MarketDeck docs

http

Inspect the cross-process HTTP request log for outgoing exchange / provider traffic.

Every outgoing HTTP request the CLI, daemon, and plugins make goes through core/http.js. The client adds a reliability layer — token bucket, retry with decorrelated jitter, circuit breaker — and writes each completed request to a JSONL log under ~/.local/share/marketdeck/ so every process shares one timeline.

Usage#

marketdeck http <subcommand> [flags]

Subcommands#

Subcommand	What it does
`log [--tail N] [--host H] [--outcome O]`	Recent request lines.
`stats`	Per-host roll-up: total / success / failure counts, avg latency, success rate.
`status`	Per-host health — `OK`, `degraded`, `down`.

Flags#

Flag	What it does
`--tail N`	Last N entries (default 200).
`--host H`	Filter to a specific host (e.g. `api.hyperliquid.xyz`).
`--outcome O`	`success`, `client-error`, `server-error`, `network`, `aborted`.

Examples#

marketdeck http log --tail 50 --format text
marketdeck http log --host api.hyperliquid.xyz --outcome server-error
marketdeck http stats --format text
marketdeck http status --format text

Where the log lives#

~/.local/share/marketdeck/http-log.jsonl

One line per request. Fields (per line):

id, pid (the MARKETDECK_PROC_TAG of the originating process).
startedAt, endedAt, durationMs.
host, path, url.
method, bodySize — the body itself is not logged, just the byte count.
tag — derived from a known-shape POST body (e.g. candleSnapshot/BTC/1h).
status, ok, retries.
outcome, errorCode, errorMessage (telegram bot tokens redacted).

Reliability counters are kept per-host in memory by every process; they reset on process restart.

Retention#

Pruning by retention age:

marketdeck prune http-log --older-than 14

Or wipe entirely:

> ~/.local/share/marketdeck/http-log.jsonl

The status / stats counters live in memory and are unaffected.