Skip to main content

CodeMie's capabilities

AI/Run CodeMie efficiently addresses numerous tasks across varying difficulty levels. Below is an overview of AI/Run CodeMie's key capabilities:

Core Features​

Comprehensive SDLC Smart Assistance and Assistants Library​

AI/Run CodeMie offers robust smart assistance across all phases of the SDLC process by leveraging a variety of AI assistant roles, such as Business Analyst (BA), Developer, Quality Assurance (QA), Project Manager (PM), and more. These pre-built AI assistants enhance performance and productivity, and automate routine work, significantly reducing process costs and accelerating the software development cycle. The platform comes with a comprehensive library of pre-built AI assistants tailored to various roles within the SDLC to suit the diverse needs within a project.

Assistants Constructor​

Provides the flexibility to create personalized assistants equipped with specific tools and abilities tailored to your project's needs.

Indexing and Data Management​

AI/Run CodeMie provides options for data indexing, including the ability to:

  • Monitor the current progress and status of the indexing process
  • Perform incremental or full reindexing
  • Manage indexed data sources effectively

Supported data sources include Jira, Confluence, various file formats (PDF, PPT, Excel, etc.), and Git.

Support for Multi-Agent Workflows​

AI/Run CodeMie supports multi-agent workflows, allowing multiple AI assistants to collaborate seamlessly on complex tasks and workflows. This capability covers use cases where different agents need to interact and share information to achieve a common goal, enhancing coordination and efficiency across various phases of the SDLC.

Ease of Use for Beginners​

Simple use cases for newcomers include:

  • Code review
  • Newcomer training
  • User story creation

These require minimal setup, such as configuring your Git token for code-related tasks or your Jira token for project management tasks.

Extensive Library of Tools​

AI/Run CodeMie includes a wide array of tools to support various aspects of software development and project management:

Version Control Systems (VCS)​

Tools for managing and tracking changes in the codebase, such as Git.

Codebase Tools​

Tools for code review, static code analysis, and automated code formatting.

Research Tools​

Tools to assist in gathering and organizing research data and documentation.

Cloud Tools​

Integration with major cloud providers (AWS, GCP, Azure) for deployment, monitoring, and management of cloud resources.

Project Management Tools​

  • Jira: For project management, task tracking, and issue tracking
  • Confluence: For documentation, knowledge sharing, and collaboration

Additional Tools​

  • Open API: Integration with various open APIs to extend functionality and connect with other services
  • Notification Tools: Tools for sending notifications and alerts via email, chat, or other communication channels
  • Data Management Tools: Tools for querying Elasticsearch indexes
  • Access Management: Keycloak integration
  • Plugin System: Extensible plugin architecture
  • File Management: Tools for file operations
  • Quality Assurance: Testing and QA tools

Best Practices​

Data Source Usage​

  • There is no priority or sequential system in place. Everything depends on the given instructions
  • You can instruct the model to use a specific data source for a particular use case
  • Provide a description for the data source when it is created
  • Data source descriptions are provided to the model so it can understand better use cases for it

System Instructions​

  • System Instructions (System prompt) extends based on data source description
  • If contradictions arise, the model will use its creative problem-solving abilities to address them
  • The data source does not have rules, only description

Context and Queries​

  • The model has a context window, but it is irrelevant to data source size
  • The model answer will depend on query quality
  • With queries that are specific and on point, there are no problems even with thousands of data sources
  • If a poor query is provided (e.g., "tell me about something"), the answer would be vague

Instructions Guidelines​

  • The smaller and clearer the instructions, the better
  • Instructions must be uncontradictory, as this will reduce the risk of confusing LLM
  • Other than the context window, there is no limit