You have a sales data spreadsheet, and your boss wants trend analysis and next quarter's forecast by this afternoon. Manually organizing data, creating charts, and writing analysis conclusions would take at least half a day. +This scenario usually happens when "data exists but explanation is missing." Users don't want a vague analysis—they want a specific business question answered. -The Data Analyst agent can complete this work in minutes: automatically identifying data formats, cleaning outliers, generating statistical summaries, creating visual charts, and providing actionable analysis conclusions. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | Weekly meeting, business review, abnormal metrics review, boss suddenly asks about a business fluctuation | +| Existing Materials | CSV/Excel data, metric definitions, target values, historical comparison periods, business context | +| Pain Point | Data cleaning, metric definitions, chart selection and anomaly attribution all pile together, very time-pressured | +| DesireCore Intervention | Data Analyst first does data overview and definition confirmation, then generates charts, comparisons and preliminary explanations by question | +| Acceptance Result | User gets metric tables, charts and conclusion hypotheses that support discussion, then decides which ones go into the report | ## Recommended Agent -**Data Analyst** -- Specializes in data cleaning, statistical analysis, trend prediction, and visual report generation, supporting CSV, Excel, JSON, and other data formats. +**Data Analyst** — Suitable for data cleaning, statistical analysis, trend comparison, anomaly detection and visual report generation; supports CSV, Excel, JSON and other common data formats. -## Teaching First: Setting Your Business Metrics +## Teaching First: Master Your Business Metrics -> Every company calculates GMV, conversion rates, and other metrics differently. Teach your definitions once, and every future analysis will use the correct standards -- no more "the numbers don't match" moments. +> Every company calculates GMV, conversion rates and other metrics differently. Teach your definitions once, and every future analysis will use the correct standards—no more "the numbers don't match" moments. ### Teaching Rules @@ -105,18 +111,20 @@ Data Analyst: West China deep dive analysis results: ## Key Steps -1. **Upload Data** -- Upload CSV, Excel, or JSON files to the agent -2. **Describe Analysis Needs** -- Explain what you want to understand (trends, comparisons, predictions, etc.) -3. **Get Analysis Results** -- The agent automatically generates statistical summaries, charts, and insights -4. **Deep Dive** -- Ask follow-up questions about discovered issues for deeper analysis +1. **Define Business Question** — First clarify what question needs answering, e.g., "which region dragged down sales targets" rather than just "analyze data" +2. **Confirm Metric Definitions** — Specify time range, metric definitions, refund/cancellation handling and target value sources +3. **Data Health Check** — First check column names, missing values, duplicates, outliers and category naming consistency +4. **Explore and Compare** — Compare by time, region, channel, product, demographic and other dimensions to find major differences +5. **Form Verifiable Conclusions** — Write conclusions as "phenomenon + data evidence + possible causes + next verification steps", avoid just giving subjective judgments ## Final Deliverable -A complete analysis report with key metrics, regional comparisons, trend charts, anomaly detection, and improvement recommendations, ready for team presentations. +An analysis report containing key metrics, data quality notes, comparison dimensions, trend charts, anomaly explanations and next-step verification suggestions. :::tip Best Practices - Ensure clear column names before uploading (e.g., "Sales Amount" instead of "col_1") so the agent can better understand data meaning - The more specific your analysis requirements, the more precise the results. "Analyze sales trends" is less effective than "Compare Q3 and Q4 growth rates by region, identifying the fastest and slowest growing regions" - If the data volume is large, have the agent do a data overview first, then dive deeper into directions of interest -- **Batch analyze multi-department data**: When you need to review data across departments or product categories on Monday morning, send everything to the Data Analyst at once -- it will analyze each dataset in parallel, all using the same metric standards and report template you taught -::: +- When batch analyzing multi-department data, first confirm all spreadsheets use the same metric definitions and time range, then have the agent process in parallel +- Don't skip data health checks. Many "business anomalies" actually come from empty values, duplicate orders, definition changes or export scope inconsistencies +::: \ No newline at end of file diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/03-email-management.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/03-email-management.md index 61004b4..5b970bf 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/03-email-management.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/03-email-management.md @@ -1,16 +1,22 @@ --- title: Email Management -description: Use the AI Assistant to intelligently classify emails, draft replies, and extract key information for efficient daily email processing. +description: Use the email assistant to classify emails, extract key information, organize to-dos and draft replies. keywords: [Email Management, Email Classification, Email Replies, Efficiency, AI Assistant] --- # Email Management -## Scenario Description +## Real Workflow -You open your inbox on Monday morning to find over 50 unread emails. There are clients催促 delivery, colleagues asking about progress, administrative notices, and a pile of promotional emails. Reading and replying to each one would take at least two hours. +This scenario isn't about "letting AI send emails for you"—it's about transforming your inbox from a pile of unread messages into a manageable work queue. -The AI Assistant can help you process these emails quickly: classifying by importance, extracting key information from each email, and drafting reply content, allowing you to focus on what truly requires your personal attention. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | After vacation, Monday morning, before/after project launch—emails pile up | +| Existing Materials | Email body, sender identity, historical communication context, attachments or contract context | +| Pain Point | Important emails mixed with notifications; to-dos, deadlines and responsible people unclear | +| DesireCore Intervention | First classify and sort, then extract summaries, to-dos and suggested replies; sensitive emails can have multiple reply versions generated | +| Acceptance Result | User sees priority list and reply drafts, handles client, deadline and contract emails first | ## Complete Conversation Example @@ -68,17 +74,25 @@ AI Assistant: Revised, please confirm: ## Key Steps -1. **Paste Email Content** -- Provide the email body to the agent -2. **Smart Classification** -- The agent automatically identifies email importance and type -3. **Extract Key Information** -- Summarize email highlights and flag items requiring your attention -4. **Draft Reply** -- Generate reply drafts based on email content and your needs +1. **Merge Context** — First merge emails from the same client or project into threads, avoid only seeing the last email +2. **First-Round Triage** — Classify into "handle immediately / handle today / delegate / wait for reply / archive for reference" +3. **Extract Commitments** — For each action email, extract task, responsible person, deadline, source email and dependencies +4. **Generate Reply Drafts** — Create drafts with different tones for clients, internal colleagues, vendors etc. +5. **Build Follow-Up Queue** — Separately list emails waiting for client reply, internal confirmation or needing reminders ## Final Deliverable -Each email has clear classification labels, key information summaries, and ready-to-use reply drafts, cutting email processing time by more than half. +Each email gets classification labels, key information summary, to-dos, reply drafts and follow-up status. You can handle urgent client matters first, then batch process internal collaboration and notification emails. + +## DesireCore Capabilities Used + +- **Email Center**: Gmail, Outlook, IMAP unified management; view context by conversation thread +- **Natural Language Teaching**: Teach the agent your client reply style, escalation rules and prohibited expressions +- **Security Confirmation**: Client commitments, contracts, complaints and other important emails retain manual confirmation before sending :::tip Best Practices -- When processing in bulk, you can paste multiple emails at once, and the agent will sort them by importance -- Tell the agent your reply habits (e.g., "I usually use a formal tone"), and it will remember and apply them to subsequent replies +- When processing in bulk, you can paste multiple emails at once—the agent will sort them by importance +- Tell the agent your reply habits (e.g., "more formal for client emails, more concise for internal emails"), and subsequent drafts will better match your style - For emails requiring careful wording (e.g., customer complaints, contract negotiations), have the agent provide multiple versions for you to choose from -::: +- Important to-dos should retain source email links or original text snippets, so you won't lose context when tracing back for accountability +::: \ No newline at end of file diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/04-meeting-assistant.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/04-meeting-assistant.md index f564967..2a4bd3f 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/04-meeting-assistant.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/04-meeting-assistant.md @@ -1,16 +1,22 @@ --- title: Meeting Assistant -description: Use the AI Assistant to complete agenda preparation before meetings and minutes organization, action item extraction after meetings. +description: Use the meeting assistant to prepare agendas, organize minutes, extract to-dos and follow-up checklists. keywords: [Meeting Assistant, Meeting Minutes, Agenda, Action Items, Efficiency] --- # Meeting Assistant -## Scenario Description +## Real Workflow -You have an important project coordination meeting this afternoon and need to prepare an agenda. After the meeting, you also need to organize meeting minutes, extract everyone's action items, and send them to all attendees. These administrative tasks, while not difficult, are very time-consuming. +The real value of a meeting assistant isn't expanding a few sentences into minutes—it's completing the information loop before and after the meeting. -The AI Assistant can help you efficiently complete the before and after of meetings: generating structured agendas before the meeting, and organizing formal minutes from your notes or transcription after the meeting, automatically extracting action items assigned to specific responsible parties. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | Launch coordination meeting, requirement review, project retrospective, client meeting—any meeting needing pre-meeting preparation and post-meeting tracking | +| Existing Materials | Meeting objectives, attendee roles, topics, handwritten notes, audio transcription, decisions made during meeting | +| Pain Point | Pre-meeting topics not organized; post-meeting decisions and to-dos mixed together; responsible persons and deadlines easily missed | +| DesireCore Intervention | Pre-meeting: generate agenda and items to be decided; post-meeting: organize minutes, to-do list and follow-up checklist | +| Acceptance Result | Host gets a minutes document ready to send to attendees, plus a task list that can be synced to project management tools | ## Complete Conversation Example @@ -71,16 +77,19 @@ AI Assistant: Meeting minutes have been organized: ## Key Steps -1. **Meeting Agenda Generation** -- Provide attendees and discussion topics, the agent generates a structured agenda -2. **Meeting Notes Organization** -- After the meeting, provide notes or transcription, the agent organizes into formal minutes -3. **Action Item Extraction** -- Automatically extract action items from minutes, assigning responsible persons and deadlines +1. **Pre-Meeting Goal Alignment** — Clarify what the meeting should produce: decisions, proposal review, risk sync or task assignment +2. **Carry Previous Action Items** — Include last meeting's action items, open issues and new topics in the agenda +3. **Distinguish Facts from Decisions** — In notes, separate discussion content, confirmed decisions, unconfirmed questions and risks +4. **Structure Action Items** — Each to-do needs task, sole responsible person, deadline, completion criteria and dependencies +5. **Post-Meeting Follow-Up Queue** — Minutes aren't just for sending; sync action items to project tools or next meeting agenda ## Final Deliverable -A formal meeting minutes document ready to be sent, including decisions, action items, and next follow-up arrangements. +A clearly structured meeting minutes draft containing decisions, to-do checklist, responsible persons, deadlines, completion criteria and next follow-up arrangements. Can be further converted to email format, Feishu announcement or project task list. :::tip Best Practices -- Meeting notes don't need to record every word, just key decisions and assignments, the agent will help organize into complete minutes -- If you have meeting recording transcription, paste it directly to the agent for better results -- You can have the agent remember your preferred minutes format (e.g., "always list action items in a table"), so you don't need to repeat instructions next time -::: +- Meeting notes don't need to record every word, but key decisions, responsible persons and deadlines should be as clear as possible +- If you have meeting audio transcription, paste it directly to the agent for better results +- Minutes format can be remembered by the agent (e.g., "always list to-dos in a table"), so no need to repeat instructions next time +- Meeting action items without a sole responsible person and completion criteria easily become "everyone knows but nobody pushes forward" items +::: \ No newline at end of file diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/05-translation.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/05-translation.md index d2d237a..3fc873b 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/05-translation.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/05-translation.md @@ -1,24 +1,30 @@ --- title: Translation Assistant -description: Use the Translator agent to translate technical documents, business emails, etc., preserving original style and professional terminology. +description: Use the translation assistant to handle technical documents, business emails and localized text with unified terminology and style. keywords: [Translation, Multilingual, Localization, Glossary, Translator] --- # Translation Assistant -## Scenario Description +## Real Workflow -You need to translate a technical document into English for the overseas team, but regular translation tools produce "Chinglish" results with inconsistent professional terminology. Or you received an email in Japanese and need to quickly understand the content and reply. +This scenario isn't about single-sentence translation—it's about delivery-oriented translation for real readers, focusing on terminology, tone and contextual consistency. -The Translator agent supports mutual translation among 12 mainstream languages, understands contextual nuances, preserves original style, and supports custom glossaries to ensure consistent professional vocabulary. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | Sending API documentation to overseas clients, replying to foreign-language emails, preparing product localization content | +| Existing Materials | Source text, target language, reader identity, product terminology glossary, historical translation style | +| Pain Point | Inconsistent terminology, stiff literal translations, code/API names/legal/business wording prone to errors | +| DesireCore Intervention | Translator generates translation based on glossary, lists key terminology, style choices and ambiguous sentences | +| Acceptance Result | User receives reviewable translation, focusing on professional terminology, client-sensitive expressions and contextual consistency | ## Recommended Agent -**Translator** -- Supports mutual translation among 12 languages including Chinese, English, Japanese, Korean, French, German, Spanish, and Russian, excelling in professional translation of business documents, technical materials, and legal files. +**Translator** — Supports mutual translation among Chinese, English, Japanese, Korean, French, German, Spanish, Russian and other mainstream languages; suitable for business documents, technical materials, product localization and cross-language email communication. -## Teaching First: Setting Your Terminology and Style +## Teaching First: Master Your Terminology and Style -> The biggest challenge in professional translation is inconsistency -- the same term translated differently across documents. Take a few minutes to upload your glossary once, and every translation after that will follow it precisely. +> The biggest challenge in long-term translation projects is terminology inconsistency. The same product name, feature name or industry term should maintain the same translation across documents. ### Teaching Rules @@ -44,7 +50,7 @@ Translator: 📝 New learning recorded You: [✓ Correct] By the way, I'll send you updated glossaries whenever they change. ``` -Once teaching is complete, the Translator has learned your glossary and style preferences. From now on, just send documents for translation -- terminology will be applied automatically. +Once teaching is complete, the Translator has learned your glossary and style preferences. From now on, just send documents for translation—terminology will be applied automatically. ## Complete Conversation Example @@ -101,19 +107,21 @@ Translator: Text type identified: Business Email (Formal) ## Key Steps -1. **Upload or Paste Document** -- Provide the text content to be translated -2. **Set Translation Preferences** -- Specify target language, style (formal/neutral/casual), and glossary -3. **Get Translation Result** -- The agent returns translated text and terminology glossary -4. **Review and Adjust** -- Provide revision suggestions for the translation, the agent makes precise adjustments +1. **Define Reader and Purpose** — The same Chinese text will be translated differently for developers, clients, legal or marketing +2. **Prepare Localization Package** — Provide glossary, product names, variable/placeholder rules, style guide and historical translations +3. **Translate by Context** — Technical docs preserve code and API names, marketing copy preserves tone, UI text considers length and button context +4. **Terminology and Format QA** — Check terminology consistency, number units, links, variables, punctuation and untranslatable content +5. **Consolidate Translation Memory** — Store confirmed terminology and expressions for consistent batch translation later ## Final Deliverable -An idiomatic, professional translated document with a terminology glossary to ensure consistency in subsequent translations. +A reviewable translation, usually accompanied by terminology mapping, style notes, untranslatable items and sentences that could be ambiguous in the source text. :::tip Best Practices -- Before translation, tell the agent the document's purpose and target audience (e.g., "API documentation for American developers"), the translation style will be more appropriate +- Before translation, tell the agent the document's purpose and target audience (e.g., "API documentation for American developers"), the translation style will be more stable - If you have internal company glossaries, provide them upfront, and the agent will strictly follow them -- For long documents, suggest translating in batches and confirming terminology consistency between batches -- For legal documents, contracts, and other critical translations, the results should be reviewed by legal professionals -- **Batch translation**: When you have multiple documents to translate, send them all at once -- the Translator will process them in parallel with globally consistent terminology, ensuring the same term is never translated differently across documents -::: +- For long documents, translate in batches and confirm terminology consistency between each batch +- Legal documents, contracts and other critical translations should be reviewed by legal professionals for final approval +- Before batch translation, confirm unified glossary and style requirements, then send multiple documents to the translator for processing +- For product interface and help center translations, check variables, button lengths, links and screenshot context—don't just look at the sentences themselves +::: \ No newline at end of file diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/06-data-analysis-agent.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/06-data-analysis-agent.md index 8347b60..cb0359d 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/06-data-analysis-agent.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/06-data-analysis-agent.md @@ -1,16 +1,22 @@ --- title: Data Analysis Report -description: Use an AI agent to automatically generate standardized data analysis reports, supporting multi-source data ingestion, intelligent analysis and visualization, and formatted output. +description: Use the data analysis report agent to organize data, templates and analysis instructions into standard reports. keywords: [Data Analysis, Report Generation, Visualization, Excel, Financial Report] --- # AI-Powered Standard Data Analysis Reports -## Pain Points +## Real Workflow -Every week, month, and quarter, business teams need to submit various data analysis reports: sales weekly reports, operations monthly reports, financial quarterly reports. Reports have strict format requirements, chart styles must be uniform, and ensure consistency in data definitions. Analysts spend a lot of time on "template filling": exporting data from Excel, adjusting formats, creating charts, writing conclusions — a single report takes at least half a day. +This scenario suits periodic reports. The real pain point isn't "can't analyze"—it's having to repeat data extraction, template filling, chart creation and writing the same structure of conclusions every time. -This use case lets an AI agent automatically generate standardized enterprise data analysis reports. Input raw data and report templates, and get professionally formatted, visually rich reports. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | Sales weekly reports, operations monthly reports, financial quarterly reports, departmental business analysis—fixed-cycle reports | +| Existing Materials | Current period data, previous report, company template, metric definitions, chart style requirements | +| Pain Point | Multi-department data formats inconsistent, report template fixed but filling tedious, anomaly explanations easily missed | +| DesireCore Intervention | Data analysis report agent generates department reports and summary version according to template, and highlights abnormal metrics | +| Acceptance Result | Analyst focuses on checking anomaly explanations and management suggestions, no longer spending time on copy-paste and formatting | --- @@ -44,6 +50,27 @@ This use case lets an AI agent automatically generate standardized enterprise da --- +## Workflow Control Points + +| Stage | Details to Confirm | +|-------|-------------------| +| Data Ingestion | Whether each department's data range, time scope, field names and units are consistent | +| Data Cleaning | Whether there are empty values, duplicate rows, outliers, merged cells and inconsistent category names | +| Metric Calculation | Whether MoM, YoY, proportion, average order value etc. use company unified definitions | +| Chart Generation | Whether charts serve conclusions, not just "looking good" by stacking charts | +| Text Conclusions | Whether each conclusion can be traced back to specific data, charts or business explanations | +| Template Output | Whether title, table of contents, headers/footers, chart styles and export format meet company standards | + +--- + +## DesireCore Capabilities Used + +- **Workflow / SOP**: Solidify monthly, weekly and quarterly report processes into fixed steps, reducing repeated instructions +- **Multi-Agent Collaboration**: Data Analyst handles calculations and charts, AI Copywriter handles report text and layout +- **Scheduled Tasks**: Can automatically generate periodic reports or remind you to supplement data sources at fixed times + +--- + ## Typical Use Cases ### Scenario 1: Consumer Industry Sales Data Analysis Report @@ -134,9 +161,13 @@ File location: ./assets/data-analysis/finance_q1_report | Metric | Manual Report Creation | Fixed Script Generation | AI Agent | |--------|------------------------|-------------------------|----------| -| Single report time | ~3 hours | ~10 minutes | ~3 minutes | -| Batch generation (10 reports) | ~30 hours | ~20 minutes | ~15 minutes | -| Template adaptation cost | Manual each time | Code modification required | Natural language description | -| Anomaly analysis capability | Relies on human experience | Requires preset rules | Intelligent identification | -| Conclusion writing | Manual | None | Auto-generated | +| Single report time | Usually takes hours | Fast after development complete | Suitable for generating reviewable first drafts | +| Batch generation (10 reports) | Easily filled with formatting and copy-paste | Suitable for fixed formats | Suitable for batch tasks with same template, different data sources | +| Template adaptation cost | Manual adjustment each time | Code modification required | Natural language description | +| Anomaly analysis capability | Relies on human experience | Requires preset rules | Assists identification | +| Conclusion writing | Manual | None | Generates first draft | | Format consistency | Error-prone | High | High | + +:::tip Usage Suggestion +These periodic reports are best suited for solidifying templates and metric definitions first. Once the template is stable, each month mainly checks anomaly explanations and management suggestions—no need to repeatedly adjust formatting. +::: \ No newline at end of file diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/07-flowchart-agent.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/07-flowchart-agent.md index 243b683..6d81288 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/07-flowchart-agent.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/07-flowchart-agent.md @@ -1,16 +1,22 @@ --- title: Flowchart Generation -description: Use an AI agent to automatically generate professional flowcharts, architecture diagrams, and sequence charts from natural language descriptions or documents, with multi-format export support. +description: Generate flowcharts, architecture diagrams, sequence charts and other editable charts from natural language descriptions, documents or code. keywords: [Flowchart, Architecture Diagram, Sequence Diagram, Visualization, Chart Generation] --- # AI-Powered Automatic Flowchart Generation -## Pain Points +## Real Workflow -Product managers need flowcharts for PRDs, developers need architecture diagrams for technical docs, and operations teams need business process diagrams for SOPs — every time you have to open Visio, ProcessOn, or draw.io, dragging nodes one by one, connecting lines, aligning, and adjusting styles. For a moderately complex flowchart, drawing and beautifying takes at least half an hour. +The flowchart scenario works because: logic already exists—it's just hidden in text, meeting discussions or code, causing different roles to have inconsistent understanding. -This use case lets an AI agent automatically generate professional flowcharts based on natural language descriptions or document content, supporting multiple export formats, eliminating tedious manual work. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | PRD review, SOP creation, technical proposal review, API integration alignment | +| Existing Materials | Verbal processes, PRD paragraphs, SOP documents, API specifications, code call chains | +| Pain Point | Branches, exception paths and role boundaries unclear; manual drawing consumes lots of time | +| DesireCore Intervention | First extract process structure, then generate flowcharts, swimlane diagrams, sequence diagrams or architecture diagrams, and adjust based on feedback | +| Acceptance Result | Team gets a diagram usable for review and discussion, focusing on logic review rather than dragging nodes | --- @@ -44,6 +50,19 @@ This use case lets an AI agent automatically generate professional flowcharts ba --- +## Workflow Control Points + +| Stage | Details to Confirm | +|-------|-------------------| +| Input Clarification | First confirm process boundaries: where it starts, where it ends, who participates | +| Node Extraction | Separate actions, decisions, waits and exception handling into different nodes, don't mix into one sentence | +| Role Separation | For cross-department processes, prefer swimlane diagrams, clarify which role owns each node | +| Branch Review | Focus on checking "No", "Exception", "Timeout", "Rejection" paths are complete | +| Style Output | Only adjust layout, colors and export format after logic is confirmed | +| Future Maintenance | Save editable formats together, avoid having to redraw next time | + +--- + ## Typical Use Cases ### Scenario 1: Product Requirements Flowchart @@ -144,9 +163,9 @@ File location: ./assets/flowchart/microservice-architecture.drawio | Metric | Manual Drawing (Visio/draw.io) | AI Agent | |--------|--------------------------------|----------| -| Simple flowchart (5-10 nodes) | ~15 minutes | ~10 seconds | -| Complex flowchart (20+ nodes) | ~45 minutes | ~30 seconds | +| Simple flowchart (5-10 nodes) | Manual drag, connect, align | Suitable for quick initial generation | +| Complex flowchart (20+ nodes) | High modification and alignment cost | Suitable for confirming structure first, then multi-round adjustment | | Modification & Adjustment | Manual drag | Natural language description | -| Style Uniformity | Manual setup required | Auto-applied templates | +| Style Uniformity | Manual setup required | Can apply templates | | Format Conversion | Export one by one | One-click multi-format | -| Learning Curve | Need to learn tool | Zero barrier | +| Learning Curve | Need to learn drawing tool | Mainly through description and review \ No newline at end of file diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/08-web-scraping-agent.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/08-web-scraping-agent.md index 1078c87..9083224 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/08-web-scraping-agent.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/01-general/08-web-scraping-agent.md @@ -1,49 +1,43 @@ --- title: Web Data Collection -description: Use an AI agent to automatically browse web pages, extract information, and organize data, outputting structured results with scheduled tasks and change monitoring support. +description: Use an AI agent to browse public web pages, extract fields, and organize data into structured results. keywords: [Data Collection, Web Scraping, Data Extraction, Competitor Monitoring, Structured Data] --- # AI-Powered Web Data Collection -## Pain Points +## Real Workflow -Market research requires monitoring competitor prices, operations teams need to collect industry news, and sales teams need to organize publicly available client information — every day spent switching between different websites, copying and pasting, and formatting. Traditional web scrapers require writing code and maintaining scripts; a single website redesign breaks them. Manual collection is inefficient and prone to missing key information. +Web data collection scenarios are typically not long-running, stable large-scale scrapers, but rather temporary, low-frequency information gathering tasks where fields frequently change. -This use case makes an AI agent your data collection assistant. Just tell it "where to collect what," and it automatically completes web browsing, information extraction, and data organization, outputting structured results. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | Competitor research, job posting aggregation, public company information collection, industry news gathering | +| Existing Materials | Target URLs, field requirements, filtering criteria, historical spreadsheets or output templates | +| Pain Point | Website structures are inconsistent, manual copying easily misses fields, developing traditional scrapers isn't cost-effective | +| DesireCore Intervention | Web Data Collection agent browses pages, extracts fields, preserves source links and outputs tables | +| Acceptance Result | User receives a verifiable data table, then confirms collection scope and data usage per compliance requirements | ---- - -## What It Can Do - -### 🌐 Intelligent Web Browsing - -- **Multi-Page Parallel**: Open multiple web pages simultaneously for batch collection -- **Dynamic Content Handling**: Supports JavaScript-rendered pages, waits for content loading -- **Login State Preservation**: Supports Cookie management for content requiring login -- **Anti-Scraping Countermeasures**: Intelligent access frequency control, simulating human browsing behavior - -### 🎯 Precise Information Extraction +## Recommended Agent -- **Natural Language Instructions**: "Extract product names, prices, and ratings from this page" -- **Table Data Recognition**: Automatically identifies web tables and fully extracts row and column data -- **List Content Collection**: News lists, product lists, search results — all batch collected -- **Nested Data Processing**: Detail page links automatically followed for complete information collection +**Web Data Collection Agent** — Handles competitor pricing monitoring, public company information gathering, industry news aggregation, and other web browsing and data extraction tasks. Uses natural language instructions to specify collection targets and output formats. -### 📋 Structured Output +## Work Flow Control Points -- **Excel / CSV Export**: Standard table format for subsequent analysis -- **JSON Format**: Integration with databases or other systems -- **Custom Templates**: Define output fields and formats on demand -- **Incremental Updates**: Compare with historical data, output only changes +| Stage | Details to Confirm | +|-------|-------------------| +| Scope Confirmation | Target website, sections, page count, fields, and filtering criteria are clearly defined | +| Compliance Check | Whether collection is allowed; whether login state, personal information, copyright content, or prohibited terms are involved | +| Extraction Rules | Field names, price units, date formats, handling of missing values, and whether to preserve source links | +| Data Cleaning | Deduplication, standardization, abnormal prices, empty fields, and duplicate products are handled | +| Output Validation | Spot-check several records to confirm no field misalignment or omissions | +| Ongoing Maintenance | Re-calibrate rules when page structures change; preserve collection logs for scheduled tasks | -### 🔄 Scheduled Tasks +## DesireCore Capabilities Used -- **Periodic Collection**: Execute automatically daily or weekly -- **Change Monitoring**: Real-time alerts for price changes and content updates -- **History Records**: Retain collection history for trend analysis - ---- +- **Desktop Automation / GUI Control**: Browse pages through the browser, perform screenshot recognition and form interactions when needed +- **Scheduled Inspection**: Periodically check for changes in prices, announcements, job postings, etc.; notify only when changes occur +- **Security Audit**: Preserve collection sources, execution records, and output files for traceability ## Typical Use Cases @@ -57,7 +51,7 @@ File location: ./assets/web-scraping/gucci_handbags.xlsx ├── Collection scope: All products in handbags category └── User instruction: "Collect names, prices, and links for all handbags on Gucci official site" -⬇️ Agent processing (approx. 3-5 minutes) +⬇️ Agent processing 📊 Output: gucci_handbags.xlsx (33 products) ├── Product Name @@ -80,7 +74,7 @@ File location: ./assets/web-scraping/gucci_handbags.xlsx ├── Keyword filtering rules └── User instruction: "Collect today's AI industry news, sorted by importance" -⬇️ Agent processing (approx. 5-8 minutes) +⬇️ Agent processing 📊 Output ├── Today_News_Summary.md @@ -98,7 +92,7 @@ File location: ./assets/web-scraping/gucci_handbags.xlsx ├── Target company list (50 companies) └── User instruction: "Collect basic information, funding status, and main products for these companies" -⬇️ Agent processing (approx. 15-20 minutes) +⬇️ Agent processing 📊 Output ├── Company_Information_Database.xlsx @@ -118,7 +112,7 @@ File location: ./assets/web-scraping/gucci_handbags.xlsx ├── Job keywords, city, salary range └── User instruction: "Find qualified Product Manager positions" -⬇️ Agent processing (approx. 8-10 minutes) +⬇️ Agent processing 📊 Output ├── Job_Listing.xlsx @@ -128,25 +122,32 @@ File location: ./assets/web-scraping/gucci_handbags.xlsx └── Salary distribution analysis chart ``` ---- +## Key Steps + +1. **Define Collection Scope** — Specify target website, page sections, number of pages, fields to extract, and filtering criteria; the more specific, the more accurate the results +2. **Compliance Review** — Before collection begins, confirm that the target website's terms of use and robots.txt permit automated access; for data involving personal information or copyright, verify authorization first +3. **Set Extraction Rules** — Define field names, units, date formats, and how to handle missing values; request source link preservation for later verification +4. **Data Cleaning and Validation** — After collection, deduplicate, standardize formats, handle outliers and empty fields; spot-check several records to confirm field accuracy +5. **Automated Maintenance** — For periodic collection needs, set up scheduled tasks and preserve collection logs; re-calibrate extraction rules when page structure changes ## Efficiency Comparison | Metric | Manual Collection | Traditional Web Scraper | AI Agent | |--------|-------------------|-------------------------|----------| -| Collect 100 data points | ~2 hours | ~5 minutes (after development) | ~10 minutes | +| Collect 100 data points | Time-consuming, easily misses items | Fast after development complete | Suited for temporary or frequently-changing collection tasks | | Technical Barrier | None | High (requires programming) | Low (natural language) | -| Website Adaptation Cost | None | High (code per site) | Low (auto-adaptation) | -| Maintenance Cost | Continuous labor | High (updates needed on redesign) | Low (intelligent handling) | +| Website Adaptation Cost | None | High (code per site) | Adjustable through field descriptions | +| Maintenance Cost | Continuous labor | High (site redesign requires updates) | Requires re-calibration per page changes | | Unstructured Content | Handleable | Difficult | Strong | -| Compliance | Human judgment | Requires configuration | Built-in frequency control | +| Compliance | Human judgment | Requires configuration | Still requires user confirmation of site terms and data usage | ---- +## Final Deliverable -## Usage Notes +A structured data table with traceable source links, covering all fields specified in your requirements. You can continue asking the agent to filter, analyze, visualize, or export the data in other formats. -⚠️ **Compliance Reminder**: -- Please comply with target website's robots.txt and terms of use +:::warning Compliance Reminder +- Comply with the target website's robots.txt and terms of use - Control collection frequency to avoid burdening target websites - Only collect publicly accessible information -- Collected data is for internal analysis use only; ensure data privacy compliance +- For data involving personal information, account content, or copyright-protected content, confirm authorization and data usage scope first +::: diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/01-legal-advisor.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/01-legal-advisor.md index a6f9954..550414e 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/01-legal-advisor.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/01-legal-advisor.md @@ -1,24 +1,30 @@ --- title: Legal Advisor -description: Use the Legal Advisor agent for contract review, regulation retrieval, and legal risk assessment. +description: Use the Legal Advisor agent to assist with contract pre-review, regulation retrieval, and legal risk assessment. keywords: [Legal Advisor, Contract Review, Regulation Retrieval, Risk Assessment, Compliance] --- # Legal Advisor -## Scenario Description +## Real Workflow -You received a supplier contract that needs to be signed tomorrow, but your legal colleague is on a business trip. The contract is filled with dense clauses, and you're worried about unfavorable "traps" for the company. +The core of the legal scenario is "pre-review" — the agent doesn't replace legal counsel to make final decisions. It's suited for organizing contract issues before formal legal review. -The Legal Advisor agent can quickly review contracts, analyze the reasonableness of key clauses item by item, flag risk points, and provide modification suggestions. It's like having an experienced legal assistant by your side at all times. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | Procurement contracts, outsourcing agreements, NDAs, and client agreements entering the approval process | +| Existing Materials | Contract text, company review standards, historical review reports, business background, and contract value information | +| Pain Point | Too many clauses, business owners unsure which points need legal counsel involvement, high communication costs with legal | +| DesireCore Intervention | Legal Advisor flags risks per company standards, suggests revisions, identifies missing background info, and marks items needing legal confirmation | +| Acceptance Result | Business owner receives a pre-review checklist, then discusses key clauses with legal counsel in a focused session | ## Recommended Agent -**Legal Advisor** -- Specializes in legal compliance, with capabilities in contract review, regulation retrieval, risk assessment, and compliance recommendations. +**Legal Advisor** — Suited for contract clause review, regulation research, compliance checklist preparation, and risk flagging. -## Teaching First: Setting Your Review Standards +## Teaching First: Teaching the Agent Your Review Standards -> Spend 10 minutes teaching once, and the agent will automatically apply these standards every time it reviews a contract — no need to repeat yourself. +> It's recommended to first teach the agent your company's contract review standards, red-line clauses, and report format. It will then apply the same rules consistently for every pre-review. ### Teach Rules @@ -72,7 +78,7 @@ Once teaching is complete, the Legal Advisor has learned your company's contract You: Please help me review this supplier contract. It needs to be signed tomorrow. [Paste contract content] -Legal Advisor: Let me help you quickly review this contract. Analyzing clauses... +Legal Advisor: Let me do a contract pre-review. Analyzing clauses... Review Summary @@ -120,22 +126,24 @@ Legal Advisor: Generated modification suggestion draft. ## Key Steps -1. **Upload or Paste Contract** -- Provide the contract text to be reviewed -2. **Get Review Report** -- The agent analyzes item by item, flagging risk levels -3. **View Modification Suggestions** -- Specific modification text provided for each risk point -4. **Generate Modified Draft** -- Optionally generate a contract version with all modification suggestions +1. **Confirm Contract Background** — First clarify contract type, value, counterparty, business purpose, and clauses you're most concerned about +2. **Apply Company Red Lines** — Check payment, breach, duration, IP, confidentiality, and dispute resolution clauses against your company standards +3. **Generate Risk Matrix** — For each issue, annotate clause location, risk level, business impact, suggested revision, and whether legal counsel confirmation is needed +4. **Output Negotiation Text** — Generate ready-to-use replacement language for high-risk clauses +5. **Prepare Legal Consultation List** — Separately list issues that AI cannot judge and require additional business or legal background ## Final Deliverable -A detailed contract review report, flagging all risk points and corresponding modification suggestions, so you know what you're getting into before signing. +A contract pre-review report containing risk points, risk levels, suggested revision text, and a checklist of items requiring legal counsel confirmation. :::warning Disclaimer -The Legal Advisor agent's suggestions are for reference only and do not constitute formal legal advice. For major legal decisions (such as large contract signing, litigation response, equity transactions, etc.), consult a licensed attorney. +The Legal Advisor agent provides supplementary analysis and does not constitute formal legal advice. For major contracts, litigation, equity transactions, labor disputes, and similar matters, final review should be conducted by a licensed attorney or in-house legal counsel. ::: :::tip Best Practices - When uploading contracts, also specify which clauses you're most concerned about (e.g., payment method, liability for breach), and the agent will focus on analyzing these sections -- If it involves specific industries (such as construction, IT outsourcing), tell the agent the industry background, and it will provide more precise suggestions based on industry practices -- You can have the Legal Advisor collaborate with the Finance Assistant, one reviewing clauses and the other calculating costs, for a comprehensive contract evaluation -- **Batch contract review**: When you receive multiple contracts at once, send them all to the Legal Advisor — it will review each one in parallel, applying the same set of standards you taught it +- If it involves specific industries (such as construction, IT outsourcing), tell the agent the industry background and applicable region so it can organize risks based on more relevant rules +- You can have the Legal Advisor collaborate with the Finance Assistant — one reviewing clauses and the other calculating costs — for a comprehensive contract evaluation +- When batch-reviewing contracts, first unify the review standards, then have the Legal Advisor output risk lists against the same criteria +- The most important thing in contract pre-review is not "pointing out many issues" but distinguishing which must be changed, which can be negotiated, and which only need business awareness ::: diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/02-code-assistant.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/02-code-assistant.md index e923780..a00e453 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/02-code-assistant.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/02-code-assistant.md @@ -1,24 +1,30 @@ --- title: Code Assistant -description: Use the Code Assistant agent to build complete projects through conversation, from architecture design to code implementation, testing, and deployment. +description: Use the Code Assistant agent for requirements clarification, architecture design, code implementation, testing, and iterative modifications. keywords: [Code Assistant, Project Development, Full-Stack Development, Code Generation, Architecture Design] --- # Code Assistant -## Scenario Description +## Real Workflow -You have a new project to build — maybe an internal tool, an API service, or a mini-program. The traditional way requires writing every line from scratch, and if you're unfamiliar with the tech stack, you're constantly switching between docs and code. +The focus of coding scenarios is not "describe one sentence and let AI build the entire project" — it's about translating clear requirements into changes that fit the existing codebase's structure, standards, and tests. -The Code Assistant agent can help you build entire projects through conversation: describe your requirements, and it designs the architecture; confirm the plan, and it writes the code; report issues, and it iterates. It's like pair programming with a full-stack engineer. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | New business modules, bug fixes, test coverage, migration scripts, internal tool development | +| Existing Materials | Requirements spec, existing code, team standards, API contracts, database schema, test commands | +| Pain Point | Code must follow existing architecture, edge cases need tests, changes must not break existing logic | +| DesireCore Intervention | Code Assistant reads context and standards first, then proposes design, writes code, adds tests, and runs verification | +| Acceptance Result | Engineer receives reviewable diffs, test results, and design decisions that need human confirmation | ## Recommended Agent -**Code Assistant** -- Proficient in Python, TypeScript, Go, Rust, and other mainstream languages, capable of handling the full development lifecycle from architecture design to code implementation, testing, and deployment. +**Code Assistant** — Suited for architecture design, code implementation, refactoring, test coverage, debugging, and deployment scripts. -## Teaching First: Setting Your Tech Stack and Standards +## Teaching First: Teaching the Agent Your Tech Stack and Standards -> Teach your project conventions once, and every feature it builds will follow your tech stack and team practices — no matter what you're developing. +> Teach your tech stack, directory structure, code style, and test requirements first — the generated code will then be much closer to team standards. ### Teach Rules @@ -34,17 +40,17 @@ You: Let me teach you our team's tech stack and development standards: 5. All functions must have type annotations — Python type hints, frontend TypeScript strict mode (preference) Code Assistant: 📝 New learning recorded - Source: team-dev-guide.md + supplementary notes - Type: Rule (Tech stack and development standards) - Level: Mandatory × 4, Preference × 1 + Source: team-dev-guide.md + supplementary notes + Type: Rule (Tech stack and development standards) + Level: Mandatory × 4, Preference × 1 - My understanding: - Backend: FastAPI + PostgreSQL + SQLAlchemy. Frontend: React + TS + - Tailwind. Project organized by feature modules (modules/xxx/), each - containing router, service, model. Unified API response format - {"code", "data", "message"}. Full type annotations throughout. + My understanding: + Backend: FastAPI + PostgreSQL + SQLAlchemy. Frontend: React + TS + + Tailwind. Project organized by feature modules (modules/xxx/), each + containing router, service, model. Unified API response format + {"code", "data", "message"}. Full type annotations throughout. - [✓ Correct] [✎ Modify] [✗ Undo] + [✓ Correct] [✎ Modify] [✗ Undo] You: [✓ Correct] ``` @@ -56,17 +62,17 @@ You: Here's our existing user module code structure. Follow this pattern for new [Paste modules/user/ directory: router.py, service.py, model.py examples] Code Assistant: 📝 New learning recorded - Type: Example (Module code structure) - Router layer handles routing and parameter validation only. All business - logic lives in the Service layer. Model layer uses SQLAlchemy declarative - base. Services call each other through dependency injection. + Type: Example (Module code structure) + Router layer handles routing and parameter validation only. All business + logic lives in the Service layer. Model layer uses SQLAlchemy declarative + base. Services call each other through dependency injection. - [✓ Correct] [✎ Modify] [✗ Undo] + [✓ Correct] [✎ Modify] [✗ Undo] You: [✓ Correct] ``` -Once teaching is complete, the Code Assistant knows your tech stack and project conventions. From now on, just describe what you need and it will write code that fits your team's style. +Once teaching is complete, the Code Assistant knows your tech stack and project conventions. From now on, describe what you need and it will generate designs and code changes according to these standards. ## Complete Conversation Example @@ -163,18 +169,26 @@ Code Assistant: Generated: ## Key Steps -1. **Describe Requirements** -- Tell the agent what feature you need in natural language -2. **Confirm Architecture** -- The agent generates an architecture design for your review -3. **Iterative Development** -- The agent generates code module by module, you confirm step by step -4. **Test and Verify** -- The agent auto-generates tests and runs them to ensure code quality +1. **Lock Down Requirement Boundaries** — Clarify user stories, API scope, data model, permission rules, and what NOT to do +2. **Read Existing Context** — Have the Code Assistant review directory structure, adjacent modules, test patterns, and team standards first +3. **Design Before Coding** — Confirm database tables, API contracts, service boundaries, error handling, and compatibility plans before writing code +4. **Implement in Small Steps** — Split into model, service, router, tests as reviewable small changes +5. **Run Verification Loop** — Execute tests, lint, build, or local startup; preserve failure logs and fix history ## Final Deliverable -A complete feature module with data models, business logic, API routes, and test cases — all following your team's conventions, ready to integrate into your project. +A reviewable set of feature module code, typically including data models, business logic, API routes, and test cases. Code review, test execution, and security checks are still recommended before merging. + +## DesireCore Capabilities Used + +- **Local File System Read/Write**: Directly read project structure, modify files, and preserve diffs in your workspace +- **Command Execution & Test Verification**: Run tests, builds, lint; preserve failure logs and fix history +- **AgentFS Memory**: Team tech stack, directory structure, code style, and test commands can be persisted as long-term rules :::tip Best Practices -- Teach the Code Assistant your tech stack and project structure first — that way the code it writes can be used directly without major refactoring -- When building new features, have it design the architecture first, confirm, then write code — avoids costly rework from going in the wrong direction +- Teach the Code Assistant your tech stack and project structure first — that way generated code will be much closer to team standards +- When developing new features, have it do architecture design first, confirm, then write code — avoids costly rework from going in the wrong direction - For unfamiliar tech (e.g., WebSocket, message queues), just describe the business requirement and the Code Assistant will choose the right technical approach and implement it -- **Develop multiple features in parallel**: When you have several features to build, describe them all at once — the Code Assistant will generate architecture designs and code for each in parallel, all following the same standards you taught it +- When multiple features are developed in parallel, first clarify priorities and dependencies, then have the Code Assistant generate designs and changes separately +- Don't ask it to "build the entire project at once." Real development works better with small commits, test verification, and code review ::: diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/03-finance-assistant.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/03-finance-assistant.md index 8452812..71ce971 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/03-finance-assistant.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/03-finance-assistant.md @@ -1,26 +1,32 @@ --- title: Finance Assistant -description: Use the Finance Assistant agent to complete budget preparation, expense review, financial report generation, and cost analysis. +description: Use the Finance Assistant agent to assist with budget preparation, expense review, report organization, and cost analysis. keywords: [Finance Assistant, Budget Preparation, Expense Review, Financial Reports, Cost Analysis] --- # Finance Assistant -## Scenario Description +## Real Workflow -It's the end of the quarter, and you need to prepare the department budget for next quarter. You need to reference historical data, consider business growth plans, and reasonably allocate various expenses. At the same time, there's a pile of reimbursement documents waiting for review. +Financial scenarios truly exist in "well-defined policies but tedious manual verification" work, such as budget drafts, expense reviews, cost attribution, and report compilation. -The Finance Assistant agent can help you scientifically prepare budgets based on historical data, intelligently review the compliance of reimbursement documents, generate standardized financial reports, and identify cost optimization opportunities. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | Month-end expense review, quarterly budget preparation, expense overrun analysis, departmental cost analysis | +| Existing Materials | Historical budgets, actual expenditures, business plans, expense policies, invoices, and approval records | +| Pain Point | Policy details are intricate, documents are numerous, over-limit reasons and missing materials require line-by-line review | +| DesireCore Intervention | Finance Assistant checks expenses against policies, summarizes over-limit items, missing invoices, budget variances, and adjustment suggestions | +| Acceptance Result | Finance staff receives a budget draft or review report, then handles exceptions and approval opinions | ## Recommended Agent -**Finance Assistant** -- Specializes in financial management, handling budget preparation, expense approval, report archiving, and cost analysis. +**Finance Assistant** — Suited for budget estimation, expense policy compliance, report archiving, cost analysis, and anomaly flagging. -## Teaching First: Setting Your Financial Policies +## Teaching First: Teaching the Agent Your Financial Policies -> Every company has different reimbursement standards. Spend a few minutes teaching the Finance Assistant your company's policies, and it will review expense reports against your specific rules instead of generic standards. +> Every company's reimbursement standards are different. Teach the company policies clearly first, and subsequent reviews will follow your rules instead of generic standards. -### Teaching Rules +### Teach Rules ``` You: I'm sharing our company's expense reimbursement policy for you to learn. @@ -32,18 +38,18 @@ You: I'm sharing our company's expense reimbursement policy for you to learn. 4. Over-limit expenses: Must include a written explanation and department manager approval (preference) Finance Assistant: 📝 New learning recorded - Source: Company_Expense_Policy_2024.pdf + supplementary notes - Type: Rule (Expense reimbursement standards) - Level: Mandatory × 3, Preference × 1 + Source: Company_Expense_Policy_2024.pdf + supplementary notes + Type: Rule (Expense reimbursement standards) + Level: Mandatory × 3, Preference × 1 - My understanding: - When reviewing expense reports, only accept e-receipts from - ride-hailing apps for transportation; meals capped at $15/person - on weekdays, $40/person for client entertainment; accommodation - tiered by city (Tier-1 $80, others $55). Over-limit claims - require written justification with manager sign-off. + My understanding: + When reviewing expense reports, only accept e-receipts from + ride-hailing apps for transportation; meals capped at $15/person + on weekdays, $40/person for client entertainment; accommodation + tiered by city (Tier-1 $80, others $55). Over-limit claims + require written justification with manager sign-off. - [✓ Correct] [✎ Modify] [✗ Undo] + [✓ Correct] [✎ Modify] [✗ Undo] You: [✓ Correct] ``` @@ -106,22 +112,24 @@ Finance Assistant: Travel expenses adjusted to 350k. Reduction suggestions: ## Key Steps -1. **Provide Basic Information** -- Historical data, business plans, budget constraints -2. **Generate Budget Draft** -- The agent generates itemized budget based on data -3. **Adjust and Optimize** -- Adjust various amounts based on your feedback -4. **Expense Review** -- Batch review reimbursement documents, check compliance +1. **Define Accounting Scope** — First clarify department, project, period, budget categories, and approval criteria +2. **Align Budget with Actuals** — Compare budget, actual, forecast, and same-period historical data; identify major variances +3. **Explain Variance Causes** — Break down variances into business growth, price changes, one-time expenses, over-limit costs, or accounting differences +4. **Review Policy Compliance** — Check expense documents for invoice types, amount limits, approval chains, and supplementary explanations +5. **Output Approval Recommendations** — Categorize into approved, returned for materials, over-limit requiring approval, and rejected ## Final Deliverable -A data-driven scientific budget plan and compliance review report. +A budget draft or expense review report containing budget-actual variances, calculation basis, over-limit items, missing materials, and items pending confirmation. :::warning Disclaimer -The Finance Assistant's suggestions are for reference only and do not constitute professional financial or tax advice. For major financial decisions, consult a professional accountant. +The Finance Assistant provides supplementary organization and preliminary analysis, and does not constitute professional financial, audit, or tax advice. For major financial decisions, tax filings, and external disclosures, consult qualified professionals. ::: :::tip Best Practices -- Provide as complete historical data as possible when preparing budgets, budget accuracy will significantly improve +- Provide as complete historical data as possible when preparing budgets — budget accuracy will significantly improve - Expense review can be configured with your company's reimbursement standards (such as travel standards, entertainment limits), and the agent will strictly review according to standards -- At month-end and quarter-end, have the Finance Assistant automatically generate financial reports with key metrics and anomaly alerts -- **Month-end batch review**: Send all expense reports for the month to the Finance Assistant at once — it will check each one against the policies you taught it, automatically flag over-limit items, and typically complete the entire review within minutes +- At month-end and quarter-end, have the Finance Assistant compile financial reports with key metrics and anomaly alerts +- During month-end batch reviews, have the Finance Assistant first summarize over-limit items, missing invoices, and items needing supplementary explanations, then have finance staff make final decisions +- Financial analysis should not only look at variance amounts — also explain the nature of variances: temporary expenses, structural changes, revenue growth-driven, or budget planning bias ::: diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/04-product-manager.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/04-product-manager.md index bf39789..cd82f62 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/04-product-manager.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/04-product-manager.md @@ -1,24 +1,30 @@ --- title: Product Manager -description: Use the Product Manager agent to complete requirement document writing, iteration management, resource coordination, and risk control. +description: Use the Product Manager agent to organize requirements, write PRDs, track iterations, and identify project risks. keywords: [Product Manager, Requirements Document, Iteration Management, Sprint, Project Management] --- # Product Manager -## Scenario Description +## Real Workflow -You're managing a product's iterative development. This week you need to write requirement documents, track Sprint progress, coordinate next week's resource allocation, and prepare a cross-departmental launch coordination meeting. None of these tasks are difficult individually, but together they leave you overwhelmed. +The Product Manager scenario is not "AI deciding product direction for you" — it's about organizing scattered requirements, progress, and risks into materials that the team can review and execute on. -The Product Manager agent can assist you with these daily tasks: writing structured requirement documents, tracking Sprint progress in real-time, reasonably allocating development resources, and identifying project risks. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | PRD review, Sprint planning meeting, pre-standup progress summary, launch coordination | +| Existing Materials | Requirements background, user feedback, PRD template, task lists, development progress, and blockers | +| Pain Point | User stories unclear, acceptance criteria vague, multiple projects competing for resources, risks discovered too late | +| DesireCore Intervention | Product Manager agent generates PRD drafts, breaks down acceptance criteria, summarizes progress, and flags resource conflicts | +| Acceptance Result | PM receives review-ready documents and risk lists, then confirms priorities, scheduling, and trade-offs | ## Recommended Agent -**Product Manager** -- Specializes in project management and coordination, excelling in progress tracking, resource allocation, risk control, and team coordination. +**Product Manager** — Suited for requirements documents, iteration planning, progress summaries, resource conflict analysis, and risk lists. -## Teaching First: Setting Your Product Workflow +## Teaching First: Teaching the Agent Your Product Workflow -> Every team has its own iteration process and documentation standards. Teach your standards once, and every PRD and schedule it produces will be ready to use without constant format adjustments. +> Every team's iteration process and documentation standards are different. Teach your templates, priority definitions, and acceptance criteria first, and subsequent outputs will be more review-ready. ### Teach Rules @@ -31,17 +37,17 @@ You: Let me teach you our team's product standards: 4. Acceptance criteria: every requirement must have testable acceptance conditions — no vague descriptions like "improve experience" (mandatory) Product Manager: 📝 New learning recorded - Source: PRD_Template_v3.docx + workflow notes - Type: Rule (Product workflow standards) - Level: Mandatory × 3, Preference × 1 + Source: PRD_Template_v3.docx + workflow notes + Type: Rule (Product workflow standards) + Level: Mandatory × 3, Preference × 1 - My understanding: - Use the uploaded template format when writing PRDs. User stories - must follow the "As a... I want... so that..." pattern. Priorities - are split into P0/P1/P2 levels. Every requirement must include - testable acceptance criteria. Iterations follow a bi-weekly cadence. + My understanding: + Use the uploaded template format when writing PRDs. User stories + must follow the "As a... I want... so that..." pattern. Priorities + are split into P0/P1/P2 levels. Every requirement must include + testable acceptance criteria. Iterations follow a bi-weekly cadence. - [✓ Correct] [✎ Modify] [✗ Undo] + [✓ Correct] [✎ Modify] [✗ Undo] You: [✓ Correct] ``` @@ -105,18 +111,20 @@ Product Manager: User Tag System -- Requirements Document ## Key Steps -1. **Check Iteration Progress** -- Ask about current Sprint completion and risks at any time -2. **Write Requirements Documents** -- Provide feature descriptions, the agent generates structured PRD -3. **Resource Scheduling** -- Analyze multi-project resource conflicts and provide allocation suggestions -4. **Risk Management** -- Automatically identify project risks and propose countermeasures +1. **Define Problem and Goals** — Clarify user problems, business objectives, success metrics, and non-goals +2. **Break Down User Stories** — Decompose into user stories by role, behavior, and value — don't just pile up feature lists +3. **Define Acceptance Criteria** — Every requirement needs testable conditions, edge cases, and analytics/data requirements +4. **Sync Development Constraints** — Write dependent systems, resource conflicts, technical risks, and launch windows into the plan +5. **Track Iteration Risks** — Daily review of completed items, blockers, scope changes, and delay risks ## Final Deliverable -Efficient iteration management support, including progress reports, requirements documents, resource allocation plans, and risk alerts. +Progress reports, PRD drafts, user stories, acceptance criteria, resource allocation suggestions, and risk lists. Final scheduling should be confirmed based on actual team capacity and business priorities. :::tip Best Practices -- Have the agent generate progress summaries before daily stand-ups to quickly understand team status -- When writing requirements documents, first describe business scenarios and user pain points, and the agent will automatically break them down into user stories and acceptance criteria -- When multiple projects run in parallel, have the agent analyze resource conflicts and recommend optimal allocation plans -- **Generate PRDs in parallel**: During quarterly planning with multiple feature requests, tell the Product Manager agent about all of them at once — it will generate PRD drafts for each feature in parallel, all using the template and standards you taught it +- Have the agent generate progress summaries before daily stand-ups to quickly see completed items, blockers, and pending decisions +- When writing requirements documents, first describe business scenarios and user pain points, and the agent will automatically decompose them into user stories and acceptance criteria +- When multiple projects run in parallel, have the agent analyze resource conflicts and recommend alternative allocation plans +- During quarterly planning with multiple feature requests, have the Product Manager agent generate PRD drafts for each feature in parallel, then conduct a unified priority review +- PRDs are not better when longer — the key is ensuring engineering, testing, design, and business have the same understanding of "done" ::: diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/05-bidding-assistant.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/05-bidding-assistant.md index e5f9a76..48860f3 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/05-bidding-assistant.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/05-bidding-assistant.md @@ -1,24 +1,30 @@ --- title: Bidding Assistant -description: Use the Bidding Assistant agent to complete bid document preparation, qualification review, pricing strategy analysis, and bid document checking. +description: Use the Bidding Assistant agent to assist with tender document interpretation, qualification self-review, bid preparation and submission checks. keywords: [Bidding, Bid Document Preparation, Qualification Review, Pricing Strategy, Bid Management] --- # Bidding Assistant -## Scenario Description +## Real Workflow -Your company received a project bidding document with 18 days until the bid deadline. You need to understand qualification requirements, write technical proposals, develop pricing strategies, and ensure the bid documents are flawless. Any oversight in any step could lead to disqualification. +The core of bidding scenarios isn't "writing bid documents"—it's managing a set of materials, qualifications, pricing and format requirements around a deadline. -The Bidding Assistant agent can help you systematically complete the entire bidding process: interpreting bidding documents, self-reviewing bidding qualifications, generating technical proposal frameworks, analyzing optimal pricing, and item-by-item checking bid document completeness. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | After receiving tender documents, need to complete qualification assessment, proposal writing and material submission before deadline | +| Existing Materials | Tender documents, company qualifications, track record, personnel certificates, bid templates, pricing worksheets | +| Pain Point | Qualification gaps discovered too late; scoring criteria not fully analyzed; signatures, stamps and attachment lists easily missed | +| DesireCore Intervention | Bidding Assistant extracts timeline, qualification requirements and scoring criteria, matches qualifications and track record, generates checklists | +| Acceptance Result | Bid lead gets qualification self-review, bid framework and pre-submission checklist, then coordinates commercial, technical and finance to proceed | ## Recommended Agent -**Bidding Assistant** -- Bidding process management expert, responsible for bid document preparation, qualification review, and pricing strategy. +**Bidding Assistant** — Suitable for tender document interpretation, qualification self-review, technical proposal framework, pricing estimation and bid document completeness checks. -## Teaching First: Building Your Company Profile +## Teaching First: Build Your Company Profile -> The most time-consuming part of bidding is repeatedly compiling company qualifications and past performance. Teach once, and every new project gets automatic matching from then on. +> The most time-consuming part of bidding is repeatedly compiling company qualifications and past performance. Maintain qualification and track record libraries in advance—subsequent projects will be easier to match materials. ### Teaching Rules @@ -112,19 +118,21 @@ Bidding Assistant: Qualification Self-Review Results: Basically Qualified (3 ite ## Key Steps -1. **Bidding Document Interpretation** -- The agent extracts key requirements, timeline, and scoring criteria -2. **Qualification Self-Review** -- Item-by-item check if bidding conditions are met, flag missing items -3. **Technical Proposal Generation** -- Generate proposal framework based on bidding requirements, highlighting scoring focus -4. **Pricing Strategy Analysis** -- Based on cost estimation and competitive analysis, recommend optimal pricing -5. **Document Completeness Check** -- Item-by-item check if all documents are complete before bidding +1. **Go/No-Go Initial Screening** — First check mandatory qualification requirements, profit floor, delivery capacity and competitive landscape to decide whether to bid +2. **Analyze Tender Documents** — Extract timeline, scoring criteria, required qualifications, response format, attachment list and disqualification risks +3. **Build Compliance Matrix** — Each tender requirement maps to response location, responsible person, evidence materials and completion status +4. **Assign Content Responsibilities** — Technical, commercial, finance and legal each handle corresponding sections; Bidding Assistant maintains master checklist +5. **Pricing and Risk Review** — Pricing range should simultaneously consider cost, profit, warranty, payment cycle and unfavorable contract terms +6. **Pre-Submission Check** — Verify signatures, stamps, bid bonds, attachments, page numbers, table of contents and electronic/hard copy format against checklist ## Final Deliverable -A complete set of bidding materials, including bidding document interpretation report, qualification self-review checklist, technical proposal, pricing analysis, and bid document check list. +A set of bid materials ready for further review and refinement, including Go/No-Go conclusion, compliance matrix, qualification self-review checklist, technical proposal framework, pricing estimation and submission checklist. :::tip Best Practices -- Have the agent analyze immediately after receiving the bidding document to discover qualification gaps early and allow time for supplementation -- Focus technical proposal efforts on high-scoring chapters, the agent will highlight which sections need focused writing based on scoring criteria +- Have the agent analyze immediately after receiving the tender document to discover qualification gaps early and allow time for supplementation +- Focus technical proposal efforts on high-scoring chapters—the agent will highlight which sections need focused writing based on scoring criteria - Be sure to do a completeness check the day before bidding to avoid disqualification due to formal issues (such as missing signatures/stamps) -- **Parallel bidding**: When handling multiple RFPs simultaneously, send them all to the Bidding Assistant at once -- it will analyze each bidding document in parallel and automatically match the required qualifications and track record from your taught company profile -::: +- When handling multiple tender projects simultaneously, have the Bidding Assistant separately output qualification gaps and timelines for each, then have project leads decide resource allocation +- Real bid management isn't "finishing the bid document"—it's ensuring every tender requirement has a response, evidence, responsible person and status +::: \ No newline at end of file diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/06-real-estate.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/06-real-estate.md index 216fcf4..94160c0 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/06-real-estate.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/06-real-estate.md @@ -1,20 +1,26 @@ --- title: Real Estate Advisor -description: Use the Real Estate Advisor agent for property analysis, price evaluation, home purchase policy interpretation, and multi-property comparison. +description: Use the Real Estate Advisor agent to assist with property comparison, price reference, policy interpretation, and purchase needs assessment. keywords: [Real Estate Advisor, Property Analysis, Price Evaluation, Home Purchase Decision, Real Estate Investment] --- # Real Estate Advisor -## Scenario Description +## Real Workflow -You're preparing to buy a house and have looked at several properties. Every agent says their recommendation is the best. You're not sure which property truly suits you or whether the quoted price is reasonable. You need objective, professional analysis to assist your decision. +Real estate scenarios truly exist in the "too much information but unclear decision dimensions" stage. Users need not a single recommendation, but a clear view of the trade-offs. -The Real Estate Advisor agent can help you analyze property advantages and disadvantages, evaluate price reasonableness, interpret changes in home purchase policies, and systematically compare multiple candidate properties, giving your home purchase decision data support. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | Multiple properties entering shortlist, need to compare for self-use, commute, school district, budget, and risks | +| Existing Materials | Budget, family plans, workplace location, property listings, mortgage terms, public transaction data | +| Pain Point | Agent information is one-sided; monthly payments, taxes, amenities, policies, and long-term needs are all mixed together | +| DesireCore Intervention | Real Estate Advisor organizes comparison tables, estimates monthly payments, lists items to verify, and outlines trade-offs for different choices | +| Acceptance Result | User gets a viewing and decision checklist — bringing questions to verify on-site rather than just accepting recommendations | ## Recommended Agent -**Real Estate Advisor** -- Real estate domain professional advisor, providing property analysis, price evaluation, and home purchase decision support. +**Real Estate Advisor** — Suited for property information organization, price reference, home purchase policy interpretation, and multi-property comparison. ## Complete Conversation Example @@ -23,7 +29,7 @@ You: I'm torn between Property A and Property B. Can you help me compare? Property A: 85k/sqm, 89 sqm 3-bedroom, 500m from subway Property B: 78k/sqm, 95 sqm 3-bedroom, 1.2km from subway -Real Estate Advisor: Okay, let me do a comprehensive comparison analysis for you. +Real Estate Advisor: Okay, let me prepare a comparison based on the information you provided. Basic Information Comparison | Dimension | Property A | Property B | @@ -43,22 +49,22 @@ Real Estate Advisor: Okay, let me do a comprehensive comparison analysis for you Property B Advantages: - 8% lower unit price, better value - 6 sqm larger, more spacious 3-bedroom - - More mature surrounding commercial配套 + - More mature surrounding commercial amenities Purchase Recommendation - Value school district and commute → Choose A - - Value value and living comfort → Choose B + - Value affordability and living comfort → Choose B What is the main purpose of your purchase? Self-use or investment? Do you have school district needs? -You: Buying as wedding home for self-use, planning to have children within 2 years, working in CBD. +You: Buying as a wedding home for self-use, planning to have children within 2 years, working in CBD. Real Estate Advisor: Based on your situation, I recommend Property A, reasons: - 1. School district needs: Property A's对口school ranks higher + 1. School district needs: Property A's feeder school ranks higher 2. Commute: 500m to subway, about 35 minutes to CBD 3. Value retention: School district homes are more resilient during market adjustments - 4. Total price difference only 160k, but school district and commute advantages are obvious + 4. Total price difference only 160k, but school district and commute advantages are clear Monthly payment reference (30% down payment, 30-year loan): Property A: Monthly payment approx 28,500 CNY @@ -68,17 +74,23 @@ Real Estate Advisor: Based on your situation, I recommend Property A, reasons: ## Key Steps -1. **Clarify Needs** -- Tell the agent budget, purpose, preferences (school district, commute, etc.) -2. **Property Analysis** -- The agent analyzes location, supporting facilities, developer background -3. **Price Evaluation** -- Reference surrounding transaction data to evaluate quote reasonableness -4. **Comparison Decision** -- Systematically compare multiple candidate properties and give recommendations +1. **Define Purchase Constraints** — First clarify budget, down payment, monthly payment cap, commute, school district, floor plan, and holding period +2. **Organize Candidate Properties** — Put total price, area, floor level, transportation, amenities, taxes, and mortgage terms into the same comparison table +3. **Verify Key Risks** — Include property rights, purchase restrictions, mortgage eligibility, historical transactions, property management, school district, noise, and quality issues in the verification checklist +4. **Calculate True Cost** — Don't just look at property price; also consider taxes, agent fees, renovation, monthly payments, interest rates, and cash flow pressure +5. **Bring Questions to Viewings** — Output an on-site verification checklist rather than just giving a "choose A / choose B" conclusion ## Final Deliverable -An objective property comparison analysis report, including price evaluation, advantage/disadvantage comparison, and personalized home purchase recommendations. +A property comparison analysis draft containing total price, true cost, commute, amenities, school district, monthly payment reference, items to verify, and on-site viewing questions. + +:::warning Disclaimer +The Real Estate Advisor agent provides information organization and supplementary analysis, and does not constitute investment, mortgage, or legal advice. Before purchasing, verify property details, policies, mortgage terms, taxes, and contract clauses. +::: :::tip Best Practices -- Try to provide complete purchase needs (budget, purpose, family situation, commute location) for more precise recommendations -- When unsure about agent quotes, have the agent query surrounding similar property transaction prices for reference -- When policies change (such as interest rate adjustments, purchase restrictions easing), consult the agent promptly to understand specific impacts on you +- Provide complete purchase requirements (budget, purpose, family situation, commute location) for more precise comparisons +- When unsure about an agent's quoted price, have the agent query surrounding similar property transaction prices for reference +- When policies change (such as interest rate adjustments, purchase restriction easing), consult the agent promptly to understand specific impacts on you +- Don't just look at unit price for property decisions — property rights, mortgage, taxes, school district eligibility, property management, and the final contract all affect the outcome ::: diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/07-data-scientist.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/07-data-scientist.md index 4babab4..8ba5a6a 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/07-data-scientist.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/02-professional/07-data-scientist.md @@ -1,20 +1,26 @@ --- title: Data Scientist -description: Use the Data Analyst agent for advanced data analysis, user behavior insights, predictive models, and operational monitoring. +description: Use the Data Analyst agent to assist with user behavior analysis, predictive modeling, operational monitoring and experiment evaluation. keywords: [Data Science, User Behavior Analysis, Predictive Models, Operational Monitoring, Visualization] --- # Data Scientist -## Scenario Description +## Real Workflow -You're the head of the product team. Recently, user churn has increased, but you're not sure why. You have a lot of user behavior data and need someone to help you find clues, build predictive models, and provide actionable improvement suggestions. +Data science scenarios aren't about "letting AI find insights"—they're about when teams already have data and competing hypotheses, using analysis to narrow down to verifiable directions. -The Data Analyst agent can not only do basic statistical analysis but also handle complex user behavior analysis, build predictive models, and set up operational monitoring dashboards to help you discover business opportunities from data. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | User churn rising, conversion declining, A/B experiment results unclear, operational strategy needs review | +| Existing Materials | Behavior logs, user profiles, channel sources, version changes, experiment groups and business hypotheses | +| Pain Point | Reasons behind metric changes unclear; different teams have different explanations; lack unified analysis framework | +| DesireCore Intervention | Data Analyst does funnels, segmentation, anomaly detection and predictive model first draft, organizes verifiable hypotheses | +| Acceptance Result | Team gets prioritized recommendations, experiment suggestions and monitoring metrics, then decides product or operational actions | ## Recommended Agent -**Data Analyst** -- Supports advanced statistical analysis, predictive models, user segmentation, anomaly detection, and real-time monitoring dashboards. +**Data Analyst** — Suitable for statistical analysis, predictive models, user segmentation, anomaly detection and operational monitoring dashboards. ## Complete Conversation Example @@ -66,18 +72,20 @@ Data Analyst: Improvement Priority Ranking (by expected effect): ## Key Steps -1. **Upload Behavior Data** -- Provide user behavior logs, event data, etc. -2. **Funnel Analysis** -- The agent automatically identifies key churn nodes in the user journey -3. **Deep Attribution** -- Analyze specific reasons for churn nodes -4. **Predictive Modeling** -- Predict future trends based on historical data -5. **Improvement Suggestions** -- Provide actionable improvement plans prioritized by importance +1. **Define Business Problem** — First clarify whether it's churn, conversion, retention, payment or activity problem +2. **Organize Event Definitions** — Confirm user ID, event names, timestamps, channels, versions and experiment groups are consistent +3. **Build Funnels and Segmentation** — Break down changes by new/returning users, channels, devices, versions and behavior depth +4. **Form Testable Hypotheses** — Match hypotheses like "onboarding too long" or "entry not obvious" with data evidence +5. **Build Models and Monitoring** — Models only serve as risk ranking; ultimately land on monitorable metrics and experiment plans +6. **Output Action Priority** — Sort by impact scope, implementation cost, verification cycle and risk ## Final Deliverable -A complete user behavior analysis report, including funnel analysis, churn attribution, user segmentation, predictive models, and priority-ranked improvement plans. +A user behavior analysis report containing data definitions, funnel analysis, churn attribution, user segmentation, predictive model first draft, monitoring metrics and prioritized experiment suggestions. :::tip Best Practices -- The more complete the data, the more accurate the analysis. Ideally include user behavior events, timestamps, device information, and other dimensions +- The more complete the data, the more accurate the analysis. Ideally include user behavior events, timestamps, device information and other dimensions - Start with global analysis to understand the overall situation, then dive deep into specific issues. Avoid getting into details from the start -- Analysis results should translate into executable product improvements, you can have the agent prioritize by "input-output ratio" -::: +- Analysis results should translate into executable product improvements—have the agent prioritize by impact scope, implementation cost and verification method +- User behavior analysis shouldn't jump directly to "reasons"—first confirm tracking definitions, version changes and sample composition are stable +::: \ No newline at end of file diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/03-solutions/01-ai-copyright.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/03-solutions/01-ai-copyright.md index f28e6b9..affb69b 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/03-solutions/01-ai-copyright.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/03-solutions/01-ai-copyright.md @@ -1,21 +1,27 @@ --- title: AI Copyright Registration -description: Use DesireCore to fully automate software copyright registration, from requirement description to document generation, all in one place. +description: Use DesireCore to assist with preparing software copyright application materials, including documentation, user manuals, screenshots, and code materials. keywords: [Copyright Registration, Software Copyright, Automation, Intellectual Property, Copyright Writing] --- # AI Copyright Registration -## Scenario Description +## Real Workflow -You need to apply for software copyright (Software Copyright) for your company's software product, for high-tech enterprise certification or intellectual property protection. The traditional method requires preparing a lot of materials: source code identification materials, software description, user operation manual, interface screenshots...... Having an agency do a set costs at least two to three thousand yuan, and doing it yourself requires familiarity with the process. +The software copyright scenario truly exists in the "functionality is ready, but application materials aren't assembled yet" stage. The problem is usually not being unable to write a description, but not knowing how to align the documentation, manual, screenshots, and code materials. -DesireCore's Software Copyright Writer agent can fully automate this process: you only need to describe the software's functions in natural language, and it will automatically develop a demo application in an isolated sandbox environment, automatically take screenshots, and then generate the complete set of application materials. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | High-tech enterprise certification, intellectual property archiving, preparing copyright application materials after product launch | +| Existing Materials | Software name, version number, feature descriptions, existing code, interface screenshots or prototypes | +| Pain Point | Unfamiliar with material formats, documentation and screenshots inconsistent, confusing module naming, multiple rounds of corrections | +| DesireCore Intervention | Copyright Writer organizes modules, generates documentation and manual drafts, completes demo screenshots and code materials | +| Acceptance Result | Responsible person receives a verifiable material package, then checks ownership, version numbers, screenshot and code consistency | ## Involved Agents -- **Software Copyright Writer** -- Core agent, responsible for requirement understanding, automatic development, screenshots, and material generation -- **Code Assistant** -- Optional collaboration to ensure the quality and standardization of generated code +- **Copyright Writer** — Core agent, responsible for feature organization, demo interfaces, screenshots, and material drafts +- **Code Assistant** — Optional collaboration for checking demo code structure and consistency ## Complete Process @@ -25,7 +31,7 @@ DesireCore's Software Copyright Writer agent can fully automate this process: yo You: I developed a Markdown editor supporting real-time preview, PDF export, and cloud sync. Software name is "MoXie", version V1.0. -Software Copyright Writer: Okay, I'll generate copyright application materials for "MoXie V1.0". +Copyright Writer: Okay, I'll generate copyright application materials for "MoXie V1.0". Based on your description, I've planned the following functional modules: @@ -35,29 +41,29 @@ Software Copyright Writer: Okay, I'll generate copyright application materials f 4. Cloud Sync -- Login/registration, sync status, version history 5. Settings -- Theme switching, shortcut configuration - Confirm to start automatic development. Estimated time: 15-20 minutes. + Confirm to start generating demo interfaces and material drafts. ``` -### Step 2: Automatic Development and Screenshots +### Step 2: Generate Demo Interfaces and Screenshots -After confirming the function planning, the agent will automatically in the sandbox environment: -- Generate complete frontend application code +After confirming the function planning, the agent will in the sandbox environment: +- Generate demo frontend interface code - Take screenshots page by page -- Ensure screenshots contain software name identification +- Verify that screenshots contain the software name identifier -### Step 3: Generate Complete Materials +### Step 3: Generate Material Drafts ``` -Software Copyright Writer: All completed! +Copyright Writer: Material drafts generated. - Copyright Materials List: + Copyright Materials Checklist: 1. source_code_identification_materials.txt (first 30 pages + last 30 pages) 2. software_description.txt (detailed function description) 3. user_operation_manual.txt (operation guide based on screenshots) 4. screenshots/ (5 interface screenshots, 1920x1080) 5. softreg_materials.zip (all packaged) - You can directly download the ZIP package and submit to the Copyright Protection Center. + Please download the ZIP package, verify contents, then submit to the Copyright Protection Center. ``` ### Step 4: Submission and Correction @@ -67,36 +73,43 @@ If the Copyright Center requires corrections, just tell the agent the correction ``` You: The copyright was returned, saying screenshots are unclear and description is too simple. -Software Copyright Writer: I'll help you fix it: +Copyright Writer: I'll help you fix it: 1. Regenerate 1920x1080 HD screenshots 2. Expand description from 500 words to 2000+ words 3. Supplement functional flowchart description - Updated, can resubmit. + Updated, please verify before resubmitting. ``` ## Key Steps -1. **Describe Functions** -- Describe your software functions and name in natural language -2. **Confirm Planning** -- The agent generates functional module planning, you confirm or adjust -3. **Auto Generate** -- The agent automatically develops, takes screenshots, generates complete materials -4. **Download and Submit** -- Download the packaged ZIP file and submit to the Copyright Protection Center +1. **Confirm Application Entity and Ownership** — First clarify applicant, software name, version number, development completion date, and rights ownership materials +2. **Organize Feature Modules** — Break software functions into interface, data, permissions, import/export, and settings modules to avoid inconsistent documentation +3. **Generate Documentation Identification Materials** — Generate software description or user manual; headers, software name, version number, and application information must stay consistent +4. **Prepare Source Code Identification Materials** — Prepare consecutive pages from the beginning and end of source code per requirements; submit complete source code if less than 60 pages +5. **Align Screenshots with Description** — Screenshots must correspond to main features in the documentation; avoid name, version, or module inconsistencies +6. **Correction Loop** — If corrections are requested, preserve the original submission, modify against correction comments item by item, avoid changing one thing and breaking others ## Final Deliverable -A complete set of copyright application materials, including source code identification materials, software description, user operation manual, and interface screenshots, ready to be submitted to the National Copyright Protection Center. +A verifiable software copyright application material package containing application information, source code identification materials, documentation identification materials, user operation manual, and interface screenshots. ## Notes -- Software name suggested format "XX Software" or "XX System" +- Software name recommended format: "XX Software" or "XX System" - Version number for first application starts from V1.0 -- Screenshots must contain software name or identification -- Source code identification materials need first 30 pages + last 30 pages totaling 60 pages -- For enterprise batch applications, supports parallel generation of multiple materials +- Screenshots must contain the software name or identifier +- Source code and documentation identification materials typically consist of the first and last 30 consecutive pages each; submit all if less than 60 pages +- Before submission, verify that application entity, ownership information, software name, version number, and material content are all consistent + +:::warning Disclaimer +The Copyright Writer agent provides material preparation assistance, does not guarantee application approval, and does not constitute intellectual property legal advice. For ownership disputes, commissioned development, work-for-hire, or external licensing, consult qualified professionals. +::: :::tip Best Practices -- The more detailed the function description, the more the generated materials match reality. Including specific functional points (e.g., "supports Markdown syntax highlighting") is much better than vague descriptions (e.g., "an editor") -- If your software has specific industry background (e.g., "smart agriculture", "education platform"), tell the agent, and it will reflect industry characteristics in screenshots and descriptions -- Before application, you can have the Legal Advisor agent review material compliance together -- **Teach your style preferences upfront**: If you have specific requirements for description style or module naming conventions (e.g., "use four-character phrases for all module names"), teach them to the Software Copyright Writer first — no need to adjust every time when batch-generating materials +- The more detailed the function description, the more the materials match reality. Including specific features (e.g., "supports Markdown syntax highlighting") is much better than vague descriptions (e.g., "an editor") +- If your software has specific industry background (e.g., "smart agriculture", "education platform"), tell the agent, and it will reflect industry characteristics in screenshots and documentation +- Before application, you can have the Legal Advisor agent review material compliance +- If you have specific requirements for documentation style or module naming conventions, teach them to the Copyright Writer first — subsequent batch generation will be more consistent +- The most common issue with copyright materials is consistency: the software name and version number must match exactly across application forms, documentation, source code headers, and screenshots ::: diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/03-solutions/02-ai-bidding.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/03-solutions/02-ai-bidding.md index ba654f3..6c00284 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/03-solutions/02-ai-bidding.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/03-solutions/02-ai-bidding.md @@ -1,24 +1,30 @@ --- title: AI Bidding -description: End-to-end bidding solution with multi-agent collaboration, covering the entire process from bidding document analysis to bid submission. +description: Multi-agent collaboration to assist with bidding document analysis, bid preparation, pricing estimation and submission checking. keywords: [AI Bidding, Bid Document Preparation, Multi-agent Collaboration, Bid Management, Quality Control] --- # AI Bidding -## Scenario Description +## Real Workflow -A large bidding project involves many aspects: bidding document interpretation, qualification review, technical proposal writing, pricing strategy formulation, compliance checking, document format review...... requiring coordination across legal, financial, technical, and business domains. The traditional approach is for each department to work separately and then manually integrate, which is time-consuming and error-prone. +The AI Bidding scenario applies to large projects requiring multi-person collaboration. The problem isn't being unable to write individual bid sections—it's that legal, finance, technical and business each advance in parallel, and final integration easily misses items. -DesireCore's AI Bidding solution enables multi-agent collaboration, allowing agents from various professional domains to work in parallel, cross-review, compressing the bidding cycle by more than half. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | Large bidding project kick-off, requiring multi-role parallel handling of qualifications, proposals, pricing and compliance | +| Existing Materials | Bidding document, company qualifications database, past performance, cost estimation, contract clauses, bid templates | +| Pain Point | Complex task dependencies, version conflicts, Q&A, pricing, signatures and attachment checklists must not be missed | +| DesireCore Intervention | Core dispatcher assigns tasks to bidding, legal, finance, data and copywriting agents, and consolidates progress | +| Acceptance Result | Project lead sees daily progress, gaps and pending confirmations, ultimately receiving a complete bid materials package and checklist | ## Involved Agents -- **Bidding Assistant** -- Coordinates the entire bidding process, prepares bid documents -- **Legal Advisor** -- Reviews compliance and contract terms of bidding documents -- **Finance Assistant** -- Cost estimation, pricing strategy analysis -- **Data Analyst** -- Competitor analysis, market data support -- **AI Copywriter** -- Copy polishing, layout integration +- **Bidding Assistant** — Coordinates the entire bidding process, prepares bid documents +- **Legal Advisor** — Reviews compliance and contract terms of bidding documents +- **Finance Assistant** — Cost estimation, pricing strategy analysis +- **Data Analyst** — Competitor analysis, market data support +- **AI Copywriter** — Copy polishing, layout integration ## End-to-End Process @@ -45,6 +51,14 @@ Bidding Assistant: Interpreting bidding document... Bidding Assistant completes qualification self-review, Finance Assistant completes cost estimation, Legal Advisor provides compliance suggestions. Analysis results from all agents are compiled into a comprehensive evaluation report. +This phase should produce a compliance matrix: + +| Bidding Requirement | Response Location | Responsible | Supporting Evidence | Status | +|---------------------|-------------------|-------------|---------------------|--------| +| ISO9001 Certificate | Business Section 2.1 | Business | Certificate scan | Available | +| 3 similar projects in past 3 years | Business Section 3.2 | Business | Contract + Acceptance | Need 1 acceptance doc | +| 3-year warranty | Technical Proposal 6.4 / Pricing | Technical + Finance | After-sales plan + Cost estimation | Cost TBD | + ### Phase 3: Technical Proposal Writing (Days 6-12) ``` @@ -88,37 +102,30 @@ Bidding Assistant: Bidding document completeness check results: ## Multi-Agent Collaboration Flowchart -``` -Bidding Document ──→ Bidding Assistant (Coordination) - ├──→ Legal Advisor (Compliance Review) - ├──→ Finance Assistant (Cost Estimation) - └──→ Data Analyst (Competitive Analysis) - │ - Comprehensive Analysis Report - │ - ├──→ Bidding Assistant (Technical Proposal) - └──→ AI Copywriter (Layout Integration) - │ - ├──→ Legal Advisor (Cross Review) - └──→ Finance Assistant (Pricing Review) - │ - Final Bidding Documents -``` +![Multi-Agent Collaboration](/img/use-cases/ai-bidding-flow.svg) ## Quality Control -- **Cross Review**: After each agent completes their work, other agents cross-check -- **Risk Flagging**: All high-risk items are automatically flagged and must be confirmed by you -- **Version Management**: Each modification is automatically saved as a version, can be traced back at any time -- **Completeness Check**: Item-by-item verification of document checklist before submission +- **Go/No-Go Records**: Preserve the judgment basis for whether to bid, avoiding consuming massive resources on unsuitable projects +- **Compliance Matrix**: Every bidding requirement maps to a response location, responsible person, supporting evidence and status +- **Cross Review**: Technical, business, finance and legal each check clauses that affect the others +- **Version Management**: Bid documents, pricing and attachments should have version numbers to avoid each file being modified independently +- **Submission Checklist**: Verify signatures/stamps, deposits, digital format, physical copies, page numbers, table of contents and sealing requirements item by item + +## DesireCore Capabilities Used + +- **Multi-Agent Team Collaboration**: Bidding, legal, finance, data and copywriting agents work in parallel +- **Workflow / SOP**: Formalize qualification review, compliance matrix, section writing and submission checks into a process +- **Super Document**: Bid modifications presented in a reviewable format, facilitating accept, reject and rollback per item +- **Security Approval & Audit**: Key checkpoints like pricing, commitment clauses and stamp submissions preserve human confirmation and records ## Final Deliverable -A complete set of bidding materials, cross-reviewed by multiple agents, covering technical proposals, business pricing, qualification proofs, and compliance review. +A reviewable bid materials package covering Go/No-Go judgment, compliance matrix, technical proposal, business pricing, qualification proofs, compliance review and pre-submission checklist. :::tip Best Practices - Start analysis immediately after receiving the bidding document to allow sufficient time for subsequent writing -- The Q&A session is important, unfavorable terms found by the Legal Advisor can be clarified by requesting the bidder to clarify -- Allow at least half a day for completeness check before final submission, don't wait until the last minute -- **Upload your company profile in advance**: Teach the Bidding Assistant your business license, qualification certificates, and past winning bids upfront — it will automatically match relevant qualifications for each new project without you having to compile them every time -::: +- The Q&A session is important—unfavorable terms found by the Legal Advisor can be clarified by requesting the bidder to answer questions +- Allow at least half a day for completeness check before final submission; don't wait until the last minute +- Upload your company qualifications database, past performance and bid templates in advance so subsequent projects can match materials faster +::: \ No newline at end of file diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/03-solutions/03-enterprise-coordination.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/03-solutions/03-enterprise-coordination.md index 5c2d40b..a556102 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/03-solutions/03-enterprise-coordination.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/03-solutions/03-enterprise-coordination.md @@ -1,24 +1,30 @@ --- title: Enterprise Service Coordination -description: Use DesireCore's core scheduling agent to achieve cross-department collaboration, multi-agent orchestration, and project management integration. +description: Use DesireCore's core scheduling agent to decompose complex tasks, delegate to professional agents and track progress. keywords: [Enterprise Coordination, Multi-agent Collaboration, Cross-department, Task Orchestration, DesireCore] --- # Enterprise Service Coordination -## Scenario Description +## Real Workflow -Your company is releasing a new product next month. This involves Legal reviewing product copy, Marketing creating promotional materials, Finance calculating cost pricing, Project Manager establishing release timeline...... The traditional approach requires you to contact each department individually, follow up on progress, and integrate information, resulting in extremely high coordination costs. +Enterprise service coordination scenarios apply to cross-role, cross-material, cross-timeline tasks. The real problem isn't "no one is doing it"—it's that goal decomposition, progress synchronization and result integration costs are too high. -DesireCore's core scheduling agent is like a project manager. It can understand your overall requirements, automatically break down tasks and delegate them to the most suitable professional agents for parallel execution, monitor progress in real-time, and finally integrate all deliverables for you. +| Dimension | Real Situation | +|-----------|----------------| +| Trigger Point | New product launches, customer complaint escalations, compliance rectification, quarterly business reviews and other cross-department tasks | +| Existing Materials | Overall goal, deadlines, existing documents, related policies, pending decisions and involved roles | +| Pain Point | Many subtasks, unclear dependencies, scattered responsible parties; managers constantly chase progress and consolidate results | +| DesireCore Intervention | Core dispatcher decomposes tasks, delegates to agents, tracks blockers and centralizes items needing human decisions | +| Acceptance Result | Manager sees unified progress, deliverables and pending decisions rather than searching for information across multiple chat groups | ## Involved Agents -- **DesireCore Core** -- Task orchestration, agent scheduling, global monitoring -- **Legal Advisor** -- Compliance review -- **AI Copywriter** -- Marketing material writing -- **Finance Assistant** -- Cost accounting and pricing -- **Product Manager** -- Release timeline and task allocation +- **DesireCore Core** — Task orchestration, agent scheduling, global monitoring +- **Legal Advisor** — Compliance review +- **AI Copywriter** — Marketing material writing +- **Finance Assistant** — Cost accounting and pricing +- **Product Manager** — Release timeline and task allocation ## Complete Conversation Example @@ -29,10 +35,10 @@ You: We're releasing a new product next month and need to prepare all materials DesireCore: I'll help you coordinate this product release. I've broken it down into the following work: - 1. Legal Compliance Review -- Check product copy, user agreement - 2. Marketing Material Preparation -- Product introduction, promotional copy, social media content - 3. Financial Cost Accounting -- Release costs, pricing strategy - 4. Project Planning -- Release timeline, task allocation + 1. Legal Compliance Review — Check product copy, user agreement + 2. Marketing Material Preparation — Product introduction, promotional copy, social media content + 3. Financial Cost Accounting — Release costs, pricing strategy + 4. Project Planning — Release timeline, task allocation Delegated to respective agents to start work: → Legal Advisor: Reviewing product compliance @@ -95,11 +101,12 @@ DesireCore: Received emergency task, adjusting resource priority... ### Task Orchestration -DesireCore can automatically break down a complex requirement into multiple subtasks, identify dependencies between tasks, and arrange serial or parallel execution: +DesireCore can decompose a complex requirement into multiple subtasks, identify dependencies and arrange serial or parallel execution: - Tasks without dependencies execute in parallel to shorten total time - Tasks with dependencies are connected in series in sequence - Automatically monitor completion status of each task +- Bind deliverables, responsible parties and deadlines to each subtask ### Smart Scheduling @@ -108,6 +115,7 @@ When system resources are tight or unexpected situations occur, DesireCore will: - Reallocate resources based on task priority - Emergency tasks automatically preempt resources from low-priority tasks - When an agent fails, automatically reassign tasks to backup agents +- Centralize blockers for your review rather than scattering them across multiple conversations ### Intent Recognition @@ -116,12 +124,29 @@ DesireCore can distinguish whether you are **Teaching** or **Delegating Tasks**: - "Use a relaxed tone for emails in the future" → Teaching, records preference and syncs to all copywriting agents - "Help me write an email to the client" → Delegating, assigned to AI Copywriter for execution +### Decision Point Management + +Cross-department tasks most commonly get stuck at "no one to make the call." DesireCore will list items needing your decision separately: + +| Decision Point | Source | Impact | Next Step | +|---------------|--------|--------|-----------| +| Pricing: 299 or 399 | Finance Assistant | Affects margin and market positioning | After you confirm pricing, AI Copywriter updates promotional materials | +| Accept data processing clauses in user agreement | Legal Advisor | Affects launch compliance | After you confirm, sync with Product Manager to schedule launch | +| Delay release date | Product Manager | Affects marketing cadence and CS preparation | After you confirm, readjust task dependencies | + ## Final Deliverable -An efficiently operating multi-agent collaboration system that compresses cross-department coordination from days to tens of minutes, with all professional work advancing in parallel, cross-reviewed, and finally integrated into unified deliverables. +A trackable multi-agent collaboration workflow: tasks are decomposed, delegated, consolidated and cross-checked, ultimately forming unified deliverables, progress reports, blockers and pending decision lists. + +## DesireCore Capabilities Used + +- **Intelligent Task Orchestration**: Identify intent, decompose subtasks, delegate to agents, track progress and consolidate receipts +- **Team Mode / Swarm Mode**: Long-term roles handle stable processes, temporary roles handle one-off complex tasks +- **Decision Tree & Human Gate**: Key branches and high-risk actions require your confirmation before continuing +- **Replay & Audit**: Review what each agent did, which step took the most time and where optimization is needed :::tip Best Practices -- When describing requirements, state the global goal (e.g., "release new product"), let DesireCore determine which agents need to participate, which is more efficient than specifying individually -- For projects with clear deadlines, inform time constraints from the start, and the scheduler will prioritize tasks on the critical path -- Regularly check progress summaries and promptly respond to items requiring your decision to avoid blocking subsequent work of other agents -::: +- When describing requirements, state the overall goal (e.g., "release new product"), deadline, existing materials and non-negotiable constraints +- For projects with clear deadlines, inform time constraints from the start—the scheduler will prioritize tasks on the critical path +- Regularly check progress summaries and promptly respond to items requiring your decision to avoid blocking other agents' subsequent work +::: \ No newline at end of file diff --git a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/index.md b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/index.md index 40c4de3..70be92d 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/index.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/03-use-cases/index.md @@ -1,14 +1,16 @@ --- title: "Use Cases" -description: "See how DesireCore helps in real-world scenarios" -keywords: [Use Cases, Use Cases, DesireCore] +description: "Understand DesireCore's common use cases, input materials, and deliverables by real workflow" +keywords: [Use Cases, Scenarios, DesireCore, AI Agent] slug: /use-cases sidebar_position: 0 --- # Use Cases -DesireCore can handle various real-world work scenarios. Whether you need to process daily documents, analyze data, or tackle complex tasks in professional domains, you can find the right agent to help you. +This chapter doesn't introduce DesireCore by feature buttons, but organizes by real workflows: why a task occurs, what materials users already have, where the original process gets stuck, which steps the agent takes over, and what result counts as complete. + +If a page just says "AI helps you write documents, analyze data, handle emails," that's not yet a real scenario. A real scenario must answer: at what work node does this happen? Where do upstream materials come from? Who receives the deliverable? Which steps are most time-consuming? Which results need human confirmation? --- @@ -16,6 +18,47 @@ DesireCore can handle various real-world work scenarios. Whether you need to pro | Category | Description | Included Scenarios | |------|------|---------| -| [General Scenarios](./01-general/01-document-writing.md) | Daily work applicable to all industries | Document Writing, Data Analysis, Email Management, Meeting Assistant, Translation | -| [Professional Scenarios](./02-professional/01-legal-advisor.md) | Deep applications for specific industries | Legal Advisor, Code Assistant, Finance, Product Manager, Bidding, Real Estate, Data Scientist | -| [Industry Solutions](./03-solutions/01-ai-copyright.md) | End-to-end complete solutions | AI Copyright Registration, AI Bidding, Enterprise Service Coordination | +| [General Scenarios](./01-general/01-document-writing.md) | Daily office work and information processing, suitable for individuals or small teams to get started quickly | Document Writing, Data Analysis, Email Management, Meeting Assistant, Translation, Flowchart, Web Scraping | +| [Professional Scenarios](./02-professional/01-legal-advisor.md) | Auxiliary work requiring industry rules, team standards, or professional judgment | Contract Review, Code Development, Financial Audit, Product Management, Bidding, Real Estate Analysis, User Behavior Analysis | +| [Industry Solutions](./03-solutions/01-ai-copyright.md) | Long-process tasks completed by multiple agents collaborating | Software Copyright Materials, Bidding Collaboration, Enterprise Service Coordination | + +## How to Choose a Scenario + +| If you are | What you're dealing with | Recommended Entry | +|------------|--------------------------|-------------------| +| Project Manager / Product Manager | Writing PRDs, holding launch meetings, drawing business flows, tracking Sprint risks | [Document Writing](./01-general/01-document-writing.md), [Meeting Assistant](./01-general/04-meeting-assistant.md), [Flowchart Generation](./01-general/07-flowchart-agent.md), [Product Manager](./02-professional/04-product-manager.md) | +| Operations / Data Analyst | Explaining sales fluctuations, generating monthly reports, analyzing user churn | [Data Analysis](./01-general/02-data-analysis.md), [Data Analysis Reports](./01-general/06-data-analysis-agent.md), [Data Scientist](./02-professional/07-data-scientist.md) | +| Sales / Business / Bidding Lead | Handling client emails, writing proposals, interpreting tender documents, checking bidding materials | [Email Management](./01-general/03-email-management.md), [Document Writing](./01-general/01-document-writing.md), [Bidding Assistant](./02-professional/05-bidding-assistant.md), [AI Bidding](./03-solutions/02-ai-bidding.md) | +| Legal / Finance / Manager | Reviewing contract risks, auditing reimbursements, estimating budgets, coordinating cross-department tasks | [Legal Advisor](./02-professional/01-legal-advisor.md), [Finance Assistant](./02-professional/03-finance-assistant.md), [Enterprise Coordination](./03-solutions/03-enterprise-coordination.md) | +| Startup Team / Admin / IP Manager | Preparing software copyright materials, organizing product descriptions, completing screenshots and manuals | [AI Copyright Registration](./03-solutions/01-ai-copyright.md) | + +## How to Read a Scenario + +Each scenario follows the same workflow framework to understand: + +| Dimension | What to Look For | +|-----------|------------------| +| Trigger Point | Why this task must be done now, e.g., regular meeting, approval, bidding deadline, monthly report cycle | +| Existing Materials | Real inputs users bring, e.g., spreadsheets, contracts, meeting notes, templates, policies, historical projects | +| Pain Point | Where the original process is time-consuming or error-prone, e.g., inconsistent metrics, scattered materials, multi-person coordination, format checking | +| Intervention Method | Which steps DesireCore connects to, not replacing the entire role | +| Acceptance Result | What's delivered at the end, who decides if it can proceed to the next step | + +## DesireCore Capabilities in Use Cases + +DesireCore isn't just a single chat assistant—it's an Agent OS that "teaches rules, breaks down tasks, enables multi-agent collaboration, solidifies workflows, ensures controllable execution, and accumulates assets." Use cases should be understood through these capabilities: + +| Official Capability | How It Manifests in Scenarios | Typical Pages | +|---------------------|-------------------------------|---------------| +| Natural Language Teaching | Teach rules, templates, terminology, policies, and styles first; subsequent tasks follow the same standards | Document Writing, Translation Assistant, Legal Advisor, Finance Assistant, Bidding Assistant | +| Multi-Agent Collaboration | Split one task to different roles for parallel processing, then consolidate into a unified deliverable | AI Bidding, Enterprise Coordination, Data Analysis Reports | +| Workflow / SOP | Solidify repetitive processes into executable steps, suitable for weekly reports, inspections, bidding checks, compliance reviews | Data Analysis Reports, Web Scraping, Enterprise Coordination | +| Super Document | Document modifications are visible, controllable, and reversible; suitable for reports, tenders, contract review reports | Document Writing, AI Bidding, AI Copyright Registration | +| Email Center | Multi-email unified inbox, categorization and summaries, draft replies, confirmation before sending important emails | Email Management | +| Scheduled Inspection | Periodically check data sources or page changes, notify only when changes occur | Web Scraping, Data Analysis Reports | +| Code & Desktop Automation | Read/write local files, execute commands, operate GUI, screenshot recognition, save results to real workspace | Code Assistant, AI Copyright Registration, Web Scraping | +| Security Approval & Audit | Low-risk auto-execute, high-risk wait for confirmation, process traceable | Legal Advisor, Finance Assistant, Enterprise Coordination | + +:::tip Usage Suggestion +First teach your agent fixed standards, such as report templates, terminology glossaries, expense policies, code standards, or bidding qualification libraries. The more repetitive rules appear, the more worthwhile it is to consolidate them in advance—subsequent tasks will be more stable. +::: \ No newline at end of file