NexusQuant - KV Cache Compression
Drop-in KV cache compression for unlimited context windows. One line: with nexusquant(model): model.generate(...).
Core
| Project |
Description |
Install |
| NexusQuant |
KV cache compression: 7x training-free, beats TurboQuant |
pip install nexusquant-kv |
Results
| Model |
Compression |
PPL Delta |
Context (128Kâ) |
| Llama-3 |
5.3x |
-0.002% |
680K tokens |
| Mistral-7B |
7x |
-2.26% |
896K tokens |
Pipeline: NSQ Normalization â Hadamard Rotation â E8 Lattice VQ â Temporal Predictive Coding
asqav - AI Agent Governance
Open-source tools for AI agent compliance, audit trails, and policy enforcement.
Core
| Project |
Description |
Install |
| asqav SDK |
Python SDK for AI agent governance |
pip install asqav |
| asqav MCP |
MCP server for Claude Desktop, Cursor, Claude Code |
pip install asqav-mcp |
| asqav Compliance |
GitHub Action - scans PRs for governance gaps |
Add to .github/workflows/ |
Examples
Resources
Quick start
from asqav import audit
@audit
def call_agent(prompt):
return llm.invoke(prompt)
# Every call now has a tamper-evident audit trail
result = call_agent("Analyze Q1 revenue")
Compliance frameworks supported
- EU AI Act (Articles 9-15)
- DORA (Digital Operational Resilience Act)
- ISO 42001 (AI Management System)
- OWASP Top 10 for Agentic AI