Gateway configuration

Every flag warpllm-server accepts, and the environment it reads.

The HTTP gateway has landed on main but is not part of a published release. To run it today you build it from source or from the Dockerfile in the repository root. Flags, endpoints, and the error envelope may still change before it ships.

Configuration is built-in defaults overridden by command-line flags. There is no config file today.

warpllm-server --host 127.0.0.1 --port 9090 --timeout-secs 120 \
  --stream-read-timeout-secs 90

--flag value and --flag=value both work.

Flags

--host · string · default: "0.0.0.0"

Bind address.

The default listens on every interface, and the gateway has no caller authentication. Use 127.0.0.1 unless something in front of it is handling auth.

--port · u16 · default: "8080"

Listen port.

--timeout-secs · u64 · default: "600"

Total deadline on an upstream request, in seconds. Matches the OpenAI SDK’s default.

--stream-read-timeout-secs · u64

Seconds a stream may go without a byte before it is abandoned. Unbounded by default.

Must exceed the slowest time-to-first-token you expect — the wait before the first chunk is a gap like any other, and a reasoning model can think for minutes before emitting a token. See Timeouts.

--help

Print usage and exit 0.

--version

Print the version and exit 0.

Anything else is a parse error: an unknown flag, a missing value, or a value the type will not take (--port eighty) exits non-zero with a message naming the flag.

Why unbounded by default

No stream read timeout is right for every deployment, and one set too tight cuts off a model that paused to think — a failure that looks like a provider problem and is not. The flag is opt-in for that reason. Set it once you have measured your slowest time-to-first-token.

Environment

The gateway reads the same variables the SDKs do, one per provider, at startup. Set the ones you route to; the rest can stay unset.

Provider prefix	Environment variable	Base URL
anthropic/	ANTHROPIC_API_KEY	https://api.anthropic.com/v1
deepseek/	DEEPSEEK_API_KEY	https://api.deepseek.com
kimi/	MOONSHOT_API_KEY	https://api.moonshot.ai/v1
mistral/	MISTRAL_API_KEY	https://api.mistral.ai/v1
openai/	OPENAI_API_KEY	https://api.openai.com/v1
opencode/	OPENCODE_API_KEY	https://opencode.ai/zen/v1
openrouter/	OPENROUTER_API_KEY	https://openrouter.ai/api/v1

This table is generated from the model registry, so it is exactly what the gateway looks for.

The gateway builds its client with no providers declaration, so every provider talks to its own API and the whole roster is routable. Narrowing is deliberately not a flag: a key on the command line would sit in ps for a benefit the environment already provides.

What is not configurable yet

	Status
Caller authentication	None — put your own in front
Per-caller API keys	None — the gateway uses its own provider keys
Narrowing which providers are served	Client-side only; the gateway serves the whole roster
Config file	None — flags only
CORS	Not configured
Rate limiting	None
Load balancing	On main as a Rust client, not a gateway flag
Failover, retries, caching	Planned — #27, #28

Next

Connect a client

Point the official OpenAI SDKs at a running gateway.