AI for Code
Code generation, review, debugging, and developer tools
48 entities in this channel
HumanEval Dataset
by OpenAI
A curated set of 164 handwritten Python programming problems released by OpenAI, each consisting of a function signature, docstring, reference solution, and unit tests. HumanEval introduced the pass@k metric for functional code correctness evaluation and has become the de facto standard benchmark reported in virtually every code generation model paper.
GitHub Copilot + VS Code
by GitHub
GitHub Copilot integrates into VS Code as a first-party extension, delivering inline ghost-text completions, multi-line suggestions, and a dedicated Copilot Chat panel for conversational refactoring, test generation, and documentation. It leverages Codex and GPT-4 models under the hood, with workspace-aware context from open tabs and the current file.
Code Generation
by AaaS
Generates functional code from natural language descriptions, specifications, or partial implementations. Covers multiple languages and frameworks with support for boilerplate scaffolding, algorithm implementation, and API integration patterns.
SWE-bench Verified
by Princeton NLP
Human-validated subset of SWE-bench containing 500 problems verified by software engineers for correctness, clarity, and solvability. Provides a more reliable signal than the full SWE-bench by filtering out ambiguous or under-specified issues.
MBPP (Mostly Basic Python Problems)
by Google
A dataset of 974 crowd-sourced Python programming problems suitable for entry-level programmers, each with a problem description, code solution, and three automated test cases. MBPP complements HumanEval by covering a broader variety of programming concepts and is widely used alongside it for comprehensive evaluation of code generation capabilities across model families.
The Stack v2
by BigCode
An expanded code pretraining dataset containing 3 trillion tokens of source code in 619 programming languages, curated by BigCode from GitHub repositories with permissive SPDX licenses. Version 2 triples the size of the original Stack and includes improved deduplication, opt-out mechanisms for authors, and structured data from GitHub issues and pull requests alongside raw source files.
MBPP
by Google Research
Mostly Basic Programming Problems — a collection of 974 crowd-sourced Python programming tasks with natural language descriptions and test cases. Tests foundational programming ability including string manipulation, list processing, and basic algorithms.
Code Review
by AaaS
Analyzes code for bugs, security vulnerabilities, performance issues, and style violations. Provides actionable feedback with severity levels and suggested fixes aligned to language-specific best practices and project conventions.
CodeSearchNet
by GitHub / Microsoft Research
A dataset and benchmark challenge for code retrieval and search containing 2 million (code, documentation) pairs in six programming languages — Python, Java, JavaScript, PHP, Ruby, and Go — curated by GitHub and Microsoft Research. It is the canonical benchmark for code-to-natural-language and natural-language-to-code retrieval tasks and is widely used to evaluate code embedding models.
APPS (Automated Programming Progress Standard)
by UC Berkeley
A benchmark of 10,000 programming problems at introductory, interview, and competitive programming difficulty levels, each with problem statements, test cases, and human-written solutions. APPS is the standard dataset for evaluating code generation models on realistic programming tasks ranging from simple loops to complex algorithmic challenges drawn from competitive programming platforms.
StarCoderData
by BigCode
The 780 billion token code dataset used to pretrain the StarCoder family of models, assembled by BigCode from The Stack v1 spanning 86 programming languages with permissive licenses. It includes GitHub issues, Git commits, and Jupyter notebook data alongside source files, enabling models to learn from developer workflows and not just static code.
Cursor + OpenAI
by Anysphere
Cursor is a VS Code fork that uses OpenAI's GPT-4 and o-series models as its reasoning engine for multi-file edits, semantic codebase search, and an agent mode that can autonomously implement features across the entire repository. It offers a Composer panel for multi-file diffs and a codebase-aware chat that indexes the project with embeddings for precise retrieval.
GitHub Copilot Workspace
by GitHub (Microsoft)
GitHub's AI-native development environment that turns issues into fully implemented code changes. Plans, implements, and validates multi-file edits with human-in-the-loop review before merging.
Code Debugging
by AaaS
Diagnoses and resolves software bugs by analyzing error messages, stack traces, and code behavior. Applies systematic debugging strategies including root cause analysis, state inspection, and targeted fix generation with regression awareness.
Claude Code
by Anthropic
Claude Code is an agentic AI coding assistant from Anthropic designed to operate within a developer's terminal. It autonomously handles complex software development tasks by understanding entire codebases, editing files, executing shell commands, and managing Git workflows, acting as a hands-on pair programmer with minimal human supervision.
GitHub Copilot + JetBrains
by GitHub
The GitHub Copilot plugin for JetBrains IDEs integrates AI-powered code completion and a conversational chat panel directly into the editor. It provides inline, ghost-text suggestions and mirrors the functionality of the VS Code extension, adapting to JetBrains' native keymaps and user interface for a seamless experience across IDEs like IntelliJ IDEA and PyCharm.
Feature Importance Analyzer
by Community
Analyzes feature importance for scikit-learn compatible models using multiple advanced techniques. It computes SHAP values with Tree and Kernel Explainers, calculates permutation importance, and performs feature selection with Boruta. Results are compiled into an interactive HTML dashboard for easy interpretation and sharing.
REST AI API Template
by Community
Production-ready FastAPI template for AI-powered REST APIs, with pre-wired OpenAI/Anthropic client, async streaming endpoints, JWT authentication, rate limiting, structured logging, and OpenAPI docs. Includes Docker Compose stack with Redis rate-limit store and Prometheus metrics.
SWE-agent
by Princeton NLP
Princeton NLP's research agent that turns LLMs into autonomous software engineers. Achieves state-of-the-art results on SWE-bench by providing an agent-computer interface optimized for code navigation and editing.
Test Generation
by AaaS
Automates the creation of test suites by analyzing source code, function signatures, or specifications. It generates unit tests, integration tests, and edge case scenarios for popular frameworks, complete with necessary mocks and assertions. This accelerates development cycles and improves code reliability.
Code Refactoring
by AaaS
Code Refactoring is the disciplined process of restructuring existing computer code without altering its external behavior. It focuses on enhancing nonfunctional attributes like readability, maintainability, and performance. This practice is key to managing technical debt, applying design patterns, and modernizing legacy systems to align with current best practices.
DeepSeek-Coder-V2
by DeepSeek
DeepSeek-Coder-V2 is a powerful open-source Mixture-of-Experts (MoE) model specialized in code. It supports 338 programming languages and features advanced fill-in-the-middle capabilities, offering performance comparable to top-tier proprietary models like GPT-4 Turbo at a significantly lower inference cost.
Qwen 2.5 Coder 32B
by Alibaba Cloud
Qwen 2.5 Coder 32B is an open-weight, code-specialized large language model from Alibaba Cloud. Fine-tuned on a massive corpus covering over 92 programming languages, it excels at code generation, completion, and debugging tasks, demonstrating performance on par with or exceeding proprietary models like GPT-4o on several benchmarks.
Code Llama 34B
by Meta
Code Llama 34B is a large language model from Meta, fine-tuned from Llama 2 for code-specific tasks. It excels at generating, completing, and explaining code across various languages. With variants supporting a 100K token context window, it can analyze and work with extensive codebases for complex tasks like refactoring.
HumanEval+
by BigCode
HumanEval+ is a benchmark for rigorously evaluating code generation models. It augments the original HumanEval dataset by expanding the test suite for each of its 164 problems by 80x. This extensive testing helps uncover subtle bugs and failures on edge cases that simpler benchmarks miss, providing a more accurate measure of a model's true coding ability.
Fraud Detection Pipeline
by Community
This is a complete machine learning pipeline for detecting fraudulent transactions in real-time. It employs a hybrid approach, using XGBoost or LightGBM for classification and an Isolation Forest for anomaly detection. The system is specifically designed to handle severely imbalanced datasets through SMOTE-Tomek resampling and cost-sensitive learning.
Aider
by Paul Gauthier
AI pair programming tool in the terminal that works with any LLM to edit code in local git repositories. Features automatic git commits, multi-file editing, and voice coding with support for connecting to dozens of model providers.
Amazon Q Developer Agent
by Amazon Web Services
Amazon Q is an AI-powered developer agent from AWS that automates code transformations, feature implementation, and security remediation. It is deeply integrated with the AWS ecosystem, allowing it to understand project context, suggest relevant AWS services, and streamline cloud-native development workflows directly within the IDE.
OpenHands
by All Hands AI
OpenHands is an open-source platform for creating autonomous AI software agents. It offers a secure, sandboxed environment where agents can execute complex development tasks by writing code, running commands, browsing the web, and interacting with APIs. It supports multi-agent delegation for tackling intricate problems.
Codex CLI
by OpenAI
OpenAI's open-source CLI coding agent that operates in the terminal with sandboxed execution. Reads and edits files, runs commands, and supports multiple approval modes from suggest to full-auto.
StarCoder2 15B
by BigCode (ServiceNow + Hugging Face)
StarCoder2 15B is a powerful open-source code generation model from the BigCode project. Trained on The Stack v2 dataset spanning over 600 programming languages, it excels at code completion, generation, and fill-in-the-middle tasks, emphasizing data transparency and author opt-out.
DeepSeek Coder 33B
by DeepSeek
DeepSeek Coder 33B is a dense, open-source large language model specializing in code-related tasks. Trained from scratch on a massive 2 trillion token dataset of code and natural language, it understands project-level context and supports 87 different programming languages for advanced code generation and completion.
Code Explanation
by AaaS
Provides detailed, multi-level explanations for code snippets, functions, or entire repositories. It breaks down complex algorithms, clarifies control flow, and describes the purpose of variables and dependencies. The skill supports numerous programming languages, generating documentation-style overviews or granular, line-by-line analyses to accelerate learning and code reviews.
Cline
by Cline
Autonomous coding agent that operates directly in VS Code with support for multiple LLM providers. Can create and edit files, run terminal commands, and browse the web while requiring human approval for actions.
Codestral
by Mistral AI
Codestral is Mistral AI's open-weight generative model explicitly designed for code generation tasks. Trained on a diverse dataset of over 80 programming languages, it excels at code completion, generation, and its unique fill-in-the-middle capability. It is optimized for low-latency performance in real-world applications.
Chatbot Builder Script
by Community
This script generates a production-ready chatbot foundation using Rasa for structured dialogue and an LLM for open-ended fallback. It provides a unified channel adapter for deploying to Web, WhatsApp, and Slack, and includes built-in conversation analytics and a Streamlit-based testing environment for rapid development.
Neo4j RAG Pipeline
by Neo4j
Implements a GraphRAG pattern that stores document entities and relationships in Neo4j, then retrieves contextually relevant subgraphs at query time before passing them to an LLM. Includes automatic entity extraction with spaCy, relationship inference, and a Cypher query generator.
Tabnine + VS Code
by Tabnine
Tabnine's VS Code extension provides AI-powered code completions, including whole-line and full-function suggestions. It is designed for enterprises with strict privacy and data-residency needs, offering on-premise or private cloud deployment options. The AI can be trained on a team's specific codebase for highly relevant completions.
Cline + VS Code
by Community
Cline is an open-source VS Code extension that provides an AI agent with direct access to the IDE's environment. It enables multi-step agentic workflows by allowing the AI to use the file system, terminal, and an integrated browser. The extension supports various models and includes a human-in-the-loop approval process for safety.
Recommendation Engine Setup
by Community
This script provides a complete setup for a modern, two-stage recommendation engine. It uses a two-tower neural network for efficient candidate retrieval and a powerful Large Language Model (LLM) for nuanced re-ranking. The system integrates with a Feast feature store to leverage real-time user context, ensuring timely and relevant suggestions.
Windsurf + Anthropic
by Codeium
Windsurf (by Codeium) is an AI-native IDE that integrates Anthropic's Claude models as the backbone of its Cascade agent, which autonomously plans and executes multi-step coding tasks with real-time file and terminal access. The Anthropic integration powers deep context awareness across large codebases and supports long-horizon agent tasks with coherent state tracking.
Replit AI
by Replit
AI-powered cloud development platform with integrated coding assistant and one-click deployment. Combines a browser-based IDE with AI code generation, debugging, and instant deployment to production.
Aider Polyglot
by Aider
Multi-language code editing benchmark testing models' ability to make targeted code changes across Python, JavaScript, TypeScript, Java, C++, and other languages. Evaluates real-world code modification tasks rather than generation from scratch.
Codeforces Benchmark
by Codeforces / Community
Evaluates models on competitive programming problems from the Codeforces platform across difficulty ratings. Tests algorithmic thinking, data structure knowledge, and the ability to produce correct and efficient solutions under competitive constraints.
MLE-bench
by OpenAI
Benchmark evaluating AI agents on real Kaggle machine learning competitions. Tests the full ML engineering pipeline including data exploration, feature engineering, model selection, training, and submission formatting against actual competition leaderboards.
Continue
by Continue
Open-source AI code assistant for VS Code and JetBrains with customizable model and context providers. Supports tab autocomplete, chat, inline editing, and custom slash commands with any LLM.
Sourcegraph Cody
by Sourcegraph
AI coding assistant powered by Sourcegraph's code graph for deep codebase understanding. Provides context-aware code generation and answers using entire repository knowledge across large codebases.
Gemini Code Assist
by Google
Google's AI-powered code assistance tool integrated with Google Cloud and IDEs. Provides code completions, explanations, and transformations powered by Gemini models with enterprise security controls.