Skip to content
~/gateway

This is the same click-to-stream chat as KobeLLM, but every message is routed through ts-llm-gateway, a multi-provider LLM gateway in strict TypeScript, to AWS Bedrock (with OpenAI as automatic failover). Behind the stream sit real production policies: per-key/IP rate limiting, retry with backoff, a circuit breaker that quarantines a downed provider, and an LRU response cache. It's the same class of system I built at work, a real-time streaming pipeline, re-expressed as an LLM proxy. Your browser only ever talks to my Go backend, which holds the gateway key; the counters below are the gateway's own live /stats, proxied through that backend.

Checking…Streaming · via ts-llm-gateway

A live chat routed through ts-llm-gateway → AWS Bedrock.

Routing
AWS Bedrock
Failover
OpenAI
Policies
rate-limit · retry · breaker · cache
Transport
SSE · cancellable
ts-llm-gateway · live /stats
gateway offline
requests
-
served
cache hit rate
-
LRU cache
p50 latency
-
ms
p99 latency
-
ms
failovers
-
provider switches
providers
-
bedrock / openai
tokens
-
in / out
rejected
-
rate-limit + auth + caps

Disclaimer: responses are generated by a third-party foundation model via AWS Bedrock and may be inaccurate. Conversations are rate-limited, cost-capped, and not stored. Do not enter sensitive information.

Gateway Chat · Kobe Young