Changelog

Click the "Create AI Suggestion" button on the thread list screen, and Squadbase AI will explore your data sources and dashboards to suggest what analysis to tackle next.

Squadbase's Editor has received a major overhaul. We've reorganized complex elements to deliver a more intuitive experience.

Threads

Threads are now displayed at the top level of the chat area. You can start a new thread and move on to your next task—even while a previous chat is still running. Status badges let you see the state of each thread at a glance.

Suggested Next Actions

When Squadbase AI completes a task, it now suggests next actions. This lets you kick off a new task without typing, or quickly try ideas you hadn't even considered.

Data Source Management

Managing data sources during dashboard development is always a chore. The new Data Sources tab gives you a clear overview of connected sources and makes adding new ones a breeze.

You can now duplicate projects from the project card view.

The following items are carried over to the new project:

• Source code
• System Environments + Environment Variables
• Data Sources
• Squadbase DB (schema definition)

We’ve added a new MySQL connector.

When you connect via the connector, the MCP server is automatically configured so you can get started right away.

We’ve added a guided tutorial project to help you learn the end-to-end workflow of building dashboards with Squadbase. In the tutorial, you’ll upload a spreadsheet, prepare the data, and build a dashboard step by step—while picking up practical vibe-coding skills like data preparation and context management along the way.

You can create a tutorial project from the Project Creation screen.

Squadbase AI can now connect to Squadbase DB. This means you can complete an end-to-end workflow entirely in chat—clean up uploaded CSV/Excel files, load the data into Squadbase DB, and build a dashboard from it.

When you create a database in Squadbase DB, the appropriate MCP server is automatically configured.

With the new time-based limits, usage automatically opens up again within at most 24 hours, so you can resume work without a long interruption.

When you create a new connection from a connector, Squadbase now automatically passes the appropriate MCP servers and other dashboard implementation details to the agent.

You can also create multiple connections from the same connector—for example, connecting several Google Analytics properties to a single dashboard. Existing connections will continue to work as they are, but we recommend creating new connections via the connector-based flow.

We currently offer templates for major data sources such as Google Analytics, Snowflake, and BigQuery — with more on the way.

Squadbase now updates the product UI based on the language you select in your Portal settings.

At this stage, Japanese is the first fully localized experience. For all other language choices, the interface remains in English as a common default across locales.

Dashboard Mode, previously offered as a dedicated mode for dashboard implementation, has been merged into Thinking Mode.

When using Auto Mode, tasks that require deeper reasoning will now be automatically routed to Thinking Mode.

The chat display in the app remains unchanged, while the context passed to the AI is compacted in the background. This keeps responses stable and accurate even for long-running tasks and extended threads, without requiring any manual cleanup from the user.

We’ve added a new Knowledge feature that lets you teach Squadbase AI project-specific context.

From the Settings tab in the Editor, you can define shared instructions for the project, and from your personal settings in the Portal you can configure preferences such as language and time zone.

By registering data source descriptions, business logic, and frequently used aggregation patterns as Knowledge, you can avoid repeating the same explanations in every chat and get more accurate, consistent answers from the AI.

You can now include dashboard error logs as context for AI chat.

By passing full error messages and stack traces to the AI, it can more accurately understand what went wrong and suggest concrete next steps for investigation and fixes.

Even for complex issues, you can point the AI directly to the relevant log entries from chat, making it easier to identify root causes and significantly improving troubleshooting efficiency.

You can now undo changes and restore previous versions of your work.

All revision history is preserved, including versions you roll back from, so you can safely return to an earlier state without losing any history.

We’ve added an Auto Mode that automatically switches between the General, Think, and Fast modes.

Auto Mode analyzes each user prompt in real time and selects the most suitable mode, enabling a smoother dashboard-building experience without needing to manually change modes.

Support for connecting to Google Analytics is now available. You can set environment variables, configure the MCP server, and install required packages directly from chat, allowing you to start building dashboards immediately.

We’ve added new AI modes that allow users to control how the AI behaves based on their goals. These modes fall into two main categories: conversational modes and task-oriented modes.

Conversational Modes

• General – A balanced mode offering strong reasoning, speed, and cost efficiency.

• Fast – A speed-optimized mode suitable for simple tasks.

• Thinking – A high-reasoning mode designed for complex tasks.

Task-Oriented Modes

• Dashboard – Optimized for implementing dashboards, especially when adding pages or charts. This mode provides a strong balance of speed and cost efficiency tailored for dashboard development.

• Excel/CSV – Designed for cleansing and preparing data from uploaded Excel or CSV files. It offers high performance and cost efficiency for spreadsheet-based data processing.

To use these features, you will need to update your project template. Please update it via "Update template" in the toolbar.

With the new UI, you can upload a file from the chat and then switch to the Data tab, where your uploaded data and your conversation with Squadbase AI are displayed side by side. This layout enables a more interactive and seamless data preparation experience.

We have moved the following configuration items from the Portal screen to the Editor screen.

• Outbound IPs
• Git Repository
• Database
• Environments

We’ve introduced Mode as a new parameter that controls how the AI agent behaves. When sending a chat request, you can now specify a mode to tailor the agent’s actions to your goal.

This release includes the following two modes, with more coming soon:

• General — A flexible mode suitable for a wide range of tasks. The AI agent follows your instructions to execute actions based on your prompts.

• Implement Dashboard — Optimized for creating new dashboard pages and analytical charts. This mode streamlines everything from dashboard design to execution.

Previously, chat-based setup was only available for BigQuery, Snowflake, and PostgreSQL. With this update, every Connection can now be fully configured through chat, enabling a faster and more conversational workflow.

You can now easily stay informed about Squadbase’s latest features and improvements directly from the “What’s New” panel. Access it anytime via the bell icon in the sidebar to browse recent releases and update details.

We have also added an Auto-Charge feature that automatically purchases credits when the remaining balance reaches a configurable threshold. The threshold can be freely adjusted based on your operation needs, enabling stable usage even as team demand increases.

The Purchase History page now allows you to review and manage both manual and automatic credit purchases. Credit purchase operations can only be performed by users with Admin permissions.

You can now perform major actions—such as setting up data sources and publishing applications—from the chat interface instead of the Editor.

Additionally, when you make a natural-language request like “I want to see the time-series trend of revenue by location,” the system automatically plans and executes the entire workflow, including SQL generation and dashboard creation.

With this update, the Task feature has been removed.

You can now seamlessly switch between Managed Repository (managed by Squadbase) and Linked Repository (connected to GitHub). In the Settings page, use the “Switch to Managed Repository” or “Connect to GitHub” buttons to toggle between configurations.

We’ve added the Component Selector, allowing you to select any component directly from the app preview and give instructions through chat.

You can now reference charts and tables on the screen as they are, making it easier to request additional analysis, modify visualizations, or inspect the underlying SQL.

Examples:

• “Add gender and age as axes to this chart and visualize it as a bar graph.”
• “Show me the SQL used in this chart.”

By interacting with components as you see them, you can enjoy a smoother and more intuitive conversational analytics experience.

Squadbase has become a Vibe Coding Platform for BI. All features previously offered in the Technical Preview are now generally available to all users.

Task

In data analysis workflows, ensuring the accuracy of data aggregation is critical — far more so than in typical web or app development. While creating a beautiful UI matters, verifying that the logic produces the intended results is the foundation of trustworthy analytics. This GA release introduces the Task system to support that process. Each step required for analysis is defined as an individual task, allowing users to complete the necessary validation without complex instructions for AI. By executing tasks in order, users can easily build reliable dashboards with minimal manual work.

Dashboards from Excel, CSV, and Jupyter

The GA version adds the ability to automatically generate dashboards from files such as Excel, CSV, and Jupyter notebooks. When a file is uploaded, Squadbase AI automatically preprocesses the data, prepares it for analysis, and builds the initial dashboard. Users can then update or extend the dashboard at any time using additional prompts.

Long-Term Memory and Business Knowledge

Squadbase AI now features a long-term memory system that learns and remembers business terms and context from previous user interactions. This allows users to have natural, context-aware conversations with the AI and receive insights based on accumulated knowledge. The Editor also includes an internal knowledge database that stores and refines user-specific information, enabling the AI to become more accurate and efficient the more it’s used. Through long-term memory and active communication with connected data sources via MCP, Squadbase AI autonomously plans and executes the steps required to build dashboards.

Free Viewer Licenses

Vibe Coding aims to empower everyone to become a dashboard creator. We believe that anyone involved in business decision-making should be able to explore and understand data directly. To support this vision, Viewer licenses are now free.This change allows dashboards created by teams to be shared more widely across organizations, promoting truly data-driven decision-making at every level.

Vibe Coding Platform for BI is now available as a Technical Preview, introducing a new way to “build dashboards with words.” This release adds the Editor, enabling users to connect data and create dashboards directly through natural language prompts..

Editor

The new Editor enables end-to-end dashboard creation and sharing directly in the browser. Squadbase AI builds dashboards in real time using leading frameworks such as Next.js, Streamlit, and Django. Users can view the process live, give additional instructions, and make adjustments as the dashboard is generated. No local development setup is required — dashboards can be collaboratively created, updated, and securely shared with the team in one click.

Data Connections and MCP Integration

The Editor includes pre-defined connectors for major SaaS applications and databases, allowing data to be fetched in real time. Through integration with the MCP server, the AI deeply understands each data structure and automatically generates high-accuracy SQL based on prompt instructions. This enables instant visualization and analysis without the need to manually specify data schemas or query details.

Django is now officially supported on Squadbase. Django is one of the most powerful and popular choices for building rich web applications in Python. With Squadbase, deploying a Django app takes just a single configuration file.

Here is an example of squadbase.yml for Django.

# squadbase.yml
version: '1'

build:
  framework: django
  runtime: python3.11
  package_manager: uv
  entrypoint: my_project.asgi:application

Django Dashboard Starter Template

Along with official support, we’ve also added a Django starter template to help you quickly get started building dashboards on Squadbase.

App Templates Released

We’ve added CRM Starter / Dashboard Starter templates to make it easier to start new projects.

Both Next.js and Streamlit starter templates are available. From the dashboard, you can select a template, automatically clone it to your GitHub repository, and deploy to Squadbase in one step.

More frameworks and template types will be added over time.

Build & Deploy Settings from the GUI

Previously, build and deployment settings had to be configured by writing a squadbase.yml file.

With this update, you can now manage build and deployment settings directly from the Setting tab of each project page in the dashboard.

This allows you to configure frameworks and cloud settings without preparing a YAML file.

Email Notifications for Deployments & Invitations

We’ve added email notifications for important Squadbase events to help you stay on top of your projects.

Notifications are sent when:

• A deployment is triggered by a Git push
• A deployment succeeds or fails
• You are invited to a project

You can manage these notifications from your account settings.

Deprecation of Connection Feature for Morph

As previously announced, the Connection feature (Project > Settings) has been deprecated in this release.

The dashboard entry points have been removed, and new connections can no longer be created via the UI.

However, existing APIs remain available for current users.

In addition to GitHub, Google accounts, and email login, you can now sign in with your Microsoft account.

This makes Squadbase easier to use in enterprise environments and ensures secure login aligned with your organization’s account management policies.

We’re excited to announce the new Deploy to Squadbase button, which lets you deploy any GitHub repository with a single click. One press handles the entire flow—clone, build, set environment variables, and go live on Squadbase. OSS maintainers can spin up instant sandboxes for newcomers, while internal teams can share apps with non-engineers without ever touching the CLI.

Snippets

You can add various URL query parameters to the Deploy to Squadbase button to control and customize the deployment experience for users, depending on your project's requirements.

Use the snippets below in your Git repositories or your dashboards for users to deploy.

[![Deploy to Squadbase](https://app.squadbase.dev/button.svg)](https://app.squadbase.dev/new/clone?repository-url=https://github.com/username/repo-name)

Highlights

• Drop a single markdown / HTML badge into a README, blog, or dashboard
• Configure the user flow with URL parameters such as project-name, subdomain, env-var-keys, root-path, and config-path
• Works with monorepos and private repositories accessible to Squadbase
• Ideal for OSS demos or sharing internal tools with non-engineers

→ For the full parameter list and advanced options, see the developer docs.

Squadbase now supports API Key authentication for accessing your deployed applications. Previously, while we supported secure app publishing by safely authenticating full-stack frameworks like Streamlit and Next.js, implementing server-side applications with frameworks such as FastAPI required separate deployments to AWS or Google Cloud. This was a challenge because it prevented users from fully benefiting from Squadbase's deployment experience when they wanted to use common backend processes across multiple applications.

This new feature makes it possible to extract common processes into server-side applications, enabling teams to operate larger-scale applications.

• Access to applications via API Key
• Official support for FastAPI as a framework in squadbase.yml

Of course, deployments using Dockerfiles still allow for the deployment of other server-side frameworks and API Key access.

See the resources below for more details:

• 📖 Blog Post = 📚 API Key Authentication
• 📚 FastAPI Deployment Guide

squadbase.yml for FastAPI

# squadbase.yml
version: '1'

build:
  framework: fastapi
  runtime: python3.11
  package_manager: uv
  entrypoint: main:app

How to Create and Access API Keys

Creating API Keys is straightforward through the Squadbase dashboard. When you create an API Key, you can specify the teams and projects it can be used with in detail. Since you can issue multiple API Keys for a single account, managing them by project ensures secure inter-app communication.

Squadbase now supports multi-environment deployments. Previously, you could deploy to only a single environment, but now we've introduced "Production" and "Preview" environments by default. Additionally, you can create as many custom environments as needed.

Multi-environment deployment is essential when running applications in production. By deploying features from "Staging" or "feature/" branches into separate environments, developers can collaborate easily without worrying about the production environment.

• Link GitHub branches to each environment using exact or wildcard matching.
• Log monitoring is available across all environments.
• User analytics are supported only in the Production environment.
• The Preview environment automatically tracks all branches not linked to any specific environment.

Check out the resources below to get started:

• 📖 Blog Post – Learn how Multi-Environment Deployments simplify your workflow.
• 📚 Documentation – Dive deeper into the setup and capabilities.

Shareable Links per Branch and Commit

From this release onwards, every deployment generates shareable URLs for each environment, branch and commit. Developers can review the impact of their changes in real-time, isolated environments.

You can find each shareable URL in the deployment details section of your deployments list.

AI App Showcase

Squadbase has just launched an AI App Showcase page to highlight the real-world potential of AI-powered applications. Through two sample apps—one built with Streamlit and the other with Next.js—you’ll get a hands-on sense of how AI can transform internal workflows.

Check out the Showcase page here → https://www.squadbase.dev/en/showcase/

Sales Coach AI Chat

A Streamlit-based coaching chatbot that pulls in Markdown files stored in your project’s codebase alongside meeting notes ingested from your CRM (Attio). By feeding it the right context, this chatbot offers tailored advice to salespeople and can be fine-tuned to mirror your own sales processes.

AI Analyst for GA4

A Next.js-powered data-analysis chat app connected to BigQuery with your Google Analytics data. Ask a question, and the AI generates the necessary Python analysis code, runs it safely on a virtual machine, and returns the results in real time. We’ve chosen E2B for the secure execution environment.

Collaborate and Iterate

AI-powered coding agents are ushering in an era of rapid, custom internal app creation—but true impact comes from running those apps in production and continuously iterating on them. To bridge that gap, we’ve built collaboration features for the operational phase.

1. AI-driven scaffolding that jump-starts your app
2. Automated deployments via Git integration
3. Unified runtime and access logs tied to user identities
4. Automated log analysis for actionable insights
5. In-app feedback channels linked directly to developers

In this blog, we share each features in detail.

User Authentication & Role Management

Project roles can be defined per deployed app, using arbitrary strings. These roles are primarily consumed in code to control data access and feature permissions.

Log Collection & Monitoring

Runtime and access logs are automatically collected and made searchable via the dashboard. This enables quick troubleshooting and helps you uncover insights for app optimization.

User Analytics

Squadbase combines its built-in authentication and logging features to analyze and visualize user activity, giving you actionable analytics right out of the box.

Feedback Collection

Squadbase can automatically inject a feedback comment box into deployed apps. Team members can submit feedback directly, and all responses are aggregated in the dashboard for developers to review and act on.

Deploying Next.js on Squadbase

Squadbase supports Next.js, the React-based full-stack web application framework. Vercel is typically the primary hosting environment for Next.js due to its excellent developer experience and high-performance infrastructure. However, Vercel may not always be ideal, especially when restricting public access and providing app access only to specific organizational users. Squadbase launches production-ready infrastructure for internal business apps and all its best practices—with a single click.

In this blog, we share how to make the most of Squadbase’s features when working with Next.js.

Deploying Streamlit on Squadbase

Squadbase supports Streamlit, a framework for building full-stack web applications with Python. Streamlit’s architecture makes UI updates instantaneous and coding straightforward, but production deployment still poses challenges. Standing up a robust environment often means stitching together ten or twenty AWS or GCP services—work that quickly burdens any dev team. Squadbase launches production-ready infrastructure—and all its best practices—with a single click.

In this blog, we share how to make the most of Squadbase’s features when working with Streamlit.

Introducing Squadbase, a New Cloud Platform for AI Apps

Today we officially opened Squadbase, a cloud platform focused on the safe, internal operation of AI applications.

Push code written in Streamlit, Next.js, LangChain, Mastra, or the Vercel AI SDK to Git and, within minutes, your app will be live at an internal-only URL. Behind the scenes, user authentication, RBAC, audit logs, and usage analytics are all wired up automatically. Anyone can build line-of-business apps, and once deployed, those apps keep improving through AI. That vision is the heart of Squadbase.

In this blog, we share the story behind the development of Squadbase, outline our vision, and explain why a new type of cloud platform is essential in today's evolving technological landscape.

What’s changed

🚀 New Features

• Plan Updates:
  • Free Plan now allows inviting up to 5 users with Viewer permissions.
  • Pro Plan users can invite additional Viewer-permission users for $7 per user.
• Users can now select cloud deployment targets (AWS, GCP) and configure CPU and memory resources.
• Added support for deploying Streamlit applications to Morph Cloud.
• Added support for multipart/form-data in the deployment environment.

🐛 Bug Fixes

• Fixed an issue where certain environment variable updates were not fully applied.
• Fixed a bug causing incomplete log details from the build server.
• Fixed display-related bugs during user invitations.
• Resolved pagination issues when the number of invited application members exceeded a certain limit.
• Corrected an issue where the latest morph-data package version was always selected during builds.
• Fixed a bug causing multiple unintended project creations.

🔧 Improvements

• Enhanced build performance by reducing snapshot frequency during Docker builds.
• Improved user experience when deploying from the dashboard.
• Improved build performance by incorporating frontend code builds directly into Docker build processes.

🚀 New Features

GitHub Login & Automatic Deployment via GitHub Integration

Users can now login to Morph Cloud using their GitHub accounts.

By integrating with GitHub, deployments can be performed automatically without the need for manual setup of CI/CD or using morph deploy via CLI.

Deploying Local LLMs with Ollama

With Ollama, users can deploy local LLMs such as deepseek-r1, llama3.2, phi4, and qwen on the Morph platform.

This allows for the development of AI applications utilizing LLMs while ensuring data privacy and security.

🛠 Maintenance

Stopped New Built-in PostgreSQL Instances

The built-in PostgreSQL feature will no longer be supported after the end of March.

As a result, the option to create new instances has been removed.

What’s changed

🐛 Bug Fixes

• Fixed environment-specific morph deployment failures.
• Fixed binary file response failures in project pages.

What’s changed

🚀 New Features

• Added public release settings for the app
• Added activity display feature for project creation and successful deployment
• Added usage history display feature for built-in DB for Pro plan and above

🐛 Bug Fixes

• Fixed multiple project role inputs
• Fixed developers couldn't invite project members
• Fixed project update time wasn't being updated

🔧 Improvements

• Display member icons in the project list
• Display member icons and available connector icons in project details

What's Changed

🚀 New Features

• Access control for individual members on created projects
• Role assignment for members within projects
• Permission management for using Database/SaaS connectors per project
• Execution log overview with filtering by date and time
• Display of deployment logs
• Fine-grained control for issuing API keys on a per-user basis
• Project search functionality
• Granular permission settings for members within teams

🔧 Improvements

• Enhanced performance and stability through infrastructure upgrades

🛠 Maintenance

• API key issuance changed from team-level to user-level
• Closing of the cloud editor feature
• Closing of the built-in PostgreSQL REST API feature
• Suspension of the job scheduling feature

What's Changed

🔧 Improvements

• In Viewer mode, "Jobs" tab now allows you to check logs for scheduled executions.

🐛 Bug Fixes

• Fixed an issue in the scheduled_jobs section of morph_project.yml where wildcard usage for days in CRON expressions was not working.

Launch Week #2 updates.

LLM Component

New Component for Pages

• Added the LLM component to build LLM apps like chat and text generation. Docs

Added Streaming Function to Python SDK

• Added morph_lib.stream.create_chunk to return function results as a text streaming format. Docs

Link: https://www.morph-data.io/blog/llm-component-launch-week-2

AI Functions

New Helper Functions for AI Integration

Added shorthand functions to simplify AI usage.

• Added text_to_sql function. Docs
• Added text_to_matplotlib function. Docs
• Added text_to_plotly function. Docs
• Added text_to_html function. Docs
• Added text_to_markdown function. Docs

Link: https://www.morph-data.io/blog/ai-functions-launch-week-2

Open-Source Framework

An open-source version of the Morph framework is now available for local use.

Github repository

• Integrated SQL, Python, and Pages frameworks into a single framework.
• Enabled morph serve command to launch data apps in development mode locally.
• Enabled morph sync command to deploy locally built data apps to the cloud.

Link: https://www.morph-data.io/blog/open-source-framework-launch-week-2

Viewer Mode

• Added a mode to display data apps built with Morph in view-only mode.

Link: https://www.morph-data.io/blog/viewer-mode-launch-week-2

Other Updates

New Integrations Added

• Added MS SQL Server as a connection option for SQL Connection.

Scheduled Execution Support

• The scheduling feature, previously available only in the cloud version, can now be configured in the framework via morph_project.yml. Docs

Link: https://www.morph-data.io/blog/other-updates-launch-week-2

This is the first public version of Morph.

Features

• Workspace
• Page
• Morph AI
• Data Connection
• Templates

More details can be found in the Launch Week #1 article.