8.4K ·GitHub
What is this?
It is a 'capsule' (plugin) for Astrid OS / Unicity AOS agents that gives them cross-session memory: it reads a local memory file from the workspace and injects its content into the system prompt so the agent remembers what happened in previous sessions.
It is a 'capsule' (plugin) for Astrid OS / Unicity AOS agents that gives them cross-session memory: it reads a local memory file from the workspace and injects its content into the system prompt so the agent remembers what happened in previous sessions.Is this a standalone app or a plugin for something bigger?It's a small Rust hook module ('capsule') that plugs into Astrid OS's prompt-building pipeline — it only works as part of the larger Unicity AOS agent framework, not on its own.
It's a small Rust hook module ('capsule') that plugs into Astrid OS's prompt-building pipeline — it only works as part of the larger Unicity AOS agent framework, not on its own.How do I get this running?There's no separate install step: on each prompt assembly cycle it automatically reads '.astrid/memory.md' from the workspace via the VFS, wraps it in a '# Memory' section, and publishes an appendSystemContext hook response.
There's no separate install step: on each prompt assembly cycle it automatically reads '.astrid/memory.md' from the workspace via the VFS, wraps it in a '# Memory' section, and publishes an appendSystemContext hook response.Does it need other Astrid OS capsules to work?Yes — this capsule only reads memory; writing to '.astrid/memory.md' is done through the filesystem tools (writefile, replaceinfile) provided by the separate astrid-capsule-fs capsule.
Yes — this capsule only reads memory; writing to '.astrid/memory.md' is done through the filesystem tools (writefile, replaceinfile) provided by the separate astrid-capsule-fs capsule.Are there any limits on how much memory it can inject?Yes, content is hard-capped at 32KB to protect the context window; anything beyond that gets cut at a UTF-8 boundary and marked with a truncation notice.
Yes, content is hard-capped at 32KB to protect the context window; anything beyond that gets cut at a UTF-8 boundary and marked with a truncation notice.Who is this capsule meant for?Developers building or extending agents on Astrid OS / Unicity AOS who want their agents to retain context between sessions.
Developers building or extending agents on Astrid OS / Unicity AOS who want their agents to retain context between sessions.What is Astrid OS / Unicity AOS?It's an agent operating-system framework where this project is one of several modular 'capsules' that provide OS-like primitives (filesystem, memory, etc.) to AI agents, with this one acting as the persistent swap file for context.
It's an agent operating-system framework where this project is one of several modular 'capsules' that provide OS-like primitives (filesystem, memory, etc.) to AI agents, with this one acting as the persistent swap file for context.What happens if memory.md doesn't exist yet?The capsule simply becomes a no-op and injects nothing into the prompt until the file exists and has content.
The capsule simply becomes a no-op and injects nothing into the prompt until the file exists and has content.Why not just use a human-written AGENTS.md style memory file?Because agent-written content can grow without limit, unlike human-authored AGENTS.md, so this capsule enforces a hard 32KB cap to keep the context window bounded.
Because agent-written content can grow without limit, unlike human-authored AGENTS.md, so this capsule enforces a hard 32KB cap to keep the context window bounded.What are similar projects?EverOS, memcode, Athena-Public, mem0, GLM-5.3-Flash-J-Space-Capability-Realization-Report, agentmemoryrepo, harness-sdk, hindsight
EverOS, memcode, Athena-Public, mem0, GLM-5.3-Flash-J-Space-Capability-Realization-Report, agentmemoryrepo, harness-sdk, hindsightWhich topics is it in?system-prompt, persistent-memory, session, context-management, memory-system, long-term-memory, hooks, agent-memory, memory, prompt-engineering, artificial-intelligence, agentic-ai
system-prompt, persistent-memory, session, context-management, memory-system, long-term-memory, hooks, agent-memory, memory, prompt-engineering, artificial-intelligence, agentic-ai