MarketDeck docs
analysis
Read stored AI analyses — every agent run leaves a row, this is how you fetch them back.
Every agent run writes one row to the analyses table containing the coin, the timestamp, the provider + model, the AI's full output, and a frozen snapshot of the market data. marketdeck analysis is the read surface.
Usage#
marketdeck analysis <subcommand> [args]Subcommands#
| Subcommand | What it does |
|---|---|
list <coin> | Stored analyses for one coin, newest-first. |
show <ref> | Full payload. <ref> accepts db:<id>, snap:<id>, a legacy file path, or a short numeric id. |
Examples#
marketdeck analysis list BTC --limit 10 --format text
marketdeck analysis show 42 --format json | jq .output
marketdeck analysis show snap:abcd1234 --format textWhere analyses live#
The analyses table holds metadata; the snapshot blob is gzipped to a sibling file when it would otherwise bloat the DB:
~/.local/share/marketdeck/snapshots/<uuid>.json.gzThis started in schema v3 — pre-v3 rows still have their snapshots inline as JSON in the row. The reader handles both transparently.
Pruning#
Old analyses can be removed by retention age:
marketdeck prune analyses --older-than 90paper_analysis_snapshots (the per-position frozen analyses) is not pruned by analyses-targeted prunes — those rows are evidence for paper positions and outlive the rolling history. They are removed via FK cascade when their parent position is deleted.