diff --git a/docs/zh/docs/dak/data-analysis/qa-quality.md b/docs/zh/docs/dak/data-analysis/qa-quality.md index 0bedb196..4703f050 100644 --- a/docs/zh/docs/dak/data-analysis/qa-quality.md +++ b/docs/zh/docs/dak/data-analysis/qa-quality.md @@ -4,7 +4,7 @@ d.run 支持统计用户对问答质量的反馈，并以报表的方式展示 ## 查看问答质量详情 -1. 在 **数据分析** 栏中点击 **问答质量** ,找到需要查看的问答,点击该问答记录可进入详情页面。 +1. 在 **数据分析** 栏中点击 **问答质量** ，找到需要查看的问答，点击该问答记录可进入详情页面。 ![详情菜单](images/qa-quality.jpg) diff --git a/docs/zh/docs/dak/data-analysis/shard-hit-ratio.md b/docs/zh/docs/dak/data-analysis/shard-hit-ratio.md index 5d5a0631..7f573e7c 100644 --- a/docs/zh/docs/dak/data-analysis/shard-hit-ratio.md +++ b/docs/zh/docs/dak/data-analysis/shard-hit-ratio.md @@ -4,11 +4,12 @@ d.run 统计了用户对于分片数据的使用情况，将分片按命中次 ## 查看分片命中率详情 -1. 在 **数据分析** 栏中点击 **分片命中率** ，通过 **搜索** 找到您关注的分片,点击该分片可进入详情页面，可查看该分片被引用的详细情况。 +1. 在 **数据分析** 栏中点击 **分片命中率** ，通过 **搜索** 找到您关注的分片，点击该分片可进入详情页面，可查看该分片被引用的详细情况。 ![选择详情](images/shard-hit-ratio.jpg) 2. 可以查看以下内容： + - 分片的基本信息：包括文本分片ID以及具体的分片内容。 - 详细信息：该分片被问题命中的问题内容、应用名称、提问人以及提问时间。 diff --git a/docs/zh/docs/dak/user-guide/app-center/create-app.md b/docs/zh/docs/dak/user-guide/app-center/create-app.md index cc52d59f..6928f43e 100644 --- a/docs/zh/docs/dak/user-guide/app-center/create-app.md +++ b/docs/zh/docs/dak/user-guide/app-center/create-app.md @@ -7,11 +7,7 @@ hide: 本页介绍如何在应用中心创建一个应用。 - - -### 创建 RAG 应用，具体步骤如下 +## 创建 RAG 应用 1. 在 **应用中心** 页面中，点击 **创建** 按钮，选择创建 RAG 应用 @@ -79,7 +75,7 @@ hide: ![发布应用](images/publish-app.jpg) -### 创建全文阅读应用，具体步骤如下 +## 创建全文阅读应用 1. 在 **应用中心** 页面中，点击 **创建** 按钮，选择创建 **全文阅读** 应用 @@ -109,7 +105,7 @@ hide: ![创建全文阅读应用](../../images/create-read.jpg) -### 创建组合应用，具体步骤如下 +## 创建组合应用 1. 在 **应用中心** 页面中，点击 **创建** 按钮，选择创建 **组合应用** @@ -146,4 +142,4 @@ hide: - 点击 **发布** 按钮，应用将会被发布到应用中心，用户可以在应用中心中找到并使用您的应用。 - ![创建组合应用](../../images/create-combined.jpg) \ No newline at end of file + ![创建组合应用](../../images/create-combined.jpg) diff --git a/docs/zh/docs/en/dak/data-analysis/number-of-qa.md b/docs/zh/docs/en/dak/data-analysis/number-of-qa.md new file mode 100644 index 00000000..00cc5232 --- /dev/null +++ b/docs/zh/docs/en/dak/data-analysis/number-of-qa.md @@ -0,0 +1,20 @@ +--- +hide: + - toc +--- + +# Number of Q&As + +d.run supports recording the number of Q&As created in this workspace. + +When users engage in Q&As, the records will be saved under the workspace where the app was created. Click the **Refresh** button in the upper right corner to view the latest Q&A count data. + +![Refresh](images/refresh-number-of-qa.png) + +You can summarize the Q&A counts of all users in this workspace into a table for export. + +1. On the **Number of Q&As** page, click the **Export** button in the upper right corner. + + + +2. The Q&A counts of all users in this workspace will be compiled into a xlsx file for download. diff --git a/docs/zh/docs/en/dak/data-analysis/qa-quality.md b/docs/zh/docs/en/dak/data-analysis/qa-quality.md new file mode 100644 index 00000000..dd5073de --- /dev/null +++ b/docs/zh/docs/en/dak/data-analysis/qa-quality.md @@ -0,0 +1,33 @@ +# Q&A Quality + +d.run supports statistics on user feedback regarding the quality of Q&A, displaying it in reports and also allows data to be exported to xlsx format. + +## View Q&A Quality Details + +1. In the **Data Analysis** section, click **Q&A Quality** to find the Q&A you wish to view. Click the Q&A record to enter the details page. + + + +2. You can view the following content: + + - User's question, including the name of the user who asked. + - App's answer: the answer generated by the app in response to the user's question. + - The prompt used for the question. + + + +3. When new Q&As are generated, you can click the **Refresh** button in the upper right corner to view the latest Q&A details. + + + +## Export Q&A Quality + +You can summarize all user feedback on Q&A quality within this workspace into a table for export. + +1. On the **Q&A Quality** page, click the **Export** button in the upper right corner. + + + +2. Click **Export Records** to view the exported Q&A quality records. Select the records you need to download and export them as an Excel file. + + diff --git a/docs/zh/docs/en/dak/data-analysis/shard-hit-ratio.md b/docs/zh/docs/en/dak/data-analysis/shard-hit-ratio.md new file mode 100644 index 00000000..d39470c1 --- /dev/null +++ b/docs/zh/docs/en/dak/data-analysis/shard-hit-ratio.md @@ -0,0 +1,32 @@ +# Shard Hit Ratio + +d.run tracks user interactions with shard data, ranking shards by hit count. It supports viewing shard content and exporting shard usage statistics to a xlsx file. + +## View Shard Hit Ratio Details + +1. In the **Data Analysis** section, click **Shard Hit Ratio**. Use the **Search** function to find the shard you are interested in. Click the shard to enter the details page, where you can view detailed information about how this shard has been referenced. + + + +2. You can view the following content: + + - Basic information about the shard: including the text shard ID and specific shard content. + - Detailed information: the questions that hit this shard, app name, questioner, and the time the question was asked. + + + +3. When new Q&As are generated, you can click the **Refresh** button in the upper right corner to view the latest shard hit ratio results. + + + + You can also **Refresh** in the shard hit ratio details to see the latest results. + + + +## Export Shard Hit Ratio + +1. On the **Shard Hit Ratio** page, click the **Export** button in the upper right corner. + + + +2. This will export the shards created in this workspace and compile feedback on these shard usage statistics into a xlsx file for download. diff --git a/docs/zh/docs/en/dak/data-analysis/shard-quality.md b/docs/zh/docs/en/dak/data-analysis/shard-quality.md new file mode 100644 index 00000000..e374374f --- /dev/null +++ b/docs/zh/docs/en/dak/data-analysis/shard-quality.md @@ -0,0 +1,27 @@ +--- +hide: + - toc +--- + +# Shard Quality + +A shard refers to dividing larger data into multiple segments, which facilitates model training by making it easier to locate and hit the data. +d.run supports viewing the quality of shards. The specific steps are as follows: + +1. In the **Data Analysis** section, click **Shard Quality**, and use the **Search** function to find the shard you are interested in. Click the shard to enter the details page, where you can view detailed information about this shard. + + + +2. You can view the following content: + + - Corpus: Which corpus the shard belongs to. + - Update Time: The last update time of the shard file. + - Shard ID: The unique identification code of the shard. + - Shard Content: The specific content of the shard after slicing. + - Additional Information: Additional content related to this shard. + + + +3. When new shard files are evaluated, you can click the **Refresh** button in the upper right corner to view the latest shard files. + + diff --git a/docs/zh/docs/en/dak/data-analysis/suggestions.md b/docs/zh/docs/en/dak/data-analysis/suggestions.md new file mode 100644 index 00000000..f849f69f --- /dev/null +++ b/docs/zh/docs/en/dak/data-analysis/suggestions.md @@ -0,0 +1,44 @@ +# Suggestions + +d.run provides an entry for users to submit feedback, allowing administrators to see what issues need improvement in the apps within this workspace. + +## Handling Feedback + +1. In the **Data Analysis** section, click **Improvement Suggestions**, find the feedback that needs to be addressed, and click the **┇** button on the right. + +2. In the pop-up menu, select **Handle Feedback** (clicking the ticket information within the range will take you to the details page). + + + +3. You can view the following content: + + - User feedback on the Q&A. + - Relevant information about the Q&A: assistant name, model name, number of citations, processing time, and Q&A details. + + + + - Latest Citations: You can view the details of all citations referenced in the answer. + + + + - Details of ticket processing. + +4. Click the status of the ticket marked as **Pending** to change the status to **In Progress** or **Completed**, and fill in the **Processing Description** below. + +5. Click the **OK** button in the lower right corner to modify the feedback processing status. + + + +6. When new improvement suggestions are submitted, you can click the **Refresh** button in the upper right corner to view the latest suggestions. + + + +## Export Feedback + +Aggregate all Q&A details and feedback into a table for export. + +1. Click the **left side of Q&A feedback**, select the feedback you want to export, and click the **Export** button in the upper right corner. + +2. The Q&A content will be compiled into a xlsx file for download. + + diff --git a/docs/zh/docs/en/dak/features.md b/docs/zh/docs/en/dak/features.md index fe2e329d..04b1567f 100644 --- a/docs/zh/docs/en/dak/features.md +++ b/docs/zh/docs/en/dak/features.md @@ -9,19 +9,19 @@ The features of the Intelligent Q&A are summarized in the table below: | Module | Features | Description | | --- | ---- | --- | -| Conversation | Context-Aware Intelligent Q&A | Connects application corpus to achieve context-aware intelligent Q&A functionality | +| Conversation | Context-Aware Intelligent Q&A | Connects app corpus to achieve context-aware intelligent Q&A functionality | | | User Interaction Optimization | Supports users to **evaluate, copy, regenerate, delete conversations, or submit feedback** on AI responses, enhancing user experience | -| Application Center | Application Lifecycle Management | Provides full lifecycle management from **creation, deployment, monitoring to maintenance** | -| | Environment Isolation | Supports binding or unbinding of workspaces to achieve **environment isolation**, ensuring application security | -| | Model Service Integration | Integrates **GLM, Llama, and other model services**, enhancing the efficiency of RAG (Retrieval-Augmented Generation) applications | +| App Center | App Lifecycle Management | Provides full lifecycle management from **creation, deployment, monitoring to maintenance** | +| | Environment Isolation | Supports binding or unbinding of workspaces to achieve **environment isolation**, ensuring app security | +| | Model Service Integration | Integrates **GLM, Llama, and other model services**, enhancing the efficiency of RAG (Retrieval-Augmented Generation) apps | | | Diverse Configuration Options | Offers options for AI configuration, corpus association, retrieval strategies, etc., to optimize AI response quality | -| | Application Key Support | Supports generation of **application keys** for secure access to OpenAPI conversations | -| | Application Link Sharing | Allows **link sharing of applications** and supports H5 page display for mobile use | +| | App Key Support | Supports generation of **app keys** for secure access to OpenAPI conversations | +| | App Link Sharing | Allows **link sharing of apps** and supports H5 page display for mobile use | | | Conversation Process Viewing | Supports viewing the complete **conversation process**, enhancing user experience and troubleshooting efficiency | -| | Combination Application Support | Supports **combination applications** for intent recognition, enhancing the ability to handle complex scenarios | +| | Combination App Support | Supports **combination apps** for intent recognition, enhancing the ability to handle complex scenarios | | | Language Translation Support | Supports **translation** of user input content, improving the accuracy of corpus matching | | | Multilingual Response Selection | Allows users to choose the **language** of AI responses, catering to multilingual scenarios | -| | Full Text Reading Application Support | New full text reading application, supports displaying **image content** | +| | Full Text Reading App Support | New full text reading app, supports displaying **image content** | | | Support for Numbers and Chemical Formulas | Supports the display of **mathematical formulas and chemical equations**, enhancing the user experience in academic and research contexts | | Corpus Management | Multiple Import Methods | Supports various methods for managing the corpus, including **standard import, formatted import, manual import, and image-text import** | | | Intelligent Sharding | Supports **automatic sharding** by delimiter or size, with customizable sharding logic through plugins | diff --git a/docs/zh/docs/en/dak/index.md b/docs/zh/docs/en/dak/index.md index 2215894b..1dc4de03 100644 --- a/docs/zh/docs/en/dak/index.md +++ b/docs/zh/docs/en/dak/index.md @@ -8,8 +8,8 @@ hide: The d.run intelligent Q&A product utilizes RAG technology to vectorize user data, and then combines a large language model with similarity search to query the user data repository, achieving knowledge-based Q&A. This method addresses the issues of user privacy data protection and model hallucination, resulting in more accurate and personalized responses. -Here, you can create applications and plugins to build a corpus for AI conversations, and then optimize user experience through data analysis feedback. +Here, you can create apps and plugins to build a corpus for AI conversations, and then optimize user experience through data analysis feedback. -![Application Center](./images/app-center.jpg) +![App Center](./images/app-center.jpg) [Register and Try d.run](https://console.d.run/){ .md-button .md-button--primary } diff --git a/docs/zh/docs/en/dak/user-guide/app-center/conversation.md b/docs/zh/docs/en/dak/user-guide/app-center/conversation.md new file mode 100644 index 00000000..36a88d00 --- /dev/null +++ b/docs/zh/docs/en/dak/user-guide/app-center/conversation.md @@ -0,0 +1,73 @@ +--- +hide: + - toc +--- + +# App Dialogue Instructions + +**Dialogue** is the most common way to obtain information since the emergence of ChatGPT. In d.run, once an app is published, the dialogue function can be used. You can freely ask questions in the dialogue after linking to a corpus, consult historical records at any time, and the navigation bar also lists the recently used dialogues. + +## Daily Dialogue + +### RAG App + +1. In the left navigation bar, click on **App Center**, select the published RAG type app, and click the dialogue icon in the lower right corner. + + + +2. Enter your question in the dialogue box, click **Send**, or press the Enter key to start the conversation. + + + +### Full Text Reading App + +1. In the left navigation bar, click on **App Center**, select the published **Full Text Reading** type app, and click the dialogue icon in the lower right corner. + + + +2. Upload a file in the lower right corner of the dialogue box and ask questions based on the content of the file. Click **Send**, or press the Enter key to start the conversation. + + + +### Combination App + +1. In the left navigation bar, click on **App Center**, select the published **Combination App** type app, and click the dialogue icon in the lower right corner. + + + +2. Enter your question in the dialogue box, click **Send**, or press the Enter key to start the conversation. + + + +## Dialogue Management + + + +- **Pin**, **Rename**, and **Delete**: In the history pane, click the **┇** on the right side of a dialogue to pin, rename, or delete the dialogue. +- **Clear**: In the upper left corner of the dialogue page, click the 🧹 icon to **clear** the dialogue. +- **Link Corpus**: In the lower left corner of the input box, click the 📖 icon to link the corpus. + + Selection of corpus: The **chat app** will match the most similar corpus in the selected corpus as a reference to answer your questions. A successful operation will indicate that the corpus has been selected. You can also deselect one or more corpora. + + !!! info + + If you cannot select or change the linked corpus, you can only add a corpus. The vectorization model of the app must be consistent with the vectorization model of the corpus to be usable. + +- **Stop Dialogue**: After asking a question, click the icon on the right side of the input box to stop the dialogue, allowing the assistant to cease outputting content. + +## Some Useful Icons + + + +- **Rating**: You can like 👍 or dislike 👎 a response, depending on your satisfaction with the content of the answer. +- **Copy**: You can copy a response. +- **Reset**: You can reset or regenerate a specific response. + + !!! info "Randomness" + + For the reset content, the administrator can adjust the model's **randomness** to control the consistency of the assistant's responses over multiple attempts. + + If the randomness is high, the assistant's responses may vary each time. If high accuracy is required, the randomness can be lowered, resulting in more consistent outputs from the assistant each time. + +- **Delete**: Click the trash can :material-delete: icon to delete a response, after which the deleted dialogue will not be remembered in the context of the assistant's conversation. +- **Feedback**: Click the last icon :bookmark_tabs: of a response to submit feedback, providing comments based on the quality of the assistant's responses. diff --git a/docs/zh/docs/en/dak/user-guide/app-center/create-app.md b/docs/zh/docs/en/dak/user-guide/app-center/create-app.md new file mode 100644 index 00000000..f2ee2586 --- /dev/null +++ b/docs/zh/docs/en/dak/user-guide/app-center/create-app.md @@ -0,0 +1,144 @@ +--- +hide: + - toc +--- + +# Creating an App + +This page explains how to create an app in the App Center. + +## Creating a RAG App + +1. In the **App Center** page, click the **Create** button and select to create a RAG app. + + + +2. Fill in the basic information for the app. + + - **Upload App Icon**: Choose an image in jpg, jpeg, or png format, ensuring the file size does not exceed 10MB. + - **Enter App Name**: Limit to 20 characters. + - **Fill in App Description**: Briefly describe the app, not exceeding 100 characters. + - **App Debugging**: Choose to enable debugging, which will show the processing of the dialogue in the dialogue interface. + +3. AI Configuration + + - **Select Large Language Model Service**: Decide whether to use a local model service or an online model service (e.g., Azure Open AI or Wenxin Yiyan). + - **Set Randomness**: Adjust the creativity/divergence level of the model's responses. + - **Repetition Penalty**: Reduce the repetition of content in the generated text, with a default value of 1; the larger the value, the less likely it is to generate repeated content. + - **Vectorization Model**: Choose the vectorization model and vectorization model service for the app. + - **System Prompt**: Specify the role of the model, its capabilities, and some rules you want it to follow when conversing with you or assisting you in generating content. + - **Template for Matching Corpus**: Provide templates for the model to retrieve and answer questions. + - **Template for Non-Matching Corpus**: Provide general prompts for the model when similar content cannot be retrieved. + + !!! note + + - Template for Matching Corpus: When similar content is retrieved, the corpus prompt template will be used. The template includes **knowledge block {corpus_search_content}** and **user input {corpus_search_content}**. These variables will be replaced with the corresponding text and sent to the large model for Q&A. + + ```template + Create a knowledge Q&A game: + 1. The response content must be within "{corpus_search_content}". + 2. If the question cannot be found in the provided information, you will respond: "Sorry, the information in the database does not contain the result you need, please try modifying the referenced database or use the model's own capabilities to answer." + {user_inputs_content} + ``` + + - Template for Non-Matching Corpus: When similar content cannot be retrieved, an embedded prompt template will be used. The embedded prompt is concatenated before the user's question, serving as a general convention to guide the app model in generating responses. + +4. Link Corpus + + - **Select Vectorization Model**: Determine the model used for question vectorization. + - **Select Corpus**: Choose one or more existing corpora for model retrieval. + +5. Set Retrieval Strategy + - **Retrieval Preprocessing**: Support custom plugins for Q&A preprocessing. + - **Number of Knowledge Blocks Retrieved**: Decide the number of knowledge blocks provided to the model. + - **Similarity**: Set the strictness of knowledge block matching. + - **Reordering**: Enable or disable the reordering model to improve result ranking. + - **Only Corpus Responses**: When enabled, the model will only respond according to the matched corpus content. + - **Image and Text Mode**: Control whether and how to output image and text content. + - **Maximum Number of Image and Text Pieces**: Determine how many images can be output in one dialogue. + - **Image and Text Similarity**: Set the strictness of image matching. + +6. Set Memory Strategy + + - **Recent Chat Memory Rounds**: Set the number of rounds or turns of dialogue history. + +7. Save and Publish App + + - Click the **Save** button; the app settings will be completed and saved as a draft. + + !!! note + + Please ensure that you fill out and configure all aspects of the app carefully according to the above steps to ensure that the app can accurately and efficiently respond to user needs. + + - Click the **Publish** button; the app will be published to the App Center, where users can find and use your app. + + + +## Creating a Full Text Reading App + +1. In the **App Center** page, click the **Create** button and select to create a **Full Text Reading** app. + + ![App Center](../../images/app-center.jpg) + +2. Fill in the basic information for the app. + + - **Upload App Icon**: Choose an image in jpg, jpeg, or png format, ensuring the file size does not exceed 10MB. + - **Enter App Name**: Limit to 20 characters. + - **Fill in App Description**: Briefly describe the app, not exceeding 100 characters. + +3. AI Configuration + + - **Select Large Language Model Service**: Decide whether to use a local model service or an online model service (e.g., Azure Open AI or Wenxin Yiyan). + - **Set Randomness**: Adjust the creativity/divergence level of the model's responses. + - **Prompt**: Provide templates for the model to retrieve and answer questions. + +4. Save and Publish App + + - Click the **Save** button; the app settings will be completed and saved as a draft. + + !!! note + + Please ensure that you fill out and configure all aspects of the app carefully according to the above steps to ensure that the app can accurately and efficiently respond to user needs. + + - Click the **Publish** button; the app will be published to the App Center, where users can find and use your app. + + + +## Creating a Combination App + +1. In the **App Center** page, click the **Create** button and select to create a **Combination App**. + + + +2. Fill in the basic information for the app. + + - **Upload App Icon**: Choose an image in jpg, jpeg, or png format, ensuring the file size does not exceed 10MB. + - **Enter App Name**: Limit to 20 characters. + - **Fill in App Description**: Briefly describe the app, not exceeding 100 characters. + - **App Debugging**: Choose to enable debugging, which will show the processing of the dialogue in the dialogue interface. + +3. AI Configuration + + - **Select Large Language Model Service**: Decide whether to use a local model service or an online model service (e.g., Azure Open AI or Wenxin Yiyan). + - **App Prompt**: Provide templates for the model to retrieve and answer questions. + + !!! note + + The prompts need to include different app IDs, which can be copied from the apps under app configuration. + +4. App Configuration + + - **Select App**: Choose an app from the App Center that can engage in dialogue normally. + - **Copy ID**: Used to fill in the prompts for selecting different apps during model dialogue. + +5. Save and Publish App + + - Click the **Save** button; the app settings will be completed and saved as a draft. + + !!! note + + Please ensure that you fill out and configure all aspects of the app carefully according to the above steps to ensure that the app can accurately and efficiently respond to user needs. + + - Click the **Publish** button; the app will be published to the App Center, where users can find and use your app. + + diff --git a/docs/zh/docs/en/dak/user-guide/system-config.md b/docs/zh/docs/en/dak/user-guide/system-config.md new file mode 100644 index 00000000..2ee399e9 --- /dev/null +++ b/docs/zh/docs/en/dak/user-guide/system-config.md @@ -0,0 +1,22 @@ +# System Configuration + +!!! note + + The parameters in the system configuration will serve as default parameters when creating apps. + + + +## Conversation Memory Rounds + +This refers to the number of rounds or turns used to track and manage conversation history in a dialogue system. Each time a user interacts with the system, the conversation round count increases by one. Here, it indicates that when a user interacts with the chat app, it will retain a portion of the conversation history in a small memory window for a certain period, with the size of the memory rounds determining the duration of retention. + +- Default value is 10 +- Minimum value is 0 + +## Corpus Prompt Template + +This template is activated when the app is associated with a template. The template content includes knowledge blocks obtained through similarity search, user inputs, and embedded prompt templates. + +## Embedded Prompt Template + +The embedded prompts are concatenated before the user's question as a general convention to guide the app model in generating responses to the question. diff --git a/docs/zh/docs/en/models/apikey.md b/docs/zh/docs/en/models/apikey.md index 74082450..258cf23d 100644 --- a/docs/zh/docs/en/models/apikey.md +++ b/docs/zh/docs/en/models/apikey.md @@ -18,7 +18,7 @@ API Key is the core credential for calling model services, used to verify user i 1. On the **API Key Management** page, click the **Create** button at the top right corner. 2. In the pop-up window, fill in the name of the API Key (e.g., test-key) to identify its purpose or associated project. -3. Click **Confirm**, and the system will generate a new API Key. +3. Click **OK**, and the system will generate a new API Key. !!! note