DGi

What is DGi?

DGi is an AI data agent developed by Datagran, intended for developers and corporate teams that need to integrate data analysis, code execution, and automation capabilities into their products. The old record domains now point to DGi’s new website and API documentation; the product’s role has shifted from a general-purpose AI programming tool to a data agent platform.

It can handle natural language tasks, connect to files, databases, APIs, or web pages, generate and execute code, and then return analysis results, files, charts, or previews of applications. DGi is not merely a code completion plugin, nor should it be regarded as an automatic decision-making system that requires no verification.

Overview of key capabilities

Data analysis and file processing

Users can upload CSV files during a session, and then use natural language to request that DGi clean, analyze, visualize the data, or generate reports. The session upload endpoint of the current API is specifically designed for CSV files; although the homepage shows data sources such as CSV and Excel, developers should not assume that all file formats can be uploaded through this same endpoint.

• Sessions organize related questions, files, and tasks within the same context.
• Asynchronous tasks can return progress messages, the final answer, a list of codes, and a preview link.
• The generated files can be listed, downloaded, or updated in the session directory.
• In conclusion, it is still necessary to verify the field definitions, missing values, time range, and statistical methods.

Code generation and execution

DGi can create a directory for Python code based on a specific task, with script.py serving as the main entry point. Developers can read or modify the files in this directory, and then use run_file to create new tasks to execute that code.

1. Create a session and submit data or connection details.
2. Describe the analysis objectives and desired output in natural language.
3. Save the directory where the code execution takes place, as returned by the task.
4. Check the generated code, dependencies, and file access permissions, and make modifications if necessary.
5. The task is initiated by executing script.py, which is located in the directory.
6. Poll the messages to view the logs; if there are no messages for the code task, then read the full output.

Running AI-generated code introduces risks such as dependencies, network access issues, high resource consumption, and data corruption. In production environments, isolated execution environments, credentials with minimal permissions, limits on resource usage, and human approval processes should be employed.

Database and external connections

The product page lists the methods for connecting to PostgreSQL, MySQL, BigQuery, Snowflake, object storage, and files; the API also enables the creation of connections, as well as querying, updating, and sharing access rights. Connection objects can store database or API credentials, and access to them can be granted or revoked on a per-user basis.

• Database accounts should preferably be given read-only permissions to prevent proxies from accidentally altering production data.
• Different credentials and API keys should be used for development, testing, and production environments.
• Shared connections allow other users to view the connection and credentials; it is necessary to verify the organization and individuals before granting such access.
• Credential rotation, retrieval upon employee departure, and alerts for abnormal access all require to be covered by corporate processes.

Scheduled tasks and workflows

DGi allows the executable files involved in a session to be combined with Cron expressions or time descriptions in natural language to create scheduled tasks. The API enables creation, listing, activation, deactivation, and manual triggering of such tasks, making it suitable for generating periodic reports, checking inventory levels, and monitoring data.

The scheduling functionality should include features such as idempotency, timeout handling, retry on failure, and alerting for results. When external write operations are involved, approval, auditing, and rollback mechanisms should also be in place for each execution.

Web Automation and Application Development

Web automation capabilities enable navigation through pages, extraction of information, filling out forms, and execution of multi-step tasks; the results of these tasks may be displayed in the browser interface. Website login processes, verification codes, terms of service, robots.txt rules, and anti-automation measures still affect the actual success rate.

DGi can also generate data dashboards, reports, or previews of interactive applications. These previews are useful for verifying requirements, but before actual deployment it is still necessary to check permissions, performance on mobile devices, version dependencies, input validation, and risks of data leakage.

Understanding: Building the project context

Understanding can establish a long-term context from GitHub repositories, support documentation URLs, or Markdown files. For in-depth analysis tasks, it examines the codebase and generates PDF documents; the estimated processing time indicated on the page is between 5 and 30 minutes.

1. Prepare a dedicated GitHub token for the private repository and store it in the connection object.
2. Create an Understanding and fill in the repository, documentation, or Markdown materials.
3. Initiate a deep research task and poll its status.
4. After the task is completed, read the generated document and use the understanding_id for subsequent questions and answers.
5. After changes to warehouse permissions or personnel, revoke the tokens promptly and update the context.

Private repository analysis involves incorporating code and documentation into external processing workflows. Companies should first determine the retention period, the sub-processors involved, the purposes of training, the methods for deletion, and the contractual responsibilities.

API v2 integration process

The current public API uses Bearer API keys and adheres to the OpenAPI specifications. The core process involves creating sessions, initiating asynchronous tasks, and polling for messages based on timestamps until their status changes to completed or interrupted.

1. Register an account and create an API key on the profile page.
2. Call the test endpoint to verify authentication and service status.
3. Create a session and save the session_id.
4. Create questions, run files, or conduct research tasks within a session, and save the task_id.
5. Poll using the timestamp from the previous message to avoid pulling all messages repeatedly.
6. Use backoff and retry for server errors, and stop polling once it is completed or interrupted.
7. Read the final answer, generate a file, output code, or preview the application.

The certification guidelines still contain old domain names and v1 examples, while the current API Reference uses new domain names and v2 paths. When implementing something, it is necessary to follow the current OpenAPI and API Reference – there should be no mixing of different examples.

API, SDK, and open-source status

Price and usage limits

The number of requests per minute represents only the interface rate; it is not equivalent to the number of tasks that can be executed free of charge, the number of model tokens available, the time required for code execution, or the storage capacity allocated. At present, no information regarding the specific amounts, the price for exceeding limits, the trial period, or the rules for refunds is available. It is necessary to verify these details in the account dashboard or in the contract before making a purchase.

Safety and privacy considerations

The product page lists security features such as double-envelope encryption, AES-256 static encryption, TLS 1.3, zero trust, audit logs, SOC 2, and private clouds. At present, no audit reports, certificate details, or comprehensive security whitepapers that are available for public review can be found; therefore, when purchasing such products, it is necessary to verify the applicable products, regions, and validity periods.

• API keys must not be included in browser code, published in repositories, or recorded in logs; they should be stored in a key management system and rotated regularly.
• Create dedicated, least-privilege credentials for databases, object storage, and GitHub.
• Remove any irrelevant personal information before uploading the data, and verify the cross-border transmission as well as the sub-processors.
• It is required to specify the retention and deletion deadlines for sessions, files, code, browser recordings, and backups.
• Human approval should be required for high-risk code execution, web page modifications, and automated decision-making.
• Create separate audit records for task, connection sharing, and organizational member changes.

The current page of DGi does not provide clear, separate links to its privacy policy and terms of service; the relevant policy updates available on Datagran are quite old. When dealing with sensitive or regulated data, it is necessary to obtain, prior to signing a contract, the current data processing agreements, security attachments, refund policies, and deletion procedures specific to DGi.

Suitable for users and typical use cases

• Data Engineer: Generates and runs data processing scripts using natural language.
• Analysis team: Quickly generate trends, charts, and reports from CSV or databases.
• Product developers: Integrate data proxies into their own platforms via APIs.
• Operations team: Generates reports and monitoring metrics on a scheduled basis, or automatically collects web data.
• R&D team: Analyze code repositories and establish reusable project knowledge contexts.
• Enterprise IT: Developing internal data tools in line with requirements regarding permissions, auditing, and private deployment.

Advantages and capabilities boundaries

Main advantages

• It extends from natural language tasks to code, data connections, scheduling, and application previews.
• APIs cover session, task, file, connection, organization, and knowledge contexts.
• Asynchronous tasks provide progress messages, which facilitate integration into the product interface.
• It supports scheduled execution and connection sharing, making it suitable for team-based data workflows.

Main limitations

• The analysis and code generated by AI may contain errors, so manual review and isolated execution are required.
• The public prices, task quotas, retention periods, and refund conditions are incomplete.
• Some certification examples use old domain names, which are inconsistent with the current v2 references.
• The security and compliance statement lacks verifiable evidence of the audit scope that is available for public inspection.
• No official SDK, product source code, or native mobile applications have been confirmed.
• Web automation is affected by website rules, login verification, and page changes.

Frequently Asked Questions

Is DGi a code editor?

No. It is a data proxy platform that can generate and execute Python code, but its main functions lie in data analysis, automation, connectivity, and API integration.

Can DGi be used for free?

The certification documents specify the Free tier and a limit of 100 API requests per minute. The number of free tasks, as well as the available computing resources and storage space, are not fully disclosed; the information provided in the account dashboard should be referred to.

How much is the Pro version?

No public figure for the price is available at the moment; you need to contact sales or check the actual settlement page. Do not use historical prices from third parties as the current quote.

Which data sources are supported?

The page lists PostgreSQL, MySQL, BigQuery, Snowflake, object storage, CSV, and Excel, among others. The specific permissions for each connector as well as the available regions must be verified in the account.

Why can’t API tasks return results immediately?

DGi tasks are executed asynchronously. After creating a task, it is necessary to save the task_id, and then keep polling for messages until the task is completed or interrupted.

Can I run the code that I have modified?

It is possible to update the files in the code directory, and then run the program by using script.py as the entry point. Before execution, the code should be reviewed, and an isolated environment along with minimal permissions should be used.

Is DGi open source?

It is not a verifiable open-source platform. The availability of public REST APIs, OpenAPI specifications, and example code does not mean that the source code of the product is made available.

Is there an official SDK?

The document on navigation mentions SDKs and client libraries, but at present the available download packages, languages, versions, and licenses have not been specified. Developers can first create clients based on OpenAPI.

Can a security statement be used directly as proof of compliance?

No. Enterprises should obtain the audit reports, the scope of the certificates, the data processing agreements, and a list of sub-processors, and ensure that they cover the DGi services comprehensively.

Will DGi use the data for model training?

The current public page does not provide clear enough information regarding the current regulations. Before sensitive data is incorporated, it is necessary to establish through a contract the purposes of its use, the duration for which it will be stored, the procedures for its deletion, and the scope of processing by third parties.

Summary

DGi is suitable for development teams that wish to integrate data analysis, code execution, scheduling, and automation into their products through APIs. Its public APIs offer a fairly comprehensive set of functions, but the pricing, data policies, SDKs, and compliance requirements still need to be verified before procurement and deployment.