Reading
Engram is an OpenCode plugin that enables agents to access complete conversation history on demand through two core modules—Context Graph and Upstream History Retrieval. It addresses the information loss issue in traditional context compression and opens up a new paradigm for multi-agent collaboration.
Section 01
Engram is an OpenCode plugin. Through two core modules—Context Graph and Upstream History Retrieval—it proposes a 'pull-based' history access paradigm, allowing agents to obtain complete conversation history on demand. This solves the information loss problem in traditional 'push-based' context compression and opens up a new paradigm for multi-agent collaboration.
Section 02
In AI applications, conversation history (including reasoning chains, rejected paths, and user constraints) is valuable task-specific experience, yet it is often 'sunken' into storage and cannot be reused. The current mainstream push-based context transfer faces three key dilemmas: cognitive asymmetry (the model doesn't know what it's missing), pre-decision risk (unknown subsequent tasks during compression), and cumulative distortion (early information disappears due to repeated summarization).
Section 03
Engram solves the problem through two modules: 1. Context Graph: Extracts lossless indexes (not lossy summaries) from original conversations, allowing agents to pull original history on demand; 2. Upstream History Retrieval: Enables sub-agents to independently retrieve the complete conversation history of upstream agents, breaking the limitations of information filtering.
Section 04
The pull-based paradigm is the opposite of the push-based one—agents obtain the required context on their own when the need arises. Its advantages include: delaying filtering until the need emerges, and having the agent itself (which best understands its needs) as the filtering subject. Engram is the first to apply this paradigm to context transfer scenarios.
Section 05
Engram is a plugin for the OpenCode platform (a non-official personal project) that leverages the platform's openness in conversation history access and plugin integration. Its architectural design considers deployment needs: structured storage of navigation data blocks supports efficient retrieval, history access interfaces are called on demand, and compliance with OpenCode specifications ensures compatibility.
Section 06
Engram's value is reflected in: 1. Long conversation maintenance: Avoids 'amnesia' caused by traditional compression and retrieves key decision points; 2. Multi-agent collaboration: Downstream agents independently explore upstream reasoning, reducing misunderstandings and omissions; 3. Debugging and auditing: Traces the agent's thinking path to locate the root cause of problems.
Section 07
Engram is not just a practical plugin; it redefines the context management paradigm: turning conversation history from a 'sunk cost' into a 'retrievable resource', laying the foundation for building more reliable, transparent, and collaborative agent systems, and pointing the way for future research.
Keep going with more reads from the same topic.
Splinter is a minimalist, high-performance key-value (KV) and vector storage system enabling zero-latency inter-process communication via shared memory and atomic operations. With only 766 lines of core code, it supports millions of operations per second and 768-dimensional vector storage, offering a new architectural approach for local LLM inference and data-intensive applications.
Folkering OS is the world's first AI-native bare-metal operating system, entirely written in Rust no_std without relying on Linux, POSIX, or libc. It can generate commands from scratch, compile them into WASM, and run them in 10 seconds, achieving true self-evolution.
This project explores how to use Large Language Models (LLMs) to detect logical vulnerabilities in Ethereum smart contracts, especially business logic flaws that are difficult for traditional static analysis tools to capture, such as economic manipulation and execution order errors.
A self-hosted MCP server that provides cross-session persistent coordination, task management, event bus, and observability for AI agents. It supports 35 tools and can be integrated into clients like Claude Code and Cursor with zero code changes.