Release notes

Your proposals, now with an AI that knows your record

VETR 3.0 turns everything you've written into an advantage. Three new capabilities, all grounded in your own data and running entirely inside the GovCloud boundary — your CUI never leaves.

What's new

• Evaluation-Factor Drafting — VETR reads the RFP's Section M evaluation factors and drafts a response to each one that cites your own past performance and is written to score. It shows the predicted evaluator rating (FAR 15.3) and the specific gaps to close — so you see your score before the government does.
• Agentic Proposal Team — put a proposal in front of an AI review team: compliance, past-performance, win-theme, and pricing specialists each review it through their lens, then a capture lead synthesizes a verdict, a readiness score, and a prioritized action plan.
• Data-driven Win Probability — Pwin is no longer a formula. VETR trains an explainable model on your own win/loss history and shows the top factors driving each estimate (a transparent baseline runs until you've logged enough outcomes, then it learns from your results).

Under the hood

• Proposal memory — your proposals, past performance, win themes, and lessons learned are now indexed for retrieval, so the AI drafts in your own voice and evidence — never generic boilerplate.
• In-boundary by design — all inference runs on Amazon Bedrock and all embeddings on Titan, inside the FedRAMP authorization boundary. Independently verified: no proposal content ever egresses to an external AI provider.

What's new

• Knowledge Base & education portal — a searchable help center with guides for every part of VETR: the VETR methodology, RFP parsing, compliance, proposals & the AI copilot, teaming, past performance, the new Delivery Studio, billing, integrations, and a GovCon glossary. Browse at /help or search instantly.
• Semantic Help Assistant — the in-app assistant now answers from the knowledge base using semantic retrieval, so it understands what you mean, not just the keywords you typed. Your private data is never mixed in — the assistant only reads published help content.
• Super-admin authoring — categories and articles are fully editable in the admin with a rich-text editor; no deploy needed to publish or update help content.
• "Was this helpful?" feedback on every article, plus per-article view + helpfulness analytics for the team.

Win it, then deliver it

VETR now follows you past the award. When you mark a proposal won, a Delivery Studio workspace opens for that contract — the only GovCon-native place to actually produce the technical work after you win.

What's new

• Delivery workspace — deliverables (CDRLs) and period-of-performance milestones seeded from your contract, with a live on-track / at-risk / off-track health signal.
• AI Studio generation — draft project briefs, charters, delivery plans, and documents grounded in your contract's scope (retrieval-augmented, so output reflects your contract — not boilerplate). Plus delivery-phase research.
• Prototype canvas — describe a tool or dashboard and the Studio builds a clickable prototype, rendered in a locked-down sandbox with no network access.
• Autonomous Delivery Agent — kick off a task like "draft the Monthly Status Report" and the agent runs it in the background: retrieve → draft → (your approval) → write → self-critique → notify. It pauses for your sign-off before writing the full draft, and emails you when the draft is ready.
• CPARS flywheel — capture delivered work as a past-performance record (AI-drafted CPARS narrative + STAR), confirm it, and it strengthens your Experience pillar on the next bid. Deliver well → win more.

The Delivery Studio is available on the Professional and Enterprise plans.

What's new

• Real-time RFP parsing — the parse screen now streams progress over WebSockets (Reverb). No more waiting on a manual reload to see the result; you'll watch the percent and stage label update live.
• Audible notifications — the bell plays a short two-tone ping when a new notification arrives; mute toggle lives inside the bell dropdown.
• Collapsible icon-rail sidebars — toggle the sidebar to a 48-px icon strip with hover tooltips instead of full hide/show; works in the customer app and super-admin.
• Redesigned error pages — 401/403/404/419/429/500/503 now share a brand-true layout (navy + crimson, dark-mode aware, Fraunces display).
• Branded email theme — system emails use VETR navy headings, brand-coloured buttons, cream backgrounds, and a consistent Acu-Elligent footer.
• Onboarding tour polish — anchors line up with the elements they describe, and finishing the tour no longer relaunches it on the next render.

Fixes

• Win Theme creation now redirects to billing with a plan-upgrade message instead of a generic Access Denied page
• Notification bell badge uses the brand crimson with a ring outline so it's visible in both themes
• Avatar dropdown email/role text contrast fixed in dark mode
• Public footer Veteran-Owned badge no longer washes out white-on-white
• Sidebar scroll position is preserved across navigation
• Google sign-in callback gracefully restarts the OAuth flow on session loss
• Notifications page no longer crashes when the paginator object is shared as a notifications prop

What's new

• New typography — Fraunces (display serif) + Mulish (body sans) bring a more institutional, trust-forward feel to the marketing site.
• Branded proposal exports — per-org logo, colors, and font in PDF and DOCX exports. Configure at Settings → Branding.
• GPT-4.1 across AI services — better structured output and more accurate RFP parsing.
• SAM.gov filters — sort by posted/response/value, filter by agency multi-select, value range, response-date window; one-click save-to-pipeline.
• TipTap editors in super-admin — blog, help-articles, release-notes now use a full rich-text editor instead of plain textarea.

Fixes

• Dropdown menu hover no longer shows red background with dark unreadable text
• Sidebar logo tagline no longer truncates on narrow widths
• 404/500/503 error pages fully light-mode consistent

Improvements

• Win Probability engine now accepts competitor count and historical win rate for more calibrated scoring
• Proposal workspace prepared for CRDT-based realtime editing (experimental flag WORKSPACE_CRDT=1)
• Integrations page: real SVG logos for all partners (no more emoji)

Security

• Login, register, forgot-password, and reset-password routes now rate-limited per-IP (5/min for login, 10/min for register)
• CORS allow-list verified for production domain only
• HSTS + security headers revalidated