Morph

Morph 1.0

Morph AI - Launch Week #1

2024-10-30
Keita Mitsuhashi
Keita Mitsuhashi
Co-founder & COO

Hello everyone! On the third day of our launch week, we'd like to introduce the AI features that assist in building data apps. Let's dive right in!

AI Features Specialized for Data Tasks

While general AI chat tools can perform analysis, Morph AI is specifically designed for data tasks, providing more accurate results for your requirements. Here's how it differs from standard AI chat:

  • Understands Data Schemas: Morph AI comprehends the schema (structure) of your database, allowing it to grasp relationships between tables and the meanings of columns. This understanding enables accurate analysis that considers data relationships, allowing you to quickly obtain appropriate results even for complex queries.
  • Considers Schemas of Previous and Subsequent Data Processing Results: It keeps track of the schema generated at each step of data processing and accordingly passes it to the next step. This ensures efficient data processing while maintaining consistency throughout the entire pipeline.
  • Allows You to Provide Knowledge Like Column Meanings and Explanations of Proper Nouns: Users can supply the AI with knowledge about the meanings and uses of columns, enabling the AI to execute more precise queries and analyses.

Pipeline Builder

By invoking Morph AI on the canvas, you can use the pipeline construction feature. When you give instructions to Morph AI, it automatically builds a multi-step data pipeline consisting of multiple SQL and Python files.

The specific steps are as follows:

  1. The user provides instructions.
  2. Based on the instructions, Morph AI plans the SQL and Python files that should be included in the data pipeline and their processing content.
  3. It generates the content of the files based on this plan.
  4. The generated files are executed sequentially. If an error occurs during the process, it automatically fixes it based on the error message.

Since users can modify the SQL or Python generated by Morph AI later, you can flexibly build data pipelines tailored to your specific needs.

By using Morph AI's pipeline construction feature in this way, you can significantly reduce the cost of building extensive data processing tasks.

Code Generation

You can add processing steps as SQL and Python files placed on the Canvas. Instead of writing code from scratch, you can request Morph AI to generate an initial version.

Morph AI generates code based on your instructions with an understanding of the results of prior data processing. This greatly reduces the cost when adding new data processing steps.

Code Editing

Morph AI excels at code editing as well! For Python and SQL you can request detailed adjustments to data processing, and for MDX you can ask for style changes.

For example, in Python, there are many convenient libraries for data processing, but understanding their APIs in detail can be challenging. Even in such cases, by requesting Morph AI to edit, you can smoothly upgrade your code.

Future Roadmap

While we are confident that Morph AI already has powerful features to support building data applications, our journey in AI has only just begun. We have several concrete ideas for updating Morph AI.

First is further improvement in accuracy.

As we provide Morph and gain real-life data on best practices in data app construction, the problems that Morph AI should solve will become narrower and more specific. We aim to adjust Morph AI accordingly and strive for further accuracy improvements.

Next is updating the method of providing context to Morph AI.

When you try to construct a data pipeline using natural language, you realize the need to handle context such as proper nouns and internal terminology. Currently, if it's expressed in the data schema, Morph AI can understand it… but it would be desirable to provide context from documents stored in PDFs or text files. We are also actively considering integration with existing data catalogs.

That’s all for today! Tomorrow, we will discuss integration with external data sources.