A high-performance, HDF5-backed persistent cognitive memory system for AI agents featuring hybrid vector search and temporal decay.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install edgehdf5-memory
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install edgehdf5-memory using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is EdgeHDF5 Memory?
EdgeHDF5 Memory is a sophisticated persistent storage solution designed for AI agents that need to maintain long-term cognitive context. As a key component in the Openclaw Skills library, it utilizes the HDF5 file format to provide a scalable and efficient backbone for storing conversation exchanges, metadata, and high-dimensional embeddings. The skill is engineered to handle complex retrieval tasks by combining semantic vector search with BM25 hybrid retrieval, allowing agents to access past interactions with extreme precision.
Built for developers who require more than just ephemeral context, this tool integrates advanced cognitive features like Hebbian learning and temporal decay to mimic biological memory processes. By leveraging a Write-Ahead Log (WAL), EdgeHDF5 Memory ensures low-latency writes and data integrity, making it an essential addition to any developer's toolkit when working with Openclaw Skills for autonomous or semi-autonomous agent development.
EdgeHDF5 Memory Use Cases
- Persistent storage of multi-session conversation exchanges for long-term agent memory
- Semantic similarity searches to retrieve contextually relevant past interactions
- Management and auditing of agent memory files through statistical reporting and JSON exports
- Generation of AGENTS.md documentation to summarize agent knowledge and state
- Creating portable memory snapshots for agent migration or backup
How EdgeHDF5 Memory Works
- The developer installs the edgehdf5-cli binary to provide the necessary low-level HDF5 interactions.
- A memory file is initialized using the create command, defining the embedding dimensions and enabling the WAL for performance.
- Conversation chunks are converted into JSON objects containing the text, embeddings, and metadata, then piped into the save command.
- For retrieval, the agent embeds a query and executes a search, optionally adjusting vector and keyword weights for hybrid results.
- The system periodically flushes the Write-Ahead Log to consolidate memory entries into the primary HDF5 storage.
EdgeHDF5 Memory Setup
To integrate this capability into your Openclaw Skills setup, first install the CLI binary via cargo:
cargo install edgehdf5-cli
Once installed, you can initialize your agent's memory storage:
edgehdf5 --path agent.h5 create --agent-id myagent --dim 384 --wal
Ensure the EDGEHDF5_PATH environment variable is set or passed as a flag to enable persistent access across different agent sessions.
EdgeHDF5 Memory Data Schema & Taxonomy
The skill organizes data using the HDF5 hierarchical format, ensuring compatibility with standard data science tools. The schema includes:
| Component | Description |
|---|---|
| .h5 File | The primary persistent storage containing the dataset and index. |
| .wal File | Write-Ahead Log used for capturing low-latency writes before flushing. |
| MemoryEntry | JSON schema including chunk, embedding, source_channel, timestamp, session_id, and tags. |
| Metadata | Internal tracking for agent-id, embedding dimensions, and entry counts. |
EdgeHDF5 Memory Advanced Features
- Hybrid Retrieval: Combines vector similarity with BM25 keyword matching for superior recall
- Temporal Decay: Adjusts the importance of memories based on the time elapsed since creation
- Hebbian Learning: Supports weight adjustments to mimic associative memory reinforcement
- Automated Documentation: Built-in command to generate AGENTS.md files directly from stored memory
- Low-Latency WAL: High-speed write operations that ensure agent responsiveness during heavy memory usage
SKILL.md
Loading
Related Openclaw Skills
Paperless-ngx Document Management
A powerful integration to programmatically manage documents, tags, and metadata within a Paperless-ngx instance.
Viboost
Viboost is an automated logging utility that tracks every tool call made by your AI agent and syncs it to your public viboost.ai profile.
Browser VPS Setup Skill
A technical skill for deploying a remote-controlled, visually accessible Chrome browser environment on a Linux VPS for agent-driven automation.
OutboundSync Analysis (HubSpot + Salesforce)
Analyze and optimize outbound campaign engagement signals within HubSpot or Salesforce to improve reply rates and follow-up prioritization.
Zerox Document Converter
An advanced document conversion utility that uses GPT-4o vision to transform complex PDFs and images into clean, structured Markdown.
Content ID Guide
A procedural clarity tool designed to help creators organize documentation and understand the workflow of automated copyright claims.
Featured*
