[Marketen]

Part of guest chat: a landing-page visitor chats without signing up. Companion PRs: dappnode/dappnode-nexus (control plane, metering, UI) and dappnode/infrastructure (settings). Supersedes #18. Stacked on #19, which supplies the prompt estimate every hold uses.

What this changes

A guest is an ordinary account (is_guest from Metering) with a small credit and no login. Its requests are ordinary requests with one difference: a chat request always carries an output limit of at most GUEST_MAX_OUTPUT_TOKENS (default 8,192), written into the request so the provider enforces it.

Why: Metering holds credit for the request's output limit. Without one, the request goes upstream unbounded and a longer answer is charged as debt (dappnode-nexus#151). A guest can't pay a debt, so the limit keeps a guest's answer within its hold.

Why 8,192 and not lower:

• Over a week of production traffic on DeepSeek V4.1 Flash, answers averaged about 1,100 output tokens; 28% used 1,024 or more, and the longest was about 15,000.
• Reasoning counts toward the limit. Against the real DeepSeek API, a hard question at 1,024 tokens spent all of them reasoning and returned an empty answer.

Everything else is untouched: any model the credit covers, tools, images, the key's PII mode. #18 restricted models, fields and request size; none of that is needed once the hold is honest.

Tests

• guest_output_test.go: the limit is written (absent, larger, smaller kept, max_completion_tokens name kept), Metering holds for it, and other fields pass through; signed-in requests get no limit.
• The is_guest flag survives the Metering response, and an older Metering without it means "not a guest".
• go test ./... and go vet ./... pass.

🤖 Generated with Claude Code

[Marketen]

Merged into #22.