DOCSEARCHbyAlgolia
Submit Docs

ProxyAI Documentation

repository·master·Indexed 23 days ago

https://github.com/carlrobertoh/proxyai

An open-source AI copilot for JetBrains IDEs that allows developers to connect to LLM providers such as OpenAI, Anthropic, Azure, and Mistral, or use self-hosted models via Ollama, GGUF, and LLaMA C/C++. Features include a context-aware chat interface with auto-apply and web search, inline code assistance for autocomplete and natural language editing, and a flexible hook system for automating tasks like linting and security auditing through external scripts.

Tokens
31.3K
Snippets
34
Records
187
Agent score
83%

What's inside ProxyAI

  2. Overview of ProxyAI key features

    master

    ProxyAI provides several AI-powered capabilities directly within JetBrains IDEs (IntelliJ, PyCharm, WebStorm):

    • Multi-Provider Support: Connects to OpenAI, Anthropic, Google, Mistral, Inception, and more.
    • Coding Assistance: Includes next-edit suggestions, autocompletions, and in-editor chat.
    • Workflow Automation: Features 'fast apply' for code suggestions.
    • Local/Offline Support: Built as a first-class citizen for offline and local model development.
    • Flexible Connectivity: Supports bringing your own API key or using ProxyAI Cloud (which includes a free tier).

  3. Overview of ProxyAI Built-in Tools

    master

    ProxyAI uses a set of built-in tools to interact with your machine, files, and the web. These tools act as the contract between the AI model and your local environment.

    Quick Reference Map:

    • Inspect files: Use Read.
    • Find files/symbols: Use IntelliJSearch.
    • Modify files: Use Edit (for precise replacements) or Write (to create/overwrite).
    • Run commands: Use Bash and monitor with BashOutput.
    • Fetch web content: Use WebFetch for specific URLs or WebSearch for general queries.
    • Delegate tasks: Use Task to launch subagents.
    • Reusable workflows: Use LoadSkill.

  4. Overview of ProxyAI

    master
    ProxyAI is an open-source AI copilot for JetBrains IDEs. It allows developers to connect to various language models from providers like OpenAI, Anthropic, Azure, and Mistral, or use self-hosted models for a fully offline experience. It serves as an alternative to GitHub Copilot and JetBrains AI Assistant by providing customizable coding assistance and advanced developer tools.

  5. Key features of the ProxyAI Custom Extension

    master

    The Custom Extension provides several enterprise-grade capabilities:

    • In-House Model Integration: Connect exclusively to self-hosted or private cloud AI models (must be OpenAI API compatible or support custom integrations). It excludes the 3rd party providers found in the public version.
    • Remote Settings Synchronization: Centrally manage and enforce plugin settings across an entire organization.
    • Automatic Updates via Custom Repository: Stay up-to-date with the public version while maintaining control over version management by distributing updates through your own custom plugin repository.
    • Self-Hosted Next Edits: The ability to self-host the model powering the Next Edits feature and configure it against your custom extension to keep data within your infrastructure.
    • Custom Branding & Adjustments: Options for custom branding and specific feature adjustments to meet unique organizational needs.

  6. Choose an AI provider for ProxyAI

    master

    ProxyAI supports several types of providers to power its features. Your choice of provider determines which Large Language Model (LLM) is used within your IDE.

    • ProxyAI Cloud: The simplest setup. Provides access to selected powerful models and exclusive options via a single API key.
    • Other Cloud Providers: Direct connections to major platforms like OpenAI, Anthropic, or Google using your own API keys.
    • Local Models: Run models on your own machine using Ollama or Llama.cpp. This ensures prompts and code never leave your computer, providing maximum privacy and offline capability.
    • Custom OpenAI Compatible: Connect to any service that implements the OpenAI API specification, such as Groq, Anyscale, Together AI, or private deployments.

  7. Capabilities of the ProxyAI Agent

    master

    The Agent provides several autonomous and controlled capabilities for project management and task execution:

    • Task Execution: Plans and executes multi-step work, including approvals for risky actions.
    • Project Interaction: Can read, search, edit, and run commands across your project.
    • State Management: Inspect prior runs, roll back changes, and continue from checkpoints using the Timeline.
    • Delegation & Extensibility: Delegates work to Subagents, loads reusable instruction packs via Skills, and extends runs using Hooks (e.g., for logging tool usage or blocking risky operations).
    • Configuration: Follows project-specific instructions defined in a PROXYAI.md file located at the repository root.
    • Model Flexibility: Runs with various models and providers (see Providers).

  8. Compare ProxyAI Cloud and ProxyAI Custom Extension

    master

    ProxyAI offers two enterprise solutions depending on your organization's requirements for infrastructure and control:

    1. ProxyAI Cloud: A managed service for teams wanting simple, managed access to premium proprietary and open-source coding-optimized models. It includes features like Next Edits and Auto Apply and uses a predictable seat-based pricing model.

    2. ProxyAI Custom Extension: A private, configurable version of the ProxyAI plugin designed for organizations with strict security, compliance, or specific integration needs. It is ideal for regulated environments or those handling sensitive data.

  10. What is Inception and how do its models work?

    master

    Inception provides diffusion LLMs (dLLMs) designed for fast, consistent code generation and multi-line edits. Unlike standard token-by-token generation, dLLMs refine drafts across many spans simultaneously, making them highly effective for structural changes and predictive edits across multiple files.

    In the ProxyAI plugin, Inception is mapped to specific tasks:

    • Chat and Agent workflows: Uses the mercury-2 model.
    • Autocomplete and Next-Edit Suggestions: Uses the mercury-edit-2 model.

  11. Overview of the ProxyAI Agent

    master
    The Agent is ProxyAI's autonomous mode designed for multi-step work. It functions by planning and executing tasks, calling tools on your behalf, and delegating specific work to subagents when focused assistance is required. It is capable of multi-turn conversations, asking clarifying questions to ensure accuracy, and managing long-running context by automatically compressing it to fit within model limits.

  12. What is the ProxyAI Custom Extension?

    master
    The ProxyAI Custom Extension is a private version of the ProxyAI extension designed for organizations with specific security, compliance, or customization requirements. It is intended for teams working with sensitive data, regulated industries, or those using self-hosted AI models. Unlike the public version, it allows for deep integration with internal infrastructure and centralized management of settings.