Changelog

Faster answers and safer rollouts

We tuned the stack for higher throughput and added rollout controls so more teams can drive production traffic without chasing incidents.

What's in this release:

• Latency tuning: The planner now prefers warmed shards during query spikes so responses stay snappy even at peak load.
• Structured snippets: Responses can include optional citations and snippets arrays for turnkey UI rendering.
• In app changelog: The dashboard surfaces these release notes beside deployment and key controls so updates stay visible.
• Sharper filters: Region and freshness filters behave consistently across the dataset browser and query playground.

Fixes:

• Filter casing: Fixed case sensitivity issues in dataset and region filters that previously returned empty results.
• Usage graph sync: Resolved a race condition where concurrent key changes could temporarily desync usage graphs from reality.

Dataset discovery and health signals

We made it easier to see what data is available and how it performs in production.

What's in this release:

• Richer dataset browser: Filter datasets by domain and topic, search by name or description.
• Dataset details: A new drawer shows descriptions, example questions, schema hints, and suggested use cases for each pack.
• Latency budget header: Responses include an x-bv-latency-ms header so you can track end to end performance from your own logs.
• Event webhooks: Ingestion and usage events emit to webhooks with built in retry, plus replay controls from the UI.
• SDK consistency: Language SDKs respect custom timeouts and auth changes instead of silently falling back to defaults.

Fixes:

• Document counts: Corrected a lag in dataset document totals after large ingestions so counts stay in sync.
• Freshness labeling: Fixed a bug that occasionally showed stale datasets as recently updated.
• Duplicate webhooks: Resolved duplicate deliveries for some users after transient network failures.

Enhanced UI polish and faster feedback

We tightened up the dashboard experience with skeleton loading, refreshed visuals, and workflow nudges that keep early setups moving.

What's in this release:

• Skeleton loading: Dataset browser, Generate API Key wizard, usage graphs, and query results now show shimmer placeholders while data loads.
• Visual refresh: Updated spacing, typography, and layout keep the dashboard readable on smaller screens while matching the new design system.
• Workflow nudges: After creating a key, inline prompts suggest next steps like adding datasets, copying the key, or running a test query.
• Wizard guidance: Domain and dataset selection now include inline validation and clearer helper text.
• Safer key handling: Copy to clipboard behavior is more explicit and avoids accidental partial copies.
• Sharper empty states: Workspaces with no datasets, keys, or usage show guided empty states instead of blank panels.

Fixes:

• Domain switching: Fixed an issue where changing domains silently cleared existing dataset selections.
• Loader flicker: Resolved flickering skeletons on fast connections when navigating between dashboard tabs.
• Copy reliability: Addressed a problem where the copy button did not always capture the full API key in some browsers.

Launch: API first domain knowledge

BlackVec is now live in early access with curated data packs you can query straight from your stack for RAG and agent workflows.

What's in this release:

• API key wizard: A three step Generate API Key flow guides you through selecting a domain, choosing datasets, and creating a key, complete with country picker and validation.
• Query endpoint: Call /v1/query from curl, Node, or Python using workspace headers for clean multitenant access.
• Curated datasets: High quality datasets ship with citations and provenance so responses are inspectable.
• Query playground: Run test queries in the dashboard using the same keys and datasets you configure for production.
• Usage overview: Workspace level metrics show request volume, latency percentiles, and error rate to track early traffic.
• Starter pricing: A pay as you go Starter tier lets teams prototype without long commitments.