This page provides auto-updatable AI knowledge assets for the FrameworX Designer MCP Server.
The Designer checks this page once per session (background, non-blocking) and downloads any updated assets automatically.

How it works: On startup, the MCP Server reads local cached files. In the background, it fetches this page, compares the table below against its local versions.json, and downloads any assets with a newer version. Files are validated (must be valid JSON) and written atomically. Hot-reload happens immediately — no server restart needed.

Knowledge Assets

The table below lists all available AI knowledge assets. To publish an update:

1. Prepare the updated JSON file (must be valid JSON).
2. Upload it as an attachment to this Confluence page.
3. Update the Version number in the table row (must be higher than the previous version).
4. Update the Download link to point to the new attachment.

Column Reference

File Specifications

Documentation.json

Offline documentation index used by the search_docs MCP tool.:

Section types follow the Diataxis framework: learning, concept, task, reference, example, support.

Page fields:

• title — Page title (used for search matching, highest weight).
• section — Section key from the sections dictionary.
• url — URL path relative to baseUrl (Confluence-style: spaces as +).
• tags — Search keywords (weighted in relevance scoring).
• labels — Confluence labels. Pages with "skill" label appear in the skills catalog.
• dbslnUrl — URL to downloadable solution export (only for "example" labeled pages).
• content — Plain text body (HTML stripped). Used for full-text search and snippet extraction.

ContextDocument.json

Solution context document that overrides the hardcoded C# baseline in SolutionTools.cs. This allows updating the AI's understanding of FrameworX architecture, syntax rules, and display specifications without a product release.

The context document structure mirrors what SolutionTools.GetSolutionContext() returns. When this file exists, SolutionTools.GetContextRefreshPayload() reads from it instead of the hardcoded C# object. The version field at the root is used for tracking.

Client Behavior

File Locations (search order — first found wins)

Update Flow

1. Startup (synchronous, instant) — AIKnowledgeUpdater.Initialize() reads local files only. Never touches the network.
2. Background check (fire-and-forget) — Task.Run(() => CheckForUpdatesAsync()). Single attempt per process lifetime. No retries, no polling.
3. Fetch this page — HTTP GET with 5-second timeout.
4. Parse table — Finds the first <table>, matches header columns by keyword.
5. Compare versions — Checks each asset against versions.json in the cache folder.
6. Download — Fetches attachment URL, validates JSON, atomic write (temp file → rename).
7. Hot-reload — DocumentationService.ReloadCache() or LoadContextDocument(). No server restart needed.

Opt-Out

Users can disable update checking by placing a settings.json file in ProductPaths.AIKnowledge/:

<ac:structured-macro ac:name="code"> <ac:parameter ac:name="language">json</ac:parameter> ac:plain-text-body{ "checkForUpdates": false }</ac:plain-text-body> </ac:structured-macro>

Failure Handling

• Network unreachable → silently continues with local/bundled files.
• Page returns non-200 → silently stops.
• Table parse failure → silently stops.
• Downloaded file is not valid JSON → rejected, not saved.
• File write failure → temp file cleaned up, continues with previous version.
• All errors are logged to LastCheckResult for diagnostics.

Diagnostics

To check update status from the MCP server, the AIKnowledgeUpdater.GetDiagnostics() method returns:

• initialized — Whether Initialize() was called.
• downloadedPath — Path to the writable cache folder.
• bundledPath — Path to the read-only bundled folder.
• updateCheckCompleted — Whether the background check finished.
• lastCheckResult — Human-readable result message.
• contextDocumentOverride — Whether a context override is loaded.
• contextDocumentVersion — Version of the loaded context document.