WeChat Intelligence Hub is only as good as the database access it can't guarantee

WeChat Intelligence Hub treats your local WeChat message database as a personal intelligence source rather than a chat archive. The core idea is straightforward: given authorized, read-only access to the encrypted SQLite database that WeChat maintains on your Mac, the system indexes conversations and surfaces actionable signals — pending replies, unfulfilled commitments, business leads, dormant contacts worth reviving, and group-chat topic summaries — all within a configurable time window you specify in plain language.

The architecture is deliberately layered. A clean-room reader component handles the low-level database work: decrypting with SQLCipher, decompressing Zstandard-packed messages, and resolving current macOS file paths for media. A CLI skill wraps that reader as an agent-callable tool. The intelligence hub layer sits on top, applying judgment rules to produce reports. That separation matters because the database access problem and the intelligence extraction problem have different failure modes and different trust surfaces.

The access story is the honest hard part. The README is unusually candid: successful installation does not mean you can read anything. WeChat's schema-2 encryption requires a salt-key that the system cannot conjure — you either already have it from a prior extraction, or you go through a five-step onboarding workflow that may involve restarting WeChat and working with a re-signed copy. The README explicitly flags that this path has not been tested on all macOS versions and makes no guarantee of universal coverage. Notification previews are available as a degraded fallback, but those only capture inbound message snippets, not full conversation history.

Output comes in two forms: a structured Markdown file suitable for archiving or further AI processing, and an interactive HTML report with search, section navigation, topic filtering, and a dark/light theme. The HTML version includes what the README calls a "business signal radar" alongside group-chat digests and priority-contact summaries. Time windows are flexible — the 24-hour daily briefing is just the default example, not a ceiling.

Personalization is built in from the start. A profile-init command ingests a personal background document and a current-goals document, then uses those to rank and filter what surfaces in reports. Without them the system still runs, but flags its output as generic. WeChat contact labels — which you define yourself, not inherited from the maintainer's taxonomy — drive priority filtering for relationship categories like clients, partners, or distributors.

The privacy boundary is explicit and repeated: the reader never sends messages, never touches the WeChat UI, and real chat data must never be committed to the repository. Fictional sample data ships with the repo for testing the full pipeline without live credentials. The license is AGPL-3.0, with a commercial licensing path for anyone who needs closed-source integration.

The version on offer is v0.9.2-preview.2. The public interface aligns with 29 read-only tools and 266 input fields from an earlier CLI version, and real-data validation against current macOS paths has been completed. Other WeChat versions are described as accumulating coverage incrementally, which is an honest way of saying the matrix is not yet complete.