AI Conversation Knowledge Center User Manual (Version: v1.6)

A One-Stop Guide to Knowledge Management and Intelligent Interaction

I. Product Overview

1.0 Product Introduction

The AI Conversation Knowledge Center is an enterprise-grade knowledge management platform built on fine-tuned large language models and semantic understanding technology, integrating knowledge management capabilities with AI large models to empower business operations.

Core Features

• Intelligent Interaction: Supports knowledge Q&A, role switching (product selection/technical consultation, etc.), and simulated examiner functionality
• Multimodal Data Processing: Integrates diverse data formats including documents, archives, records, scanned files, and spreadsheets
• Core Technical Support: Powered by large language models, knowledge extraction, knowledge linking, NLP, OCR, and more
• Distinctive Features: Intelligent document management, digital avatar training, multilingual support, cross-modal retrieval
• Core Value: Transforms passive reading into active interaction, delivering efficient knowledge management and application solutions for enterprises

Product Goals

By introducing intelligent technologies and reliable knowledge management mechanisms, this platform resolves inefficiencies, inaccuracies, and outdated information commonly encountered during enterprise knowledge transmission, retrieval, and application, empowering all business scenarios.

II. Login Guide

2.0 Login Steps

1. Access the system: Open a browser and enter the official login URL in the address bar: https://stg-esm.sayco.com.cn/
2. Enter credentials: On the login page, input your personal login information (phone number/email/account) and corresponding password
3. Complete login: After verifying your information, click the "Login" button to be redirected to the Knowledge Center homepage

2.1 CAPTCHA Verification

• Verification Method: Uses a drag-and-slide puzzle mechanism for human-machine verification
• Verification Value: Intuitive and simple operation that effectively blocks automated bots, ensures account security, and prevents unauthorized access

Notes

1. Account Lockout Policy: After five consecutive failed password attempts, your account will be temporarily locked; you may retry after unlocking
2. Agreement Declaration: Logging in or registering constitutes acceptance of the User Agreement and Privacy Policy (Version V2.0.0); click the link at the bottom of the login page for details
3. Troubleshooting:
   • If "Website unreachable" is displayed, check your network connection
   • If "Account not activated" is displayed, contact your enterprise administrator to activate your account

III. User-Side Core Operations

3.0 Intelligent Search

Entry Point

The homepage features an intelligent search bar, multilingual toggle, latest documents section, and full document directory

Workflow

1. Log in: Use your account and password to log in; the system automatically redirects you to the Knowledge Center homepage, or navigate via the main menu
2. Input query: Enter keywords into the search bar using natural language (e.g., "Find me the latest sales report")
3. Retrieve results: Leveraging the enterprise knowledge base, the AI search assistant responds instantly with 24/7 automated support, extracting relevant answers for both FAQ and technical queries

3.1 All Documents

Feature Description

Displays all document resources on the platform, categorized into major sections such as Product Documentation, IT Operations, Professional Standards, and General Files, with support for multi-level folder hierarchies

Operation Method

Click "All Documents" to view the complete document directory. Each document includes a brief description (one line max) and allows further exploration via click-through

Document Types

Covers enterprise-relevant documents including Docker fundamentals, Alibaba Cloud services, CRM systems, digital management solutions, and more; supports multiple file formats

3.2 Latest Documents

Feature Description

A dedicated area displaying recently updated documents on the platform

Core Value

Immediately surfaces newly uploaded or modified documents, ensuring users do not miss critical updates

Operation Method

Click on any file listed under "Latest Documents" to directly view its content

3.3 AI Assistant

Function Overview

Built on fine-tuned large language models and semantic understanding, this highly anthropomorphic AI assistant supports input contexts up to 1,000+ tokens and delivers intelligent Q&A services

Distinctive Features

• Role Switching: Supports switching between multiple roles including Default Assistant, Training Examiner, Onboarding Guide, and Data Security Advisor
• Scenario Adaptation: Select appropriate roles for different contexts such as simulated exams, product selection, or technical consultation

Workflow

1. Access the AI Q&A interface
2. Input your question (supports long-context input)
3. Click the role-switch button to select the desired role
4. Engage in natural conversational interaction with the AI assistant

3.4 Digital Avatar Training

Feature Description

Creates a highly anthropomorphic AI training mentor that answers questions in real time, evaluates learning outcomes, simulates multiple training scenarios, and delivers skill instruction and knowledge explanation through natural dialogue

Workflow

1. Open the target document
2. Click "Digital Avatar Tutor"
3. Begin usage (complete the full learning-to-application process in three steps)

Application Scenarios & User Value

1. Enterprise Training: Standardizes interpretation of policies and manuals, ensures consistent information delivery, and reduces repetitive training costs
2. Educational Learning: Converts textbooks, research papers, and other PDFs into "virtual lecturers" for engaging, immersive instruction suited to self-paced learning
3. Cross-Language Scenarios: Enables "term-accurate mapping + logic visualization" for foreign-language technical documents, helping international and cross-border teams rapidly understand contracts and technical materials

3.5 Continue Searching

Feature Description

After completing a search, the system generates three suggested follow-up questions to guide deeper exploration of related knowledge

Workflow

1. Review the initial search results along with the system-generated recommended questions
2. Click any suggested question to continue, or input a new query
3. To switch knowledge bases, click the "All Knowledge Bases" dropdown menu and select the desired source; the current knowledge base is used by default

Functional Value

Encourages deeper thematic exploration, reveals broader and deeper knowledge connections, and inspires discovery of previously unconsidered related content

IV. Admin-Side Core Operations

4.1 Default Knowledge Base

Feature Overview

Enables creation, editing, publishing, permission control, and agent association of knowledge bases; supports multimodal data import (documents/audio/video)

Workflow

1. Create knowledge base: Click "Add New," fill in the required knowledge base name in the pop-up window, and optionally provide metadata (identifier, default language, description, icon); icon must be square (1:1 aspect ratio), minimum width 100px, optimal resolution 400px × 400px
2. Build directory: Within the created knowledge base, click "Create Folder" to establish multi-level folder hierarchies for knowledge categorization
3. Upload files: Upload relevant files into designated folders; supported formats include pdf, txt, md, docx, xlsx, pptx, mp4, mov, jpeg, jpg, png, zip; individual file size must not exceed 120MB
4. Publish knowledge base: Navigate to the target knowledge base’s detail page and click the "Publish" button in the top-right corner to transition from draft to live status

Management Rules

• Recommend retaining only the latest version of each file; new versions overwrite historical ones
• Supports batch upload, folder relocation, download, and relearning
• Enables real-time refresh to synchronize the latest file status within the knowledge base

4.2 AI Agents

Feature Overview

Supports scenario configuration, role assignment, orchestration management, and audit logging to enable 7×24 intelligent responses and complex problem decomposition

Workflow

1. Access the Agent Management module
2. Create role: Click "Create Role," configure role name, type, language, identifier, and associated knowledge base
3. Scenario orchestration: After setting basic role info, proceed to scenario orchestration; predefine AI model prompts for different scenarios (default and specific business contexts), specifying behavioral logic, language style, and knowledge boundaries
4. Optimize settings: During use, return to the role configuration interface at any time to refine prompts; save changes after editing

Configuration Options

• Knowledge Base Query: Set similarity score (default: 0.60), number of references (default: 20), and keyword weight (default: 0.70)
• Response Rules: Enable "Answer Based on Knowledge Base" (respond only if answer exists in KB; reject otherwise), "Use Knowledge Graph" (enhance multi-hop reasoning), and "Full Text Matching" (reference entire document content)
• Model Selection: Supports configuration of different LLMs (e.g., Qwen3-32B-vllm)

4.3 Log Records

Feature Overview

Records and manages user interaction data such as searches and Q&A within the Knowledge Center, enabling administrators to monitor system usage and optimize knowledge bases and AI models

Logged Data

• Basic Info: Querying user, timestamp, language used, AI engine, search duration, start time, token count, log ID
• Core Content: User query and system’s detailed Q&A response

Operational Functions

1. Search: Enter keywords in the left-side "Search by Keyword" box to quickly locate specific log entries
2. View Details: Click any log entry in the left list to display its full Q&A content on the right
3. Copy: Click the "Copy" button in the right Q&A panel to copy content for data analysis, documentation, or reporting

V. Common Use Cases

5.1 New Employee Onboarding Training

Traditional Training Pain Points

• Overwhelming Materials: Paper or digital manuals contain excessive information, making key content hard to locate
• Inefficient Access: Lacks effective search tools, resulting in time-consuming and poor user experience
• Limited Mentor Capacity: Trainers must repeatedly answer common questions, leaving little room for personalized guidance
• Delayed Clarification: Questions go unanswered for long periods, hindering learning progress

AIGC Intelligent Q&A Center Solution

1. Build knowledge base: Upload training manuals to create comprehensive knowledge reserves
2. Digital avatar training: Generate digital mentors to demonstrate procedures and explain corporate culture
3. Intelligent agent Q&A: Enable real-time conversational interactions for fast, accurate answers

Benefits

• Convenient & Efficient: Access information anytime, anywhere, and self-pace learning
• Personalized Experience: Tailored responses accommodating individual needs
• Reduced Mentor Burden: Minimizes repetitive tasks, freeing trainers for complex issues

5.2 Pre-Sales Team Product Knowledge Updates

Traditional Management Pain Points

• Low Efficiency: Manually sifting through vast volumes of documents is time-consuming and labor-intensive
• Accuracy Risks: Manual searches risk missing critical parameters, leading to misinformation
• Outdated Knowledge: After product launches or specification changes, information synchronization lags, impacting communication

AIGC Intelligent Q&A Center Implementation

1. Structured Knowledge Entry: Organize collected materials by product line, feature module, and use case; standardize and structurally decompose for precise knowledge base construction
2. Intelligent Retrieval: Search using keywords, model numbers, etc., for fast, accurate results—no need for manual bulk searching

Core Value

• For Pre-Sales Teams: Enhances retrieval efficiency, ensures accuracy, lowers onboarding costs
• For Customers: Delivers timely, precise product information, improving communication quality and perception

5.3 Sales Representatives Mobile Workflow

Traditional Workflow Pain Points

• Difficult Information Access: Mobile users struggle with typing and retrieving information on the go
• Diverse Data Formats: Internal materials in scanned PDFs, videos, audio, and images are hard to organize and retrieve
• Language Barriers: Queries in Japanese or English documents pose comprehension challenges
• Fragmented Information: Data scattered across internet sources and cloud drives is difficult to consolidate and analyze for industry insights

AIGC Intelligent Q&A Center Solution

• Multimodal Retrieval: Enables rapid search across all formats without manual preprocessing
• Streamlined Interaction: Optimized for mobile use, minimizing text input
• Cross-Language Support: Eliminates language barriers in foreign-language documentation
• Information Aggregation: Consolidates data across sources and surfaces critical content instantly

VI. Frequently Asked Questions

6.1 Common Questions & Answers

Q1: Login fails with "Account does not exist"? A: Contact your administrator to confirm if your account has been properly activated.

Q2: Cannot find target knowledge? A: Verify the accuracy of your search keywords or confirm whether the knowledge has been uploaded by the administrator.

Q3: AI response is inaccurate? A: Use the "Feedback" function on the page to submit your issue; administrators will optimize the AI model based on feedback.

Feedback Suggestions

1. Rating Function: Rate AI responses using upvote/downvote to help continuously improve the model
2. Suggestions: Any feedback regarding product features or UI design can be submitted through administrator channels

@ 2023 Sayco All Right Reserved Lu ICP No. 2023002838-8