IMA knowledge base

What is the IMA knowledge base?

IMA is an AI knowledge management tool developed by Tencent Technology; it is a integrated knowledge workspace that enables searching, reading, and writing functions, all centered around an individual’s profile. In its early stages, this product was often referred to as the “IMA Knowledge Database” or “Tencent AI Workspace”, while its current official name is “IMA – Tencent AI Knowledge Manager”.

It integrates data collection, knowledge base queries, note-taking, audio recording, and the handling of complex tasks under a single account, making it suitable for learning and research, organizing work-related materials, creating content, and sharing team experience.

Unlike traditional cloud storage services that only store files, ima is capable of analyzing documents, web pages, images, and audio, allowing users to ask questions about the content, generate summaries, and continue writing. Compared to ordinary chatbots, it places more emphasis on the user’s own sources of knowledge, the organization of that knowledge, and its accumulation over time.

The new version of Copilot can also remember usage preferences and use various skills to carry out multi-step tasks, but it is still necessary to verify important conclusions by referring to the original sources.

Core functions

1. Personal knowledge base

Users can add local files, WeChat files, articles from WeChat official accounts, web pages, images, and audio content to their personal knowledge base, creating a space for continuous accumulation of information. Once imported, this data can be searched, read, summarized, and used for Q&A; it is suitable for organizing course materials, papers, industry reports, project documents, meeting notes, and personal collections.

The value of a knowledge base lies not only in storing information, but also in enabling the retrieval, linking, and reuse of disparate data. Users can pose questions related to a specific knowledge base, which helps to reduce the likelihood that generic online answers deviate from the existing materials.

However, scanned documents may still fail to interpret elements such as complex tables, mathematical formulas, footnotes, and small text in images completely.

2. Shared knowledge base and knowledge base plaza

IMA allows knowledge bases to be shared with others, enabling course materials, project experiences, and themed collections to be exchanged among members. Publicly available content can also be found on the Knowledge Base Plaza, where users can discover and access high-quality knowledge bases from various fields.

Before sharing, it is necessary to check the scope of access, file copyrights and privacy information; internal policies, customer materials, and unpublished research should not be made publicly available.

The content in the Knowledge Base Plaza is maintained by various creators, and therefore the timing of updates, as well as the completeness and quality of that content, vary. Joining someone else’s knowledge base does not grant ownership rights to the content, nor does it mean that the material has been reviewed by Tencent or any professional organization.

3. Copilot Knowledge Partner

The new version of IMA uses Copilot as its dedicated knowledge partner. It is capable of understanding objectives, dealing with complex problems, remembering certain habits and preferences, and utilizing a knowledge base, online information, and various tools to complete tasks.

Compared to one-time Q&A, Copilot is better suited for breaking down a research topic into multiple steps, and then generating reports, organizing results, or other deliverables.

Recent versions now support processing PDF and DOC files, creating podcasts, and MCP connections; some high-cost skills require significant computing resources. Skills or MCPs may access external services, and it is necessary to check the scope of data before granting permissions.

Operations related to sending, publishing, purchasing, deleting, or modifying accounts must be ultimately confirmed by the user.

4. Q&A on information and graphic explanations

IMA can provide summaries, explanations, and sequential questions for the imported documents, incorporating relevant content in its answers. For document-based queries, the system also automatically matches relevant charts to make longer texts easier to understand.

The answers derived from the knowledge base are not necessarily accurate; the citation locations, numbers, dates, and conclusions should all be checked against the original text.

5. Notes and AI writing

When reading material, one can extract the key points and put them in notes; thereafter, the AI can be used to generate text, expand on it, refine it, or adjust its wording. The new version allows notes to be exported in Markdown format, enables the creation of bidirectional links between notes, and also permits Markdown files to be converted into notes.

For users who need to move from collecting information to writing formally, this \"read–extract–organize–create\" process is more coherent than repeatedly copying content into different applications.

AI-generated text may include facts or citations that are not present in the original materials. Papers, contracts, medical documents, legal texts, financial statements, and promotional materials must be reviewed manually; it is not acceptable to skip this review just because the information comes from an individual’s knowledge base.

6. Audio transcript

The mobile version allows recording for up to 2 hours; it can recognize multiple languages and automatically generate transcriptions of the original text as well as meeting summaries. It is suitable for use in classrooms, interviews, meetings, and for capturing ideas.

It is necessary to obtain the required authorization before recording someone else’s speech; multiple people speaking at the same time, accents, noise, and technical terms can all affect the accuracy of the transcription. Names, numbers, and deadline dates in particular need to be verified carefully.

7. AI interpretation, podcasts, and knowledge quizzes

After the material is uploaded, IMA can generate mind maps, audio podcasts with a realistic sound quality, and knowledge quizzes, thus converting the same material into different forms suitable for learning. Podcasts are ideal for listening to during commutes, mind maps help in quickly understanding the structure of the information, while quizzes assist in reviewing what has been learned.

The generated content is a reorganization of the original material; it cannot replace a thorough reading, and it may also overlook certain constraints.

8. AI-generated images and illustrations for notes

Users can enter descriptions to create images in various proportions and styles, and they can also add images to their notes with just one click. The generated images are suitable for use as sketches, covers, or illustrations for content; however, before using them commercially, it is necessary to check for potential issues related to portraits, trademarks, fonts, and the training data used, so as to avoid treating these generated images as real news photos or evidence.

9. File format and new capabilities

The IMA format for storing data continues to evolve. The recent version for mobile devices now includes support for displaying local HTML, EPUB, and Markdown files, as well as the ability to move and copy content across different knowledge bases.

The specific format, file size limits, capacity of the knowledge base, and the quality of parsing may vary depending on the client and account privileges; it is advisable to test with a small sample before uploading large amounts of data.

Task mode and complex workflows

In previous versions of the “Tencent AI Workbench,” the task mode was described by the developers as follows: after entering a topic, it automatically breaks down the necessary steps, retrieves relevant information, and generates reports, podcasts, or PPTs. The current version places more emphasis on Copilot and skill sets; the name of the interface as well as the available functions may change depending on the version updates.

Regarding the tool catalog, it can be summarized as agent-based research and content creation; however, it cannot be guaranteed that every account or platform will have the same capabilities for creating PPTs or exporting content on a permanent basis.

Suitable tasks include organizing course topics, compiling industry information, processing reading notes, summarizing interviews, and creating preliminary drafts based on existing documents. For tasks that require complete citations, up-to-date prices, legal validity, or precise calculations, it is still necessary to use original sources and specialized tools for verification.

AI models and online searching

According to Tencent’s official information, IMA offers question-answering and knowledge-processing services based on models such as Tencent Hunyuan and DeepSeek, and it allows for the use of specific models in certain scenarios. The list of available models is updated regularly; the same function may require different models depending on the task, computing resources, and platform used. Therefore, it is necessary to follow the options provided by the client application when using these services.

Online searches can provide new information that goes beyond what is available in the knowledge base, but the reliability and timeliness of external web pages vary. When conducting research, it is advisable to clearly distinguish between information from the knowledge base and information obtained through online searches, and to check the date of publication, the author, the original source, as well as to ensure consistency across various sources.

Free use and pricing for computing power

The IMA app can be downloaded for free, and its core functions such as a personal knowledge base, notes, and basic Q&A are available without charge; however, some of the more complex models, skills, and resource-intensive tasks in the new version of Copilot require computational power.

The computational power purchases currently listed on the App Store in China are as follows:

• 75 computing power: 5 yuan;
• 225 computing power: 15 yuan;
• 375 computing power: 25 yuan;
• 750 computing power: 50 yuan;
• 1500 computing power: 100 yuan;
• 3000 computing power: 200 yuan.

The cost per unit of computing power is approximately 0.0667 yuan; however, 1 unit of computing power does not correspond to one single query-and-answer session. The actual consumption varies depending on the model used, the length of the context, the skills involved, and the task at hand, so it is necessary to check the instructions displayed on the interface before submitting a request.

The web version, the Android version, as well as any gifts offered as part of promotions or educational benefits may differ from those available on the Apple platform; the final price and benefits will be indicated on the settlement page.

Past information often mentioned a 2GB storage space for regular user accounts, while recent official initiatives have offered a 100GB knowledge base as a benefit for certified teachers and students. These figures vary depending on the version, account type, or conditions of the campaign, and therefore cannot be regarded as the fixed storage capacity applicable to all users.

Checking the actual capacity displayed on the account page is more reliable than following old tutorials.

Supported platforms

IMA offers access via web interfaces, as well as desktop versions for Windows and macOS, and apps for iPhone, iPad, Android, HarmonyOS, and WeChat Mini Programs. It also enables easy import of WeChat files and articles from public accounts, thanks to the Tencent ecosystem. The frequency of content publication may vary across different devices, and there might be differences in functions such as recording, file import, Copilot capabilities, and export options.

The Apple app for the Chinese market currently requires iOS 16.0 or iPadOS 16.0 or later, and it supports Simplified Chinese, Traditional Chinese, and English. The app size and system requirements may change with updates; therefore, it is necessary to check the latest information in the App Store before installing it on older devices.

API, MCP, and open-source status

Recent versions of ima support connection to MCP via Copilot, and there is now an ecosystem of skills or open interfaces for working with notes and knowledge bases. These capabilities are designed to enable compatible agents to read and write ima content within the scope of their permissions; they do not represent general large-model APIs that can freely replace the official client.

IMA’s clients, cloud-based knowledge base, Copilot, and model services are Tencent’s proprietary products. The company does not make the complete source code of this platform available publicly, nor does it provide an open-source version that can be deployed on one’s own. Packages, plugins, and third-party skills with similar names on GitHub cannot be automatically considered to be official Tencent projects.

Before use, check the warehouse organization, licenses, maintenance status, and the account permissions required.

The support for MCP or open interfaces does not mean that the product itself is open source.

IMA Knowledge Base Usage Guide

Complete a basic task.

1. Clarify the issue, time frame, location, source priority, and output format;
2. Upload materials for which access is permitted to the IMA knowledge base, or enter search queries;
3. First, establish a framework using the personal knowledge base, and then supplement it with evidence from shared knowledge bases and knowledge hubs.
4. It is necessary to distinguish between factual information from the source, the author’s opinions, and AI-generated conclusions.
5. Check each item for dates, numbers, the original location, and any conflicting evidence;
6. The conclusions are manually revised, the verification time is recorded, and then they are published;

Create reusable professional workflows

1. Break down complex topics into four categories of questions: background, data, comparison, and conclusions;
2. A fixed set of research steps is established by combining the personal knowledge base, the shared knowledge base, the knowledge hub, and Copilot Knowledge Partner.
3. Give priority to using the official website, research papers, regulatory documents, and raw data;
4. A second person is assigned to review conclusions that are considered high-risk;
5. Save queries, evidence, versions, and unresolved issues;
6. Re-run after the data changes and update the conclusions;

Which users is it suitable for?

• Students and teachers:Organize course materials, papers, and class recordings to create mind maps, podcasts, and review quizzes;
• Researchers:Create a specialized database, formulate questions based on various documents, and prepare research notes;
• Content creator:Collect web pages and articles from official accounts, extract the key points, create an outline, and continue to refine the content;
• Workplace users:Organize meeting recordings, project documents, industry reports, and working methods;
• Small team:Accumulate common resources and experience by sharing a knowledge base;
• Knowledge management users:Aggregate documents, web pages, audio files, and notes into a searchable personal second brain.

Product advantages

• Processing Chinese-language materials and integrating with the WeChat ecosystem is straightforward;
• Personal knowledge bases, shared knowledge bases, and public forums form a complete hierarchy;
• Put searching, reading, excerpting, writing, and Q&A in the same workspace;
• It supports various output formats such as recorded transcripts, mind maps, podcasts, quizzes, and AI-generated images;
• Copilot can handle multi-step tasks, and it integrates skills and MCP gradually;
• Covers web pages, desktops, mobile devices, and WeChat entry points;
• Core capabilities can be tried out for free, while additional computing power for complex tasks can be purchased as needed.

Usage restrictions and privacy reminders

• Misunderstandings, omissions, and fabrications can still occur in knowledge base queries, especially when there are conflicts in the information, poor scanning quality, complex tables, or questions that go beyond the scope of the document.
• Important answers should be verified in the original text; AI-generated summaries should not be used directly as contract terms, academic references, or a basis for making decisions.
• Uploading data means that the relevant content needs to be stored, parsed, and indexed within the service;
• Before importing WeChat files, meeting recordings, customer information, personal photos, and organizational documents, it is necessary to ensure that one has the authority to handle such content and to comply with the organization’s data and confidentiality policies.
• Apple’s privacy disclosures indicate that app functions may process user content and identifiers linked to identities, and diagnostic information may also be used for analysis;
• Special attention must be paid to permissions when sharing a knowledge base;
• Once it is made public, other users may read and use the information contained in it;
• It depends on the current product mechanism as to whether deleting the source file, ending sharing, or changing permissions will have an impact on existing copies.
• Sensitive information should first be masked, with a local backup kept;
• Copilot’s memory and personalization capabilities help reduce the need for repeated instructions, and they can also store preferences that are no longer needed.
• Users should regularly check the account’s history, sharing settings, and authorized connections.
• When using MCP or third-party capabilities, the principle of least privilege should be applied, and access that is no longer needed should be revoked.

Frequently Asked Questions

Is the IMA knowledge base free?

The core knowledge base, notes, and basic AI capabilities can be downloaded and used for free. Some Copilot models, skills, and complex tasks require computing power, which can be obtained through various campaigns or purchased outright.

Are the IMA knowledge base and the IMA AI workstation the same product?

They are different names for various stages of the same product. The current official name of the application is “ima – Tencent AI Knowledge Butler”; its core functionality remains the search, reading, and writing experiences based on a knowledge base, with Copilot having been enhanced as part of it.

Can articles from WeChat official accounts be imported?

Yes. The sources of content listed by the official system include local files, WeChat files, articles from official accounts, web pages, images, and audio; the specific methods for importing these contents vary depending on the device used.

Can it generate PPTs and podcasts?

The official early task mode supported the creation of reports, podcasts, and PPTs; the current version still offers podcast creation capabilities as well as AI-based analysis.

The options for importing and exporting PPT files may vary depending on the version and account used; it is necessary to follow the settings provided in the actual client software.

How long can a recording be made in one session?

The official mobile app states that the recording duration is up to 2 hours, and it offers multilingual transcription as well as automatic summarization. For long meetings, it is recommended to keep the original recording and review the key points.

What model does IMA use?

Official materials mention Tencent Hunyuan and DeepSeek, and model configurations may be available based on functionality. The list of available models is updated frequently; please refer to the list in your account.

Does IMA have an API?

The product is enhancing its capabilities to connect with notes and knowledge bases through various skills, open interfaces, and MCP; however, this does not mean that universal model inference APIs are available for all developers. The eligibility requirements, scope of use, and pricing are subject to the guidelines provided for developers at present.

Is IMA open-source software?

No. The full client, the knowledge base service, and the Copilot platform are not open source.

Third-party GitHub skills or compatible plugins do not represent the open-source nature of Tencent’s official products.