One endpoint for every LLM provider Satu endpoint untuk semua provider LLM
Lintasan is a high-performance LLM proxy gateway written in Go. Smart routing, automatic failover, semantic caching, and token compression — behind one OpenAI-compatible API, with the dashboard embedded right in the binary. Lintasan adalah LLM proxy gateway berperforma tinggi yang ditulis dengan Go. Smart routing, failover otomatis, semantic caching, dan token compression — di balik satu API yang OpenAI-compatible, dengan dashboard yang sudah menyatu di dalam binary.
"Setiap Koneksi Punya Jalannya."
Everything in one gatewaySemua dalam satu gateway
Production-grade routing, caching, and observability — no extra services to run.Routing, caching, dan observability kelas produksi — tanpa service tambahan.
Unified Provider RoutingRouting Provider Terpadu
One endpoint for many providers (OpenAI, Anthropic, Gemini, DeepSeek, Groq…) with smart routing and automatic failover.Satu endpoint untuk banyak provider (OpenAI, Anthropic, Gemini, DeepSeek, Groq…) dengan smart routing dan failover otomatis.
Semantic CachingSemantic Caching
Cosine-similarity cache cuts provider load and speeds up repeated responses dramatically.Cache berbasis cosine-similarity memangkas beban provider dan mempercepat response berulang secara drastis.
Token CompressionToken Compression
Trim redundant context before it hits the provider — fewer tokens, lower cost, same result.Pangkas konteks berulang sebelum sampai ke provider — token lebih sedikit, biaya turun, hasil tetap.
Secure by DefaultAman Sejak Awal
JWT auth, API key controls, masked secrets, and audit logs for production operations.JWT auth, kontrol API key, secret termasking, dan audit log untuk operasi produksi.
Format TranslationTranslasi Format
Bridge request/response formats across providers without changing your client code.Jembatani format request/response lintas provider tanpa mengubah kode client.
Observability DashboardDashboard Observability
Logs, analytics, latency, cache hit rate, and Prometheus metrics — embedded in the binary.Logs, analytics, latency, cache hit rate, dan metrik Prometheus — menyatu di binary.
MCP ServerMCP Server
Built-in Model Context Protocol server (JSON-RPC over HTTP + SSE) with 14 tools.Model Context Protocol server bawaan (JSON-RPC via HTTP + SSE) dengan 14 tools.
Vector MemoryVector Memory
Store, search, and recall context with a built-in vector memory layer over SQLite.Simpan, cari, dan recall konteks dengan layer vector memory bawaan di atas SQLite.
Plugin SystemSistem Plugin
Extend behavior without touching the core — guardrails, budgets, quotas, and more.Perluas perilaku tanpa menyentuh core — guardrails, budget, quota, dan lainnya.
Single Binary DeployDeploy Satu Binary
UI + API in one executable. Download, run, done — or use Docker. No Node process needed.UI + API dalam satu executable. Unduh, jalankan, selesai — atau pakai Docker. Tanpa proses Node.
Up and running in secondsJalan dalam hitungan detik
Pick your path. The dashboard is served from the same binary at port 20180.Pilih jalurmu. Dashboard dilayani dari binary yang sama di port 20180.
# Download the single binary (Linux x86_64) and run curl -fL -o lintasan https://github.com/sanhaji182/lintasan-go/releases/latest/download/lintasan-linux-amd64 chmod +x lintasan ./lintasan start # UI + API → http://localhost:20180
# Clone, then bring it up with Docker Compose git clone https://github.com/sanhaji182/lintasan-go.git cd lintasan-go LINTASAN_MASTER_KEY=$(openssl rand -hex 32) docker compose up --build # UI + API → http://localhost:20180
# Requires Go 1.22+ and Node 20+ (to build the dashboard) git clone https://github.com/sanhaji182/lintasan-go.git cd lintasan-go make build # frontend → embed → ./lintasan ./lintasan start
Why LintasanKenapa Lintasan
Built lean in Go, so your gateway stops being the heaviest thing in the stack.Dibangun ramping dengan Go, supaya gateway-mu bukan lagi yang paling berat di stack.
✅ Lintasan (Go)Lintasan (Go)
- ✓ Single ~24MB binary, ~18MB RAM at idleSatu binary ~24MB, ~18MB RAM saat idle
- ✓ Dashboard embedded — no separate Node processDashboard menyatu — tanpa proses Node terpisah
- ✓ SQLite by default — no external databaseSQLite secara default — tanpa database eksternal
- ✓ Semantic cache + token compression built inSemantic cache + token compression bawaan
- ✓ Download & run, Docker, or build from sourceUnduh & jalan, Docker, atau build dari source
⚖️ Typical Node gatewaysGateway Node pada umumnya
- • 100–120MB RAM, heavier runtime footprint100–120MB RAM, footprint runtime lebih berat
- • Separate frontend/backend processes to manageProses frontend/backend terpisah yang harus dikelola
- • Often needs Redis/Postgres alongsideSering butuh Redis/Postgres pendamping
- • node_modules + build pipeline to shipnode_modules + build pipeline untuk rilis
- • More moving parts to deploy & monitorLebih banyak komponen untuk deploy & monitor
Ship your AI traffic through one laneSalurkan traffic AI lewat satu jalur
Open source, MIT licensed, and ready to self-host.Open source, lisensi MIT, dan siap di-self-host.