Understanding PiecesOS
Read about PiecesOS, the foundational layer that supports the whole Pieces for Developers Suite, and the core functionalities—like LTM-1, Pieces Drive, and Pieces Copilot.
What is PiecesOS?
PiecesOS is a background services that runs on your machine, orchestrating local data processing and managing house-made machine learning (ML) models used within Pieces software.
There are three core pillars of Pieces functionality powered by PiecesOS: [1] LTM-1, [2] Pieces Drive, and [3] Pieces Copilot.
These components make up the Pieces experience and are essential to the modern AI-powered workflow.
The Role of PiecesOS
PiecesOS provides the intelligence and power behind Pieces software in two key ways: by (1) supplying the essential components of the Pieces infrastructure and supporting various processes, and (2) powering standalone Pieces plugins and extensions when used without the Pieces Desktop App.
This ‘brain’ is required to enable the 3 fundamental qualities of the Pieces development experience.
How Does PiecesOS Work?
Powered by PiecesOS, the LTM-1 (Long-Term Memory) Engine tracks your development and workflow context using the Pieces Drive, allowing you to import and export materials as needed. Then, with Pieces Copilot, you can use cloud-based or local LLMs through Ollama to start chats that include the context you need.
This way, all AI-powered elements work together to enhance your context and creativity, especially when you need a memory boost.
Fundamental Components
Using the Pieces Desktop App or a Pieces plugin or extension for your favorite IDE requires PiecesOS, as it is a required dependency for memory and context preservation, storing and accessing materials, and interacting with generative AI.
LTM-1
The LTM-1, or Long-Term Memory Engine, stores and recalls workflow context for up to 90 days and enables powerful contextual queries and productivity features through the Pieces Copilot or Captured Context view (if using the Pieces Desktop App).
Pieces Copilot with LTM-1 Context — Pieces Desktop App
Grounded Assistance
Long-Term Memory designed to boost developer productivity by providing assistance that’s temporally-grounded within the concrete context of your actual workflow, allowing the Pieces Copilot to better understand your development process over time.
This helps it offer more relevant and timely responses to queries, as it has a local database of information to work from.
How Context is Captured
The LTM-1 works by monitoring your workflow at the operating system level, capturing data from various activities across your development environment.
This includes interactions within Integrated Development Environments (IDEs), browsers, and collaboration tools.
The data captured by the LTM-1 is processed and stored entirely on your device, so you don’t need to worry about privacy and security—which you can read more about here.
Less Context Switching
Temporally grounding the Pieces Copilot with LTM-1 data reduces manual context input, enhances task continuity, minimizes context switching, and allows for more intuitive interactions.
You can ask questions like, “What was I working on an hour ago?” or “How can I fix the CocoaPods issue in IntelliJ?” and get informed responses based on recent activities, including relevant file paths or URLs.
On-Device Data Storage
All data captured by the LTME is stored locally on your device. At no point does this data leave your device or become accessible to anyone, including the Pieces team, unless you choose to share it.
The LTM-1 applies on-device machine learning algorithms to filter out sensitive information and secrets, maintaining high levels of performance, security, and privacy.
For advanced components that require blended processing, user preferences can be set to leverage a cloud-powered Large Language Model as the Copilot’s runtime.
Pieces Drive
Pieces Drive is an intelligent and interactive material manager baked into PiecesOS that allows you to save, manage, and share important developer resources such as notes, useful code snippets, links, and more within your personal local repository.
From this repository, you can view, edit, reuse, and share these materials via Pieces Shareable Links.
Viewing Saved Materials with Pieces Drive — Pieces Desktop App
Multi-Tool Integration
With PiecesOS, you can use Pieces Drive in tools like Obsidian, Raycast, Microsoft Teams, and your favorite code editors like IntelliJ, Visual Studio Code, and Sublime Text, and much more.
You can also view and adjust metadata saved with your materials in the Applet View of Pieces plugins or the Pieces Desktop App.
AI-Powered Enrichment
When you save code or other useful materials to Pieces Drive, AI automatically enhances them. This adds details like the programming language, descriptions, tags, and suggested searches for more information.
Pieces will also identify and flag any sensitive information, like API keys or personal data. Since coding is often a team effort, Pieces can also track related people based on Git history or who shared the code with you.
Pieces Copilot
Like Pieces Drive, you can find Pieces Copilot nested within Pieces plugins and extensions or the Pieces Desktop App, where it can be used to generate code, answer questions, add code documentation, and serves as the bridge between you and your workflow context as captured by the LTM-1.
Pieces Copilot for Code Generation — Pieces Desktop App
Introducing Context to Chats
If you encounter a scenario where you need to recall a specific URL to a Firestore database that was used during a live demo with another coworker, you can use Pieces Copilot (with LTM-1 enabled) to recall that exact link.
Understanding Code
Pieces Copilot helps you comprehend complex code snippets by providing explanations, detailing how specific functions work, and suggesting alternative implementations for that code.
Try asking it alternative ways to handle logging, or more robust methods of error handling—give it an entire file, an entire project, or just a function.
Code Generation
By specifying your requirements for a function or script, Pieces Copilot can generate code that fits into your active file seamlessly.
Use Pieces Copilot and Pieces Drive together to generate boilerplate code or adjust a function, then save it to Pieces Drive with enriched metadata for future use, saving time.
Error Handling
Debugging, checking logs, and handling errors are often the toughest parts of coding. But with Pieces Copilot, you can debug code thoroughly using the Pieces Desktop App or your preferred plugin.
Context-based debugging is powerful because generative AI, whether cloud-based or local, can understand code semantics beyond what's visible and within the project's scope to find solutions and fixes.
You might even be able to finally center that div—seriously, we’ve actually done it.