18.5k stars! An open source AI programming assistant that browses web pages to collect information, and can also break complex instructions into multiple steps to complete.
Recently, there are many AI coding software, both open source and closed source, all with their own different characteristics.
Today I would like to introduce you to Devika. Its feature is that it can split instructions into small sequential tasks. And it also integrates AI search, which can collect information from the Internet to better complete tasks.
The UI also feels very comfortable.
Project introduction
Devika is an open source AI programming engineer capable of understanding high-level human instructions, breaking them down into steps, searching for relevant information, and writing code to accomplish a given goal. Devika leverages large language models, planning and reasoning algorithms, and web browsing capabilities to intelligently develop software to handle complex coding tasks without much manual intervention, from new feature development to the construction of entire projects.
Features
Multi-model support: Compatible with Claude 3 series, GPT-4, Gemini, Mistral, Groq and other AI models.
Advanced planning and reasoning capabilities: Advanced AI planning and reasoning capabilities.
Contextual keyword extraction: Conducting focused information research.
Seamless web browsing: Convenient information gathering capabilities.
Multi-language code writing: supports multiple programming languages.
Dynamic status tracking and visualization: Track agent status in real time and display it visually.
Natural language interactive interface: interact through chat interface.
Project management: Facilitates the organization and management of projects.
Scalable architecture: Supports new features and integrations.
System architecture
User Interface: A web-based chat interface for interacting with Devika, viewing project files, and monitoring agent status.
Agent core: The core component is responsible for coordinating the planning, reasoning and execution processes of AI and communicating with various sub-agents and modules to complete tasks.
Large-scale language models: Use state-of-the-art language models such as Claude, GPT-4, and native LLMs through Ollama for natural language understanding, generation, and inference.
Planning and Reasoning Engine: Responsible for breaking down high-level goals into actionable steps and making decisions based on the current context.
Research module: Use keyword extraction and web browsing capabilities to collect task-related information.
Code writing module: Generate code based on plans, research results and user needs, supporting multiple programming languages.
Browser Interaction Module: Enables Devika to navigate websites, extract information and interact with web page elements as needed.
Knowledge Base: Store and retrieve project-specific information, code snippets, and learnings for efficient access.
Database: Persists project data, agent status, and configuration settings.
Project link
https://github.com/stitionai/devika