April 2026

April 14th

Views, a faster AI agent, three new connectors, and reliability fixes across dashboards, sharing, and incremental sync.

---

What’s New

Views

Views are a new data source type that let you define virtual tables using natural language or SQL. Describe the data you want — Querri generates the SQL and evaluates it against your connected sources. Views support multi-turn refinement, so you can iterate on the definition conversationally. A dedicated panel in the source preview modal shows view definitions and results.

In this initial release, Views can only be created via the CLI or API — the visual editor is coming in a future release. This makes Views especially powerful for data engineers, automation workflows, and teams that manage data infrastructure as code.

Views are a foundational capability for how Querri works with complex data. Rather than running every question against raw source tables, you define clean, curated virtual tables once — and the AI works from those. This makes analysis faster, more accurate, and easier to govern:

• Performance — the AI queries pre-shaped, narrower datasets instead of joining raw tables on every question. Shorter SQL means faster execution and fewer retries.
• Response quality — views eliminate ambiguity. When the AI is handed a well-named monthly_revenue_by_region view instead of raw transactions, it generates more precise answers with fewer hallucinations.
• Reusability — define business logic once (customer lifetime value, MRR, churn cohorts) and reuse it across every project, dashboard, and automation without duplication.
• Automation — views are the building block for scheduled pipelines. Materialized views can be re-run on a schedule and shared downstream via the API without any manual intervention.
• Governed access — views let you expose only the columns and rows each team or customer should see, acting as a lightweight permission layer on top of raw connectors.
• Multi-source joins — a single view can join data across connectors (e.g., Salesforce + Stripe + HubSpot), giving the AI a unified table to reason over rather than requiring it to navigate multiple sources per question.

Getting started with CLI:

# Create a view interactively with AI authoring
querri view new --prompt "monthly revenue by product line"

# Create a view directly with SQL
querri view new --name "Active Customers" --sql "SELECT * FROM customers WHERE status = 'active'"

# Refine an existing view conversationally
querri view chat <VIEW_ID> -m "add a filter for active customers only"

# Preview results without materializing
querri view preview <VIEW_ID>

# Materialize all views (or specific ones)
querri view run
querri view run --view-ids <VIEW_ID>,<VIEW_ID>

# List and manage views
querri view list
querri view get <VIEW_ID>
querri view update <VIEW_ID> --sql "SELECT ..."
querri view delete <VIEW_ID>

Python SDK:

# Create a view with SQL
view = client.views.create(
    name="Monthly Revenue",
    sql_definition="SELECT date_trunc('month', created_at) AS month, sum(amount) AS revenue FROM orders GROUP BY 1",
)

# Materialize views
client.views.run()

# Preview without materializing
preview = client.views.preview(view["id"], limit=50)

See the Server SDK guide for full documentation.

---

Faster AI Agent (Experimental)

An experimental v2 agent is now available behind the experimentalV2 toggle in organization settings. Early benchmarks show 45–80% faster responses on common queries, with significantly lower token usage. The agent reuses existing results when answering repeated questions instead of re-running analysis from scratch.

---

New Connectors

Connector	Details
Amplitude	API key connector with full event and user data sync
Mixpanel	API key connector for syncing event, funnel, and user data
Customer.io	API key connector for campaign, message, and customer sync

Connector creation now closes the modal immediately after setup — no more manual page refresh needed to see the new connector.

---

Python SDK and CLI

The official Querri Python SDK and CLI are now available on PyPI:

pip install querri          # SDK only
pip install "querri[cli]"   # SDK + CLI

CLI (requires the [cli] extra) — Upload files, create projects, ask questions, manage dashboards, sources, views, sharing, and more — interactively or from scripts.

querri auth login
querri --json file upload data.csv
querri --json project new "Sales Analysis"
querri --json project chat -m "What are the top products by revenue?"

Python SDK — Embed Querri analytics in your Python web app. The get_session() method resolves users, applies access policies, and returns a session JWT in one call.

from querri import Querri

client = Querri(api_key="qk_...", org_id="org_...")
session = client.embed.get_session(
    user={"external_id": "customer-42", "email": "alice@acme.com"},
    access={"sources": ["src_sales_data"], "filters": {"tenant_id": "acme"}},
)

Supports sync and async clients, automatic retries, typed error hierarchy, and all Querri resources (projects, dashboards, sources, views, files, users, sharing, audit logs, and more). Requires Python 3.10+.

See the Server SDK guide for full documentation.

---

Improvements

Area	Change
Smarter step execution	Analysis steps are now only re-run when their data or dependencies have changed, reducing unnecessary re-execution on project runs
Excel download	Maximized Excel workbook results now correctly offer “Download as Excel” instead of “Download as PNG”
User management	The user removal dialog now shows a loading indicator and correctly closes after deletion, preventing duplicate submissions
Themes	Custom themes now apply correctly across all components, replacing previously hardcoded styles

---

Bug Fixes

Area	Fix
Dashboard filters	Fixed a bug where dashboard filter interactions failed silently on every cache miss
Shared dashboards	Fixed 403 errors when loading images, Excel files, and other assets inside shared or embedded dashboards
Embed filters	Filter widgets in embed mode now load correctly and no longer persist state back to the backend
Incremental sync	Fixed three MySQL sync bugs: a DuckDB binding error, an S3 download error on local files, and row counts showing only the latest batch instead of the full dataset
Session refresh	Failed JWT token refreshes are now detected immediately instead of silently retrying with stale credentials for up to 30 seconds
Wrapped	Fixed a crash in the comparison slide when entity data was transiently undefined

---

Release Date: April 14, 2026

---

April 6th

Views, a faster AI agent, new connectors, smarter analysis, embedded chat customization, onboarding personalization, and dozens of reliability improvements.

---

What’s New

Improved Final Analysis

The final analysis at the end of a conversation is now more accurate and better grounded in your actual data, especially for longer conversations with multiple steps. Admins can customize analysis behavior in organization settings under Final Analysis Preferences.

---

Activation Checklist

New users now see a guided activation checklist during their first session, walking them through key setup steps — connecting data, running their first analysis, and exploring results. The checklist tracks progress and dismisses automatically once complete.

---

Save to Dashboard from Chat

After a successful chat response that includes data results, an inline Save to Dashboard button appears. Click it to add the result directly to a dashboard — Querri auto-creates “My Dashboard” if you don’t have one yet, or lets you pick from your existing dashboards.

---

Onboarding Experience

New users are greeted with a short onboarding survey (role and industry) that loads a personalized demo project tailored to their background. Users who prefer to explore on their own can skip and won’t be asked again.

---

Embedded Chat Customization

The embedded chat widget now supports fine-grained customization:

• Simple Mode — a carousel layout with granular toggles for tables, charts, suggestions, and reasoning
• Welcome screen — customizable welcome message for your embedded chat
• Action bar controls — toggle visibility of copy, share, print, and header buttons per embed instance

---

Leave Projects and Dashboards

Users can now remove themselves from projects and dashboards they’ve been invited to. Project and dashboard creators retain non-editable ownership permissions, and only creators can delete.

---

Automation Status Banner

When a scheduled automation is running on a project, the chat bar is replaced with a status banner so you know work is in progress.

---

Integrations

New Connectors

Connector	Details
Google Analytics 4	OAuth connection with property discovery and 11 pre-configured report types (sessions, events, conversions, and more)
NetSuite	Token-based authentication with 14 SuiteQL operations and incremental sync support
Attio CRM	Full OAuth connector with contact, company, and deal sync
Stripe	OAuth connector for syncing payments, subscriptions, customers, and more
Zendesk	OAuth connector for tickets, users, organizations, and satisfaction ratings
Salesforce	OAuth connector with full object sync including custom objects
Mailchimp	Sync your audience lists, campaigns, and engagement data
Intercom	In-app support widget for help and messaging

Connector Improvements

Connector	Improvement
MySQL	Users with restricted database permissions can now manually enter their database name. Special characters in passwords are handled correctly. Error messages now include actionable guidance (e.g., IP allowlisting instructions on access denied)
HubSpot	Gracefully handles missing optional object types and correctly syncs nested properties
eMaint	Fixed “Save and Load Data” not triggering source processing
All connectors	”Configure” is now the primary action on connected connectors. A new Files tab in the config modal shows connector files without leaving Settings. Save errors now display the actual backend message instead of a generic error code

Incremental Sync

• Fixed a bug where syncs with a cursor value of 0 were treated as having no cursor, causing unnecessary full re-syncs
• Added upsert support — incremental syncs can now update existing records in addition to appending new ones
• Primary keys are auto-detected for connectors that support them

---

AI Analysis Improvements

• Smarter error recovery — when a plan step fails, the AI now attempts to generate an alternative approach instead of stopping immediately. Recovery prompts include the actual error detail for better corrections
• Better SQL generation — new hints help the AI fix common SQL errors (UNION mismatches, missing GROUP BY columns) on retry instead of repeating the same mistake
• Validation against data fabrication — the AI can no longer generate analysis code that creates data from scratch instead of using your actual dataset
• Longer conversations supported — improved handling when conversations approach the context limit, with a clear message instead of an unexpected error
• Accurate timeout messages — plan timeout notifications now show the actual reason instead of a generic message

---

Improvements

Area	Change
Performance	Library and project pages load significantly faster thanks to permission caching
Reliability	Long-running analysis steps no longer falsely time out on complex queries
Streaming	Chat responses on long conversations no longer freeze mid-stream
Stale deploys	After a Querri update, stale browser tabs now show a friendly refresh prompt instead of a blank error
Offline detection	Fixed false “you’re offline” banners caused by stale browser caches after updates
OAuth errors	OAuth connection errors (e.g., QuickBooks) now show a clear error page instead of raw technical output
Permissions	Newly invited users get access immediately instead of requiring a page refresh
Dashboard cleanup	Deleting a project step now properly removes it from any dashboards that reference it

---

Bug Fixes

Area	Fix
Chat	Fixed cases where chat creation errors showed a misleading “no UUID returned” message instead of the real error
Dashboards	Dashboard updates now return a proper “not found” error instead of a generic failure when the dashboard doesn’t exist
Onboarding	Users who skipped onboarding are no longer prompted again on refresh
Connectors	Connector configuration modal no longer closes unexpectedly during setup
Sharing	Share links now correctly load subscription status without permission errors
Charts	Fixed a crash when Plotly chart data was null or undefined
File uploads	Fixed an unhandled error when uploading unsupported file types
Unicode	Fixed analysis failures when step names contained Unicode characters that looked identical but used different encodings
Large numbers	Fixed a database write error when AI analysis produced very large integer values

---

Release Date: April 6, 2026