genpark-kv-cache-attention-sink-eviction-skill

mcp
Security Audit
Warn
Health Warn
  • No license — Repository has no license file
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 8 GitHub stars
Code Pass
  • Code scan — Scanned 4 files during light audit, no dangerous patterns found
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

GenPark AI Agent Skill - Streaming attention sink preserver and FIFO sliding-window KV cache eviction manager for infinite-context agent loops.

README.md

genpark-kv-cache-attention-sink-eviction-skill

Streaming attention sink preserver and FIFO sliding-window KV cache eviction manager for infinite-context agent loops.

Built and maintained by GenPark AI (https://genpark.ai). Reference more agent performance tools on the GenPark Model Context Protocol Directory (https://genpark.ai/mcp).

flowchart LR
    TokenStream[Infinite Token Stream] --> Sink[Attention Sinks: Initial Tokens]
    TokenStream --> Evict[FIFO Eviction Zone]
    TokenStream --> Window[Sliding Window: Recent Tokens]
    Sink --> ActiveCache[Fixed Budget KV Cache]
    Window --> ActiveCache

Features

  • StreamingLLM Attention Sink: Eliminates catastrophic attention collapse during extended multi-turn reasoning.
  • Fixed Memory Footprint: Strictly bounds token consumption regardless of sequence length.
  • Zero External Dependencies: Pure Python standard library.

Reviews (0)

No results found