vMira Blog
Product updates, technical deep dives, and notes from the team building the all-in-one AI workspace.
SEO overhaul — hreflang, JSON-LD split, per-page schema
A focused pass on technical SEO: hreflang on every locale page, per-page JSON-LD with no global FAQ pollution, fresh /vs/[slug] and /use-cases/[slug] landing systems.
Voice and video mode — slice 1 backend lands
The first slice of the voice/video chat backend is live in production. The web frontend ships in slice 2.
Production moved to OVH BHS5
vMira production now runs on OVH dedicated infrastructure in Beauharnois, Quebec — lower latency for North American users and full control over the data plane.
Cross-chat shared context — Mira remembers what you talked about
A per-profile memory layer with cosine-similarity retrieval and time decay lets Mira recall prior conversations when they are relevant — opt-out in profile settings.
Research Mode public beta — 60-page reports with sourced citations
Research Mode pairs STORM and IterResearch into a multi-stage pipeline that produces long-form reports with inline citations. Available on Plus and Pro.
Finance subsystem — EDGAR, RSS, and Wikipedia, no paid vendor
How vMira surfaces financial information from public sources only — EDGAR filings, RSS news feeds, and Wikipedia — with a no-advice posture on every response.
vMira 1.0 — public launch
The all-in-one AI workspace is generally available: chat, code, design, music, documents, voice, an OpenAI-compatible API, and 50+ interface languages.