SiYuan’s most revealing number is not its 46,699 GitHub stars. It made 1,578 GitHub commits in the last 30 days, yet ToolVitals recorded 0 release events in that same period. That gap defines the current signal: a heavily worked-on self-hosted knowledge system whose public release cadence is quieter than its repository activity. Buyers should read SiYuan as active software, but they should not confuse active source development with a fresh production release.

What SiYuan is

SiYuan is built around a different unit of knowledge than a conventional page editor. The official website describes it as open-source, local-first, privacy-first note software with Markdown, block references, bidirectional links, databases, and multi-device synchronization.

The core model is the content block. The official site presents pages as blocks and says blocks can be split, regrouped, and moved without breaking existing links. It also describes more than 20 block-level elements, more than 10 inline elements, block focus, large-document editing, backlinks, graph views, and database relationships.

That design is more than a feature list. It is SiYuan’s central product bet: knowledge should remain addressable at a fine granularity, instead of being trapped inside a page-sized blob. Block-level references and two-way links are useful when a research note, task, or database entry needs to be reused in several contexts without copying the text.

The official site also puts data control at the center of the product. Its supplied description says data is stored on devices controlled by the user and remains available locally without a network connection or cloud service. It presents end-to-end encrypted synchronization as the mechanism for moving data across devices, with incremental synchronization intended to reduce transfer volume.

Those statements describe the product’s positioning, not an independent security audit. They still establish an important distinction from hosted note-taking tools. SiYuan is designed for people who want a local workspace, self-managed deployment, and control over where their knowledge lives.

The GitHub repository expands that positioning into a knowledge workspace where humans and AI agents work together. It describes fine-grained block-level reference and Markdown WYSIWYG editing, then lists databases, SQL queries, embeds, graphs, PDF annotation, exports, multi-tab layouts, and AI writing and question answering through the OpenAI API.

The repository also documents Docker deployment. The supplied instructions show a server launch using the serve subcommand, a mounted workspace, port 6806, an access authentication code, and configurable PUID and PGID values. That detail matters to operators because it makes the deployment model concrete. It also makes clear that self-hosting transfers responsibility for permissions, upgrades, backups, and access controls to the person running the instance.

ToolVitals describes SiYuan as written in TypeScript and Golang. It classifies the project as OSI-approved OSS and labels the license AGPL-3.0. That is the correct licensing distinction for this analysis. SiYuan is open source under the supplied ToolVitals classification, not fair-code, source-available, or open-core software.

The unusual signal in the metrics

ToolVitals gives SiYuan a hot score of 228.4, a health score of 94, a shipping score of 97, and an overall ToolVitals score of 95. The data confidence value is 90. These figures support a narrow conclusion: by ToolVitals’ measures, SiYuan has strong project-health and shipping signals, with enough public evidence to produce a high-confidence score.

The raw activity numbers explain why. SiYuan has 46,699 GitHub stars, 1,578 GitHub commits in the last 30 days, and 30 GitHub releases in the last 90 days. It also has 3 active contributors in the measured period. The combination points to concentrated output from a small active group, although the metric does not establish the project’s total maintainer count or the size of its broader contributor community.

The release fields need to be read carefully. ToolVitals reports 30 GitHub releases over 90 days and 0 release events over 30 days. Those are separate fields and should not be collapsed into one claim about release frequency. A project can have substantial recent release history while still having no release event in the latest measurement window.

That is what happened in the supplied evidence. The recent event list contains SiYuan releases between August 14 and August 21, 2026, including beta builds, v3.8.1, and v3.8.2 alpha builds. The payload was generated on October 10, 2026, so the August events do not contradict the 0 release events in the latest 30-day window. They show an earlier burst of public packaging, followed by a quieter current window.

The distinction matters for anyone choosing software to deploy. A high commit count tells you that code is changing. A release event tells you that maintainers have published a version for users to consume. Those signals can move at different speeds, and SiYuan currently displays that separation clearly.

The related note-taking entries provide a useful, limited comparison. Trilium is listed with 38,265 GitHub stars, a hot score of 168.0, a shipping score of 98, and 0 release events in 30 days. Trilium.cc is listed with 38,248 stars, a hot score of 168.0, a shipping score of 98, and 0 release events in 30 days. Both entries are classified as OSI-approved OSS under AGPL-3.0.

SiYuan has the larger star count and higher hot score in that comparison, while its shipping score is 97 against Trilium’s 98. This is not a feature benchmark or a user-satisfaction survey. It only says that, within ToolVitals’ related-tool data, SiYuan has a stronger activity and interest signal while showing a similar current absence of release events.

What changed in the recent releases

The v3.8.1 release notes frame the release around three areas: more flexible AI integrations, more efficient editing and database workflows, and better cross-platform reliability. That framing matches the repository’s current description. SiYuan is not merely adding an AI chat box to a notes application. It is adding agent-facing actions inside the same block, document, database, and publishing model.

The AI changes are specific. The release supports syncing custom AI editor actions across devices, shortcuts for the Agent panel, user-level skills from the shared agents directory, nested rerank API formats, and the OpenAI Responses API across AI features. Publishing services can configure plugin availability. These changes make AI a set of configurable operations inside the workspace rather than a separate destination for copied text.

That direction has practical consequences. A team evaluating SiYuan should ask whether its AI features can work with the team’s preferred provider, model, access controls, and data boundaries. The release notes establish API and agent integration points, but they do not establish that every workflow is ready for sensitive production data. The repository’s mention of MCP and agent features makes that validation more important, not less.

The editing and database work is just as substantial. v3.8.1 includes table row and column operations after merged cells, Markdown table transposition, inherited child-document sorting rules, ordered-list starting numbers, table-column sizing, database item templates for clipping, faster database grouping, and virtualized grouped table rendering. It also improves document loading and several graph and tab interactions.

The cross-platform work rounds out the release. The notes mention HEIF and HEIC previews, mobile behavior, PDF rectangle annotation quality, export handling, marketplace updates, widget resource cleanup on iOS, and fixes involving security, MCP, publishing services, and formulas. This is maintenance work with visible user impact. It targets the parts of a self-hosted application that become painful when desktop, browser, mobile, export, and server deployments do not behave alike.

The follow-up alpha releases continue that same direction. The v3.8.2-alpha.1 notes and v3.8.2-alpha.2 notes list independent execution across Agent sessions, better Agent response tags, handling for malformed structured arguments in Agent question and todo tools, backlink filtering, database interaction improvements on mobile, and improved data synchronization conflict detection.

The alpha work also reaches into the editor and deployment surface. The supplied notes mention large-list rendering virtualization, a customizable editor toolbar, multiple editor fonts, mobile marketplace support, plugin event hooks, asset upload interception, mobile navigation changes, and improvements to graph layout and database search. Alpha.2 adds fixes for embedded images pasted from Microsoft Word, duplicate asset filenames, graph rendering, marketplace behavior on mobile, and other editor, export, and synchronization paths.

The event order is revealing. The supplied release list moves from v3.8.1 beta builds in mid-August, to v3.8.1 on August 18, then to v3.8.2-alpha.1 on August 20 and v3.8.2-alpha.2 on August 21. That is evidence of fast iteration during that period. It is not evidence that the same pace continued through the latest 30-day measurement window.

The best interpretation is not that SiYuan has stopped. The 1,578 commits argue against that. The better interpretation is that source development and public release packaging have entered different phases. Maintainers may be preparing a later release, concentrating on internal changes, or simply operating on a cadence not visible in the current event metric. The supplied evidence cannot identify the reason.

What the data does not tell you

ToolVitals can measure repository activity, contributors, stars, releases, licensing signals, and other public project evidence. It cannot measure code quality, test coverage, vulnerability exposure, user satisfaction, support quality, revenue, or whether SiYuan works well for a particular team’s documents and workflows.

Stars are a visibility signal, not an active-user count. A 1,578-commit month does not say whether the work is concentrated in core features, bug fixes, generated changes, documentation, tests, or build infrastructure. The 3 active-contributor figure identifies activity in a defined period, but it does not tell you how maintainers share ownership or how quickly the project could respond to a serious issue.

The release count also says nothing about upgrade safety. A long change list can indicate useful progress, but it can also increase the number of paths that need regression testing. The supplied release notes mention synchronization conflicts, mobile workspace deletion, database references, PDF export, MCP endpoints, publishing services, and formulas. Those are exactly the kinds of boundaries an operator should test rather than infer from a shipping score.

The privacy claims need the same treatment. The official site describes local storage and end-to-end encrypted synchronization, but the supplied excerpts do not provide an independent audit, threat model, cryptographic review, or operational guarantee. A privacy-sensitive buyer should treat those claims as product documentation and verify the deployment, key-management, backup, and recovery model before placing confidential material in the system.

A skeptical engineering lead should evaluate SiYuan in a disposable or staging workspace first. Test a Docker deployment, host-mounted permissions, workspace backups, synchronization recovery, large documents, database grouping, PDF export, mobile editing, plugin behavior, and the exact AI or MCP integrations the team intends to use. Run an upgrade from the v3.8.1 reference point against a copy of the workspace, and keep the v3.8.2 alpha line out of production until its specific workflows pass those checks.

Recommendation for buyers and maintainers

For maintainers, the metrics suggest a communication problem more than a development problem. ToolVitals sees 1,578 commits in 30 days, 30 GitHub releases in 90 days, 3 active contributors, and 0 release events in the latest 30 days. A concise public distinction between stable releases, beta builds, alpha work, and unreleased repository activity would help operators understand the gap without guessing. The detailed release notes are already useful. Pairing them with explicit upgrade-risk notes, migration guidance, and current channel status would make that evidence more actionable.

For teams choosing a knowledge system, SiYuan is a credible candidate when the requirement is a self-hosted workspace that combines block-level references, databases, local control, multi-device workflows, and AI-agent integration. Its 95 ToolVitals score, 46,699 stars, and recent release history justify an evaluation. They do not remove the need to test backups, synchronization, upgrades, permissions, exports, and provider-specific AI behavior.

The concrete recommendation is simple: if your team needs to keep knowledge under its own operational control and wants agent features inside a structured notes workspace, evaluate SiYuan with v3.8.1 as the stable reference shown in the supplied release sequence. Compare its recovery and upgrade behavior against Trilium, then make the production decision from those tests, not from stars or commit volume alone.

Sources