Export formats

What you get when you export from rdn-swarm, and where to find it.

Run zip#

Exported from Output (the zip icon), from History (Share as zip), or from App Preview (the open-externally icon). Every entry sits inside one folder named after the run, so unzipping never scatters files.

<run-id>/
├── index.html, styles.css, game.js …   generated source, same tree as Output
├── docs/
│   ├── SUMMARY.md          README written by the Summary Generator
│   ├── requirements.md     the brief this run built from
│   └── notes.md            your notes, if you added any
├── tests/
│   └── acceptance-tests.js the executable acceptance tests
├── screenshots/            Visual / Interactive QA captures, e.g. iter-2-mobile.png
├── token-usage.csv         cost and token breakdown
├── token-usage.txt         the same breakdown, human-readable
└── llm-calls.csv           one row per model call

Folders only appear when the run produced something for them. llm-calls.csv is included for runs recorded with the per-call ledger.

token-usage.csv and token-usage.txt#

One row per phase of each iteration. The columns are:

runId, iteration, phase, role, provider, model, inputTokens, outputTokens, cacheReadTokens, cacheWriteTokens, cost, apiCalls, toolCalls, elapsedMs, outcome, createdAtMs, reasoningTokens, cacheSavings

The .txt file is a readable summary of the same figures.

llm-calls.csv#

One row per model call, the most detailed record of where a run's money went:

runId, iteration, phase, role, workerId, turn, provider, modelId, maxOutputTokens, inputTokens, cacheReadTokens, cacheWriteTokens, outputTokens, reasoningTokens, cost, cacheSavings, latencyMs, ttftMs, retryMs, retries, droppedAttempt, stopReason, hitOutputCap, toolCallsRequested, messageCount, bytesStreamed, startedAtMs

ttftMs is the time to first token. hitOutputCap marks a call cut off at its max output tokens. droppedAttempt marks a call whose connection dropped mid-answer and was retried. The provider still bills the tokens it sent before the drop, so they're recorded.

Fleet token usage CSV#

Data → Token Usage → CSV exports the same per-phase table as token-usage.csv, across all your runs, for your own analysis or expense reports.

Requirements .md#

Export .md in the doc library saves the active requirements doc as a UTF-8 markdown file wherever you choose. Import .md brings one in as a new doc, named after the file. Use this to back up your briefs, move them between devices, or write them on a computer.