Capability Center
The catalog of registered capabilities — view, enable, disable, and configure them.
The Capability Center is Wujie's extension surface: a catalog of capabilities registered in code, with the entry points for viewing, enabling, disabling, and configuring each one.
Specific enterprise capabilities are still under development. This page describes the Capability Center as the catalog and management surface only — it does not describe capabilities that have not landed.
What the catalog is
The capability catalog is registered by code, not assembled by hand — what you see there reflects what the platform actually ships. Each entry can be inspected to understand what it provides before turning it on.
Enabling and configuring
Each capability can be enabled, disabled, and configured from the center. An enabled capability can add its own entry to the sidebar, which is how new functional surfaces appear in the product as capabilities come online.
The workflow is deliberate: look at the catalog, enable what your workspace needs, configure it, and let it surface its own entry — or disable it again without uninstalling anything.
Next
- Connectors — the current connector entries and how they read external docs
- Knowledge — knowledge management as a product surface
Meeting notes
Browse Home, My content, Shared content and Trash. Record starts immediately; global controls keep pause/resume/stop available during navigation. Stop automatically uploads. Batch uploads create one meeting per file and can be retried individually. Preselect an existing owned local Agent for automatic processing; unavailable Agents do not prevent saving media for later manual processing. Meetings are private by default. Creators may share with the workspace; other members have read-only access.
Search transcripts and seek to approximate chunk boundaries. Minutes include summaries, topics, decisions and actions. Edit content, preview versions, explicitly adopt results and export Markdown/HTML. Converting an action to an Issue first explains workspace visibility, then opens the existing creation form.
V1 requires an online owned Claude/Codex Agent using managed personal or enterprise billing, an updated daemon, and ffmpeg/ffprobe on the device. It reuses the existing model gateway and needs no new skill or ASR key. The original meeting skill and Issue recorder remain separate.
Microphone recording supports pause/resume and device-local draft recovery. Drafts are removed only after successful upload. Supported media is limited to 500 MiB and 24 hours. Wide details display minutes and transcript together, with fixed playback controls for seeking, speed and 15-second skips. In-app playback rechecks permissions; revoking sharing or deleting a meeting invalidates in-app playback grants. All deletes move to Trash for 30 days; restored meetings are private. Expiry or confirmed permanent deletion triggers independent background cleanup with retries. Original and converted audio reuse the existing public attachment storage: S3_BUCKET and the environment’s existing S3_KEY_PREFIX, or LOCAL_UPLOAD_DIR. No separate private bucket or directory is required. Meeting records, transcripts and minutes remain access-controlled, but anyone with an audio object URL can access the recording. Revoking sharing or moving a meeting to Trash does not revoke direct object access.