Features
- Connect works from plain chat — a conversation without an agent context can now open Connect sessions when the organization holds an enabled org-scope grant, and such sessions are attributed to the organization's chat assistant.
- A truthful desktop tool contract — the desktop tool now tells the model what it really does (macOS native app control through accessibility references, with an open-the-app-first recipe) so agents attempt desktop tasks instead of refusing them, and dispatch survives real networks with connection-loss detection and message re-send with device-side dedupe.
- The files surface — Teams store-agent runs can list, read, search, and edit documents inside folders the user declares in the CHATTERgo Connect app, with per-folder Read-only or Read + edit permissions, deletes going to Trash, and containment enforced on-device; desktop app-allowlists and per-action consent holds round out the safety core.
- Conversation replay for super-admins — a replay viewer with search, transcripts showing applied-tool chips and execution events, and a per-run full-fidelity trace.
- Dashboard home rebuilt as the adoption hub — a setup-progress banner that auto-hides at completion, an analytics-gated KPI strip deep-linking the reports, a six-card feature toolkit with live adoption status, and a plan and usage rail with Billing v2.
- Binary email attachments for organization tasks — forwarding an email with Office, PDF, RTF, zip, or image attachments to a task address stores the files verbatim into the organization's workspace storage for task runs.
- A SEO/GEO content expansion on chattergo.ai — use-case landing pages (a hub plus five keyword-targeted pages in a direct-answer format), colleague role pages (a hub plus one page per commerce colleague), the AI colleagues theme with a blog series, sitewide social-preview images and structured data, and an llms.txt feed — all trilingual.
Fixes
- Ingredient searches in English no longer return an error on plain ingredient names — a search-query path that raised a database error on valid questions was replaced with a never-raising parser.
- B2B price narrowing now applies the shopper's group pricing on every search path, so narrowing by price no longer surfaces products above the group tier.
- A product-lookup tool no longer errors for organizations with customer-group pricing enabled.
- Zero-result searches now fall back to per-token probes and vector matching instead of returning nothing when one qualifier word is absent from the catalog.
- Dashboard counters — Products indexed and teammate counts no longer read zero on the dashboard home.

