hackedbyagirl/program-engineer-gpt

Interact directly with code repositories in a chat or automate the development of your code based just on a project description

program-engineer-gpt's Issues

[BUG]: Deeplake Error in Analyze Mode

Overview

When running the programengineergpt.py in analyze mode, a deeplake.util.exceptions.BadRequestException is being thrown when trying to recieve index from deeplake or create a new index. This error is happening in the get_dataset_credentials function in deeplake/client/client.py. The specific error message is:

"Invalid Request. One or more request parameters is incorrect. An error occurred (NoSuchEntity) when calling the AssumeRole operation: The entity in your request cannot be found."

This suggests that there is an issue with the parameters being passed to get_dataset_credentials. It also seems like an AWS IAM role that does not exist is being assumed, as indicated by the NoSuchEntity error.

Expected Behavior

You should enter a chat with any indexed code that is uploaded or queried from an existing index.

Current Behavior

An error occurs in deeplake/client/client.py at the get_dataset_credentials function, causing a deeplake.util.exceptions.BadRequestException to be thrown. This is due to a thrid party module.

Steps to Reproduce

1. Run programengineergpt.py in analyze mode
2. select to upload a new index or use an existing deeplake repo

Context (Environment)

• Python 3.10
• MacOS

[Feature] Integrate Saving conversation to Memory

Feature Request

Saving the conversational history so you can reference chat history and generated content

Possible Solution

saving to a local directory

Add Web Interaction

Description

Use the LangChain's tools to interact with the web for additional context. This might involve using the Search tool to search the web for information related to a user query, or the Requests tool to retrieve specific content from a website.

[Feature] Install requirements and file instructions

Feature Request

Include an action that performs writing a install file which includes all the command line code you need to setup the code base for running

Possible Solution

Add an action to the Actions class

Create Custom Agent

Description

Create a custom agent that uses the GPT model to analyze the code and answer user queries. This involves creating a PromptTemplate that instructs the model on how to analyze the code and answer queries, and an OutputParser that parses the model's output into a format that can be used by the agent.

Integrate Features

Description

Integrate all these components into a single system that can load and embed code, create a conversation over the code, answer user queries about the code, and interact with the web for additional context. This might involve creating a main function or class that initializes and coordinates all these components.

Create Modular Interface

User Story: Create Application Interface

As a developer or user of CodeAssistantGPT, I want an interactive interface when I launch the application, so that I can easily understand the capabilities of the tool and select the mode that best suits my needs.

Done Statement

The story is done when the user can:

• Launch the application and be presented with a general description of the tool and its capabilities.
• View a list of available modes: Analyze, Develop, Generate**, and Debug**.
• Select a mode to enter based on their needs.

Tasks

1. Design and implement an application entry point that provides a general description of the tool.
2. Implement a mechanism to list available modes and allow the user to select one.
3. Implement the functionality for each mode:
   • Analyze: Analyze a code repository based on a GitHub link, a local path to a code repository, or a coded index.
   • Develop: Assist with the developing of a new or existing coding project..

Steps to Complete

1. Design the layout of the application entry point and the interaction flow.
2. Implement the application entry point with a general description of the tool.
3. Implement the mechanism to list available modes and handle user selection.
4. Implement the functionality for each mode.
5. Test the application to ensure it works as expected.
6. Refine and iterate based on feedback and testing results.