Skill v1.0.1

currentAutomated scan100/100

yarlabs/hyperspace-db/hyperspacedb-memory

+1 new

──Details

PublishedOctober 25, 2026 at 04:01 PM

Content Hashsha256:ca87a37a9eb57415...

Git SHAb520a65ad3cb

Bump Typepatch

Compare with v1.0.0

──Files

Files (1 file, 4.5 KB)

SKILL.md4.5 KBactive

SKILL.md · 138 lines · 4.5 KB

version: "1.0.1" name: hyperspacedb-memory description: > Zero-overhead cognitive long-term and episodic memory for AI agents (Mem0 drop-in and mcp-hyperspace-memory). Use this skill when implementing agent memory, episodic fact storage, semantic recall with session isolation, memory update/delete, Fréchet mean consolidation, or geodesic hallucination defense. Trigger on: "agent memory", "Mem0", "episodic memory", "remember", "recall", "forget", "mcp-hyperspace-memory", "hyperspace-memory", "long term memory".

Hyperspace AI Agent Memory (Mem0 Alternative)

Hyperspace Memory is a high-performance cognitive memory layer designed for LLMs, autonomous agents, and spatial multi-agent swarms.

Compared to legacy memory tools (Mem0, Zep):

100x lower latency (< 2ms vs 2,500ms)
Zero LLM tokens on write ($0.000000 per insert vs $0.01 per insert)
61.4x RAM compression via extreme 1-bit ADC quantization
Lorentz $\mathbb{H}^{32}$ hierarchy representation + Geodesic Hallucination Guard

1. Tier 1: Zero-Code MCP Server (`mcp-hyperspace-memory`)

Connect to Claude Desktop, Cursor, Windsurf, or Antigravity with a single JSON config:

json

{
  "mcpServers": {
    "hyperspace-memory": {
      "command": "npx",
      "args": ["-y", "mcp-hyperspace-memory"],
      "env": {
        "HYPERSPACE_HOST": "the.yar.ink",
        "HYPERSPACE_API_KEY": "YOUR_YARINK_API_KEY"
      }
    }
  }
}

The 9 Cognitive Tools:

memory_remember: Store episodic facts with auto-vectorization (text, session_id, tags, importance).
memory_recall: Semantic search with session isolation (query, session_id, top_k).
memory_update: Modify an existing memory by ID (memory_id, new_text).
memory_forget: Delete a memory item by ID (memory_id).
memory_list_sessions: Enumerate all active conversation sessions.
memory_explore_hierarchy: Traverse concept taxonomy in Lorentz space.
memory_consolidate: Compute Fréchet mean of episodic memories into an abstract concept on the hyperboloid.
memory_verify_claim: Calculate Geodesic Trust Score against stored premises to block hallucinations.
memory_stats: Operational diagnostics, vector counts, and cascade settings.

2. Tier 2: Drop-in Mem0 Replacement SDKs

Python (`hyperspace-memory`)

bash

pip install hyperspace-memory

python

from hyperspace_memory import Memory
# 1-Line Drop-in Replacement for Mem0
memory = Memory(config={
    "host": "the.yar.ink",
    "api_key": "YOUR_YARINK_API_KEY",
    "quantization": "extreme"  # 1-bit ADC + Lorentz f32
})
# Add memory (<2ms, 0 tokens)
res = memory.add(
    "Patient blood glucose is 14.2 mmol/L with rapid rising trend.",
    user_id="user_alice",
    agent_id="agent_medical",
    session_id="session_shift_01",
    importance=9.0,
    tags=["glucose", "critical"]
)
mem_id = res["results"][0]["id"]
# Semantic search with user & session isolation
memories = memory.search("glucose trend", user_id="user_alice", session_id="session_shift_01", limit=5)
# Cognitive Consolidation (Fréchet Mean on Hyperboloid)
concept = memory.consolidate("glucose trend", summary_text="Patient displays recurring glycemic instability.")
# Anti-Hallucination Geodesic Verification
check = memory.verify_claim(
    "Patient blood glucose is 14.2 mmol/L with rapid rising trend.",
    "Patient blood sugar is within optimal normal range."
)
print("Trust status:", check["status"], "Trust score:", check["trust_score"])
# Update & Delete
memory.update(mem_id, "Patient blood glucose normalized to 6.5 mmol/L.")
memory.delete(mem_id)

TypeScript (`hyperspace-memory`)

bash

npm install hyperspace-memory

typescript

import { Memory } from "hyperspace-memory";
const memory = new Memory({
  host: "the.yar.ink",
  apiKey: "YOUR_YARINK_API_KEY",
  quantization: "extreme"
});
// Add memory
const res = await memory.add(
  "Deployment target: production-us-east-1 on Kubernetes v1.30.",
  { userId: "devops_bob", agentId: "infra_agent", sessionId: "deploy_01", tags: ["k8s", "prod"] }
);
// Semantic search
const results = await memory.search("Kubernetes version?", { userId: "devops_bob" });
console.log(results[0].memory);
// Fréchet Mean Consolidation
const concept = await memory.consolidate("Kubernetes version");
// Geodesic Claim Verification
const verification = await memory.verifyClaim(
  "Deployment target is production-us-east-1 on Kubernetes v1.30.",
  "Deployment target is AWS ECS Fargate."
);
console.log("Verified:", verification.status, "Score:", verification.trustScore);

← v1.0.0 All versions