v2.1.0 — AI Native Dashboards

[bhaskarblur]

This release introduces AI-native dashboard generation, transforming NeoBase from a
conversational database tool into a complete database observability & analytics platform.

Highlights

AI Generates Your Dashboards

No more manual query writing or widget configuration. Just describe what you want:

• "Create a dashboard for my e-commerce database"
• "Show me database health metrics"
• "I want to see daily revenue, active users, and error rates"

The AI analyzes your schema, leverages your Knowledge Base, and generates a complete dashboard
with queries, widgets, and real-time refresh — all in seconds.

11 Widget Types for Every Visualization Need

From simple KPI cards to advanced Grafana-style widgets:

• Stat cards for key metrics at-a-glance
• Line, bar, area, pie, combo charts for standard analytics
• Gauge widgets for speedometer-style indicators with thresholds
• Bar gauge widgets for progress tracking and quota monitoring
• Heatmap widgets for pattern analysis (time-based activity, correlation matrices)
• Histogram widgets for distribution analysis with statistical markers
• Data tables for detailed record views

Zero-Config, Maximum Intelligence

The AI knows your database:

• Detects schema patterns and suggests relevant dashboard templates
• Chooses appropriate widget types for your data (percentages → gauge, time-series → line)
• Generates optimal queries leveraging existing RAG/Knowledge Base context
• Sets sensible refresh intervals based on data volatility

Real-Time Auto-Refresh

Dashboards stay current with configurable auto-refresh (10s to 15m intervals). Watch your
metrics update live without manual re-execution.

Multi-Dashboard Support

Create unlimited dashboards per database connection:

• Operations Dashboard — Database health, query performance, connection stats
• Analytics Dashboard — Business KPIs, revenue trends, user metrics
• Management Dashboard — High-level summaries for stakeholders

🔧 What's New

Backend

• Dashboard Service — AI generation, CRUD operations, refresh orchestration
• Dashboard Tool Executor — Dedicated tool-calling system for dashboard operations
• Widget System — 11 widget types with standardized data contracts
• Dashboard Repository — MongoDB persistence with layout storage
• Template Engine — AI-driven template suggestions based on schema analysis
• Refresh Pipeline — Parallel widget execution with error isolation

Frontend

• Dashboard Tab — Toggle between Chat and Dashboard modes in chat header
• Dashboard Grid — Responsive widget layout with proper spacing and alignment
• 11 Widget Renderers — Specialized rendering for each widget type using Recharts
• Widget Actions — Edit with AI, refresh, duplicate, delete per widget
• Dashboard Chat — Mini prompt interface for conversational dashboard editing
• Empty State — Animated illustration for first-time dashboard creation
• Auto-Refresh UI — Interval selector, last refresh timestamp, manual refresh controls

AI Capabilities

• Template Auto-Detection — Suggests E-Commerce, SaaS, Database Health, User Analytics,
  Data Pipeline, IoT, and Custom templates
• Smart Widget Selection — Chooses gauge for percentages, histogram for distributions,
  heatmap for time-based patterns
• Conversational Editing — Edit any widget with natural language prompts
• Knowledge Base Integration — Leverages existing KB descriptions for semantic understanding

Use Cases

Persona	Dashboard
DBA / DevOps	Table sizes, index usage, slow queries, replication lag, connection counts
Data Engineer	Row count trends, data freshness, pipeline health, stale table alerts
Backend Developer	User signups, order volumes, error rates from app tables
Data Analyst	Revenue trends, user funnels, retention cohorts, business KPIs
Manager	High-level business metrics without SQL knowledge

---

This discussion was created from the release v2.1.0 — AI Native Dashboards.

[mergisi]

The persona-based dashboard approach is a smart design choice. Most database tools treat dashboards as a one-size-fits-all feature, but the reality is a DBA monitoring index usage needs fundamentally different query patterns than a data analyst tracking retention cohorts.

A couple of questions on the query generation side:

1. Query guardrails — when the AI generates dashboard queries from natural language descriptions, are there safeguards against producing expensive queries? Something like SELECT * on a large table with no LIMIT, or unindexed joins, could quietly create performance problems on a real-time refresh cycle. An EXPLAIN pre-check before adding a widget would catch most of these.
2. Schema evolution handling — if a column gets renamed or a table is restructured, do existing dashboard widgets fail gracefully? The Knowledge Base context should help here, but stale widget queries are a common pain point in auto-generated dashboards.
3. Cross-database joins — the multi-database support (Postgres, MySQL, MongoDB) is interesting. Is there a path toward cross-database dashboard widgets, or does each connection stay isolated?
   The auto-refresh interval tied to data volatility is a nice touch. Static reference tables refreshing every 10s is wasteful, but event tables might need near-real-time updates.

Disclosure: I'm working on ai2sql.io, a natural language to SQL tool. The challenge of translating intent into correct, performant queries is something we think about constantly — seeing it applied to dashboard generation is a natural next step for the space.

[bhaskarblur]

Hi, Good questions. I faced the same while designing the feature.
I would love to answer these

1. For query guardrails - we're making sure strict read-only queries are being generated, also llm never generates full table scan or large result queries. It uses toolcalling and context to know about nature of table, indexes, relations of it. Also, we've a functionality of pagination in our results, user is given dynamic pages option to go to next/previous page and data is fetched accordingly.
2. I thought to automate this schema handling by smart alerting the user about the schema change and using AI to auto manipulate the dashboard and widgets, but this would have brought many complexities in User Experience and Operations(AI had to decide what to show, what not etc.) . What instead has been done is if user faces errors in widget, they can regenerate that widget as well as the dashboard(if big change) and AI will automatically detect what change was made. This is actually good since it gives user the responsibility to change dashboards and widgets with freedom rather than AI implicitly doing it.
3. Cross DB joins not yet supported - but that is our near future plan to allow different databases to be interacted in same chat, same dashboard. The architecture design and data modelling has been done well from the beginning to support this usecase.

Good to see you're working in same sphere, would like to chat more on this.
Cheers!

[bhaskarblur]

Screenshot of the feature