TransacAI Insights Generation Service

This project is the codebase for the Insights Generation Service (IGS) of the TransacAI project.

TransacAI

TransacAI project is geared towards generation of enriched summaries and insights of transactional data in real-time or batch using Generative AI and Large Language Models (LLMs). It goes beyond visual and analytical processing of transactional data by generating context-aware and enriched insights in natural language using LLMs. It focuses on delivering human-centric analysis that is easier to understand and act upon, eliminating the need for multiple complex data processing steps to derive insights from raw data.

Insights Generation Service (IGS)

IGS is one of the core service of the TransacAI project. The primary goal of this service is to generate enriched insights from transactional data using Generative AI and Large Language Models (LLMs). The service is designed to be scalable and efficient to handle real-time and batch processing of transactional data.

IGS uses gRPC to communicate with the Prompt Builder Service (PBS) of the TransacAI project, to get the prompt to be used to generate LLM response.

Core functionalities of the IGS are:

1. Receives a request through gRPC from the Workload Manager Service (WMS) to generate insights from transactional data.
2. Uses gRPC to communicate with the Prompt Builder Service (PBS) to get the prompt to be used to generate insights using Large Language Model (LLM).
3. Generates insights from LLM using Gemini API or OpenAI API through QvikChat.
4. Save the generated insights in the database.
5. Send update on Kafka topic to inform clients about the generated insights.

In a nutshell, flow looks like this:

1. WMS -> IGS
2. IGS <- PBS
3. IGS <- LLM
4. IGS -> Database
5. IGS -> Kafka

This project was setup using QvikChat Starter Template. It comes pre-configured with the following features:

• QvikChat: QvikChat installed and configured to start serving chat endpoints.
• TypeScript: TypeScript to allow you to write type-safe code efficiently.
• ESLint: ESLint to enforce code quality and consistency.
• Prettier: Prettier to format your code automatically and ensure consistent code style.
• Jest: Jest to run your tests and ensure your code works as expected.
• GitHub Actions: GitHub Actions to run your tests and lint your code automatically on every push.
• SWC: For faster and more efficient TypeScript compilation.
• PNPM: PNPM to manage your dependencies efficiently.

Local Development

Prerequisites

Clone the repository.

git clone https://github.com/pranav-kural/transacai-insights-generation-service.git

Setup Environment Variables

Create a .env file in the root of the project and add the following environment variables:

GOOGLE_GENAI_API_KEY=

Alternatively, you can copy the .env.tmp file or rename it to .env and fill in the values.

By default QvikChat uses the Google GenAI, so to use QvikChat with default settings, you need to provide the GOOGLE_GENAI_API_KEY. You don't have to set values for other environment variables if you are using the default settings.

Add value to the OPENAI_API_KEY variable if you're going to use OpenAI models.

Running the Project

You can run the following commands to get started:

pnpm install
pnpm dev

Testing

This project comes with Jest pre-configured to run your tests, and some tests predefined in the src/tests directory. You can run the tests using the following command:

pnpm test

Please ensure you have the environment variables set up before running the tests.

By default, Jest is configured to test the source code in the src directory. You can change the configuration in the jest.config.js file, along with any other Jest configurations you may want to change.

Genkit Developer UI

You can run the Genkit developer UI to test the response generation through GUI interface. You can know more about the Genkit Developer UI here.

Start the Genkit developer UI:

npx genkit start

OR, you can install the Genkit CLI globally:

npm i -g genkit

Then start the Genkit developer UI:

genkit start

Linting

This project comes with ESLint pre-configured to enforce code quality and consistency. You can run ESLint using the following command:

pnpm lint

You can also run ESLint in watch mode using the following command:

pnpm lint:watch

Formatting

This project comes with Prettier pre-configured to format your code automatically and ensure consistent code style. You can run Prettier using the following command:

pnpm format

Building

You can build the project using the following command:

pnpm build

This will compile the TypeScript code in the src directory and output the compiled JavaScript code in the lib directory.

gRPC

IGS exposes a GenerateInsights method through gRPC which can be used by the Workload Manager Service (WMS) to initiate a request to generate insights. This method only returns an object for acknowledgement with a property received set to true to indicate that the request has been received successfully. The actual insights generation process is asynchronous and the generated insights are sent to the Kafka topic.

Proto file for the service is located at src/rpc/igs.proto.

Testing Locally

You can use the grpcurl CLI tool to test the gRPC service locally.

To list services:

grpcurl -plaintext 0.0.0.0:50051 list

Please note that reflections API is disabled in production, so you won't be able to list services in production.

To call the GenerateInsights method (remember to change reqId to a unique value), you can use a command similar to below:

grpcurl -plaintext -d '{"reqId":"ghgfjerherw","clientId":"test_client","promptId":2,"recordsSourceId":"SUPABASE","promptTemplatesSourceId":"SUPABASE","fromTime":"2019-12-29T06:39:22Z","toTime":"2019-12-29T23:49:22Z"}' 0.0.0.0:50051 igs.InsightsGenerationService/GenerateInsights

Above request uses test client and promptId of 2, with sources for both records and prompt templates as SUPABASE. You can change the values as per your requirement.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Issues

If you encounter any issues or bugs while using this project, please report them by following these steps:

1. Check if the issue has already been reported by searching our issue tracker.
2. If the issue hasn't been reported, create a new issue and provide a detailed description of the problem.
3. Include steps to reproduce the issue and any relevant error messages or screenshots.

Open Issue