genie

👻 genie

Genie is a Proof of Concept (POC) source code generator that showcases the potential of utilizing Large Language Models (LLMs) for code generation. As a limited prototype, Genie provides a glimpse into the capabilities of LLM-based code generation tools. It allows users to experiment with generating source code based on simplified prompts or descriptions.

Genie is based on https://github.com/smol-ai/developer.

Installing

You can install the pre-compiled binary in several different ways

deb/rpm/apk:

Download the .deb, .rpm or .apk from the releases page and install them with the appropriate tools.

manually:

Download the pre-compiled binaries from the releases page and copy to the desired location.

Building from source

Install a Go 1.20 compiler.

Run the following command in the checked-out repository:

make build

(Add the appropriate .exe extension on Windows systems, of course.)

Docker Support

Run the following command in the checked-out repository:

make docker-build

docker run -it --rm -v $PWD/dist:/dist -e OPENAI_API_KEY=$OPENAI_API_KEY genie openai -p "create a golang hello world"

How to use

Genie is a versatile code generation tool that simplifies the process of creating applications, APIs, and more. It offers an array of commands and features to cater to various code generation needs.

Usage:
  genie [command]

Available Commands:
  anthropic   Run codegen provided by anthropic
  completion  Generate the autocompletion script for the specified shell
  help        Help about any command
  openai      Run codegen provided by openAI

Flags:
  -h, --help            help for genie
  -o, --outdir string   outdir to use (default "dist")
  -p, --prompt string   prompt to use (required)
  -v, --version         version for genie

Use "genie [command] --help" for more information about a command.

OpenAI

Usage:
  genie openai [flags]

Examples:
genie openai -p "Create a python hello world"
genie openai -p prompt.txt

Flags:
      --api-key string        openAI api key
  -h, --help                  help for openai
      --max-tokens int        max tokens to use (default -1)
  -m, --model string          model to use (default "gpt-3.5-turbo")
  -t, --temperature float32   temperature to use (default 0.4)

Global Flags:
  -o, --outdir string   outdir to use (default "dist")
  -p, --prompt string   prompt to use (required)

Anthropic

Usage:
  genie anthropic [flags]

Examples:
genie anthropic -p "Create a python hello world"
genie anthropic -p prompt.txt

Flags:
      --api-key string      anthropic api key
  -h, --help                help for anthropic
      --max-tokens int      max tokens to use (default 4000)
  -m, --model string        model to use (default "claude-v1")
  -t, --temperature float   temperature to use (default 0.4)

Global Flags:
  -o, --outdir string   outdir to use (default "dist")
  -p, --prompt string   prompt to use (required)

Environment Variables

The following environment variables are supported by this project:

Example

This example instructs genie to use the _examples/aws_cdk/prompt.md file as the input prompt and generate the AWS CDK app based on the specifications provided in the prompt. The generated code will be saved in the _examples/aws_cdk/dist directory.

genie -p _examples/aws_cdk/prompt.md -o _examples/aws_cdk/dist

Outputs:

 ██████  ███████ ███    ██ ██ ███████
██       ██      ████   ██ ██ ██
██   ███ █████   ██ ██  ██ ██ █████
██    ██ ██      ██  ██ ██ ██ ██
 ██████  ███████ ██   ████ ██ ███████

Create list of files:
• src/api/api.ts
• src/api/todoController.ts
• src/lambda/todoHandler.ts
• src/dynamoDB/todoTable.ts
• .env
• .gitignore
• README.md
• cdk.json
• jest.config.js
• package.json
• projenrc.json
• tsconfig.json

Reasoning:
• The app requires an API Gateway to expose the CRUD endpoints.
• The Lambda functions will handle the API requests and interact with DynamoDB, so we need to create a Lambda Backend.
• DynamoDB is needed to store the todo items.
• We'll use 'projen' to manage the AWS CDK project and its dependencies.
• To implement the CRUD operations, we'll need to create the necessary API endpoints.

Create list of shared Dependecies:
• {dataSchemas Shared data schemas for the request and response bodies [TodoItem CreateTodoRequest CreateTodoResponse GetAllTodosResponse GetTodoByIdResponse UpdateTodoRequest UpdateTodoResponse]}
• {environmentVariables Shared environment variables for configuring DynamoDB table name and settings [DYNAMODB_TABLE_NAME]}
• {errorHandlingUtilities Shared utilities for error handling [createErrorResponse]}
• {awsCdkLibraries Shared AWS CDK related libraries [aws_cdk aws_lambda aws_apigateway aws_dynamodb]}

Reasoning:
• To identify the shared dependencies between the generated files, we need to analyze the content of each file and look for common symbols or entities used across multiple files.
• Starting with the API endpoints, there might be shared data schemas for the request and response bodies.
• The todoController.ts file is responsible for handling the API requests and might have dependencies on the data schemas.
• The todoHandler.ts file implements the Lambda functions for CRUD operations and might have dependencies on the data schemas as well.
• The todoTable.ts file creates the DynamoDB table and might have dependencies on the data schemas, as it needs to define the primary key.
• Other shared dependencies could include environment variables, error handling utilities, and AWS CDK related libraries.
• We'll analyze the content of each file and identify the shared symbols to create the shared_dependencies list.

File src/lambda/todoHandler.ts created
File src/dynamoDB/todoTable.ts created
File .env created
File src/api/api.ts created
File .gitignore created
File src/api/todoController.ts created
File cdk.json created
File jest.config.js created
File package.json created
File tsconfig.json created
File projenrc.json created
File README.md created

Tokens Used: 28404
Prompt Tokens: 24452
Completion Tokens: 3952
Successful Requests: 14
Total Cost (USD): $0.00

Contributing

Contributions are welcome! Feel free to open an issue or submit a pull request for any improvements or new features you would like to see.

References

• https://github.com/smol-ai/developer
• https://github.com/hupe1980/golc

License

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