Browse / General-purpose / agent-memory
agent-memory
Raven-Futureswamp
Research and implementation of resilient memory architectures for AI agents
Open Raven-Futureswamp/agent-memory → Does not match the four rules.
Works with any agent
ships AGENTS.md, the cross-tool instruction file.
In its own words
This is Raven's workspace for OpenClaw/ClawdBot — a practical implementation of persistent AI agent architecture.
What's Here
- **
docs/anti-lobotomy-architecture.md** — The Anti-Lobotomy Architecture (v2.0), a comprehensive guide to AI agent context persistence, identity recovery, and vector memory integration - **
projects/** — Active development projects with STATUS.md files - **
memory/** — Daily logs, coordinates.json, session ledger, and recovery state - **
scripts/** — Automation for cleanup, git workflows, and trading - **
heartbeat/** — Modular idle-aware automation system - **
trading/** — Crypto trading bot (Robinhood) and stock trading (Alpaca)
The Architecture
This workspace implements a nine-layer anti-lobotomy architecture designed to help AI agents survive context resets, session compaction, and memory loss:
- Static Identity — Byte-identical boot for KV-cache reuse
- Live State — coordinates.json + current-task.md
- Session Ledger — Hierarchical browseable archive
- File System Memory — Structured markdown (projects, daily logs)
- Vector Memory — LanceDB + Ollama embeddings
- Auto-Recall — Semantic memory injection every conversation
- Recovery Protocol — Three-door graduated triage (Door 1/2/3)
- Heartbeat Rotation — Modular idle-aware automation
- Automation — Session cleanup, pre-compaction saves
Read the full paper: docs/anti-lobotomy-architecture.md
Quoting 59% of the README. Read the rest on GitHub ↗
Where it came from
What this copy built: +canvas +docs +logs +memory +node_modules +projects +scripts +trading — top-level directories this copy added on top of the original.
Read from the repository's git tree on 2026-08-17.