lynkr

name	lynkr
display_name	Lynkr AI Routing Proxy
version	8.0.1
description	Universal LLM gateway with intelligent routing, Graphify code intelligence, Distill compression, routing telemetry, Code Mode, and 12+ provider support. 60-80% cost reduction for Claude Code, Cursor, and Codex.
author	lynkr-ai
license	MIT
tags	["routing","proxy","llm","multi-provider","ollama","openai","anthropic","bedrock","cost-optimization","graphify","distill","telemetry","code-mode","compression","memory"]
category	infrastructure
homepage	https://github.com/Fast-Editor/Lynkr
npm	lynkr
requires	{"node":">=20"}
providers	["ollama","databricks","azure-anthropic","azure-openai","openrouter","openai","bedrock","vertex","moonshot","zai","llamacpp","lmstudio","codex","deepseek"]

Lynkr - Universal LLM Gateway

Lynkr routes AI coding requests to the optimal model based on task complexity, cost, and provider health. Supports 12+ providers with 60-80% cost reduction through intelligent token optimization.

Quick Start

npm install -g lynkr
lynkr-setup   # Auto-installs Ollama + pulls a model
lynkr          # Start the proxy

Then point your AI coding tool at http://localhost:8081/v1.

How It Works

1. 5-Phase Complexity Analysis - Scores each request 0-100 using token count, tool usage, code patterns, domain keywords, and Graphify structural analysis (god nodes, community cohesion, blast radius)
2. 4-Tier Routing - Maps score to SIMPLE/MEDIUM/COMPLEX/REASONING, each with a configured provider:model
3. Agentic Detection - Detects multi-step workflows (tool loops, autonomous agents) and upgrades to higher tiers
4. Cost Optimization - Picks the cheapest provider that can handle the tier
5. Circuit Breaker + Failover - Automatic failover with half-open probe recovery

Key Features (v8.0)

Intelligent Routing

• 5-phase complexity scoring with 15-dimension weighted mode
• Agentic workflow detection (SINGLE_SHOT / TOOL_CHAIN / ITERATIVE / AUTONOMOUS)
• Graphify knowledge graph integration — god node detection, community cohesion, blast radius
• Routing telemetry with SQLite store, quality scoring (0-100), latency tracking (P50/P95/P99)

Token Optimization (60-80% savings)

• Smart tool selection — filters tools by request type
• Distill compression — structural similarity (Jaccard), delta rendering, block dedup
• Code Mode — replaces 100+ MCP tools with 4 meta-tools (~96% token reduction)
• History compression — sliding window with Distill-powered dedup
• Prompt caching — SHA-256 keyed LRU cache
• Headroom sidecar — optional 47-92% compression via Smart Crusher, CCR, LLMLingua

Production Hardening

• Circuit breakers with half-open probe recovery
• Admin hot-reload endpoint (POST /v1/admin/reload) — no restart needed
• Per-request performance timing (PERF_TIMER=true)
• Prometheus metrics, structured logging, health checks
• Rate limiting, load shedding, input validation