Artifacts
Artifacts are saved Mahout outputs that you can return to after the chat message is finished.
Artifact candidates
During a response, Mahout may propose one or more artifact candidates. A candidate is not saved until you choose to save it.
Common candidates include:
| Type | Examples |
|---|---|
| Analysis report | thesis summary, market memo, source-backed brief, reproducible sandbox-backed analysis |
| Backtest result | assumptions, code-backed metrics, charts, trade log |
| Chart | price chart, equity curve, drawdown chart |
| Data table | comparison table, metrics table, scenario matrix |
Save only the outputs you expect to revisit.
Saved artifacts
Saved artifacts appear in the Artifacts view. Each saved artifact belongs to your account and includes its content plus metadata such as type, source session, and creation time.
Use saved artifacts for:
- preserving useful research from long chats
- comparing outputs over time
- rerunning code-backed work
- returning to a chart or table without replaying the whole conversation
Reruns
Some artifacts can be rerun in a fresh sandbox. Reruns are useful when you want to verify that code-backed analysis still works or when you want a fresh execution result with the same saved artifact context.
A rerun is refused with an ARTIFACT_SOURCE_REMOVED error when the artifact was built with a data source that has since been removed from Mahout. The saved artifact itself is kept — only the rerun is blocked.
Reruns may also fail if:
- the underlying data source is temporarily unavailable
- the sandbox provider is unavailable
- the artifact was not designed to be rerun
- your account has reached a rate or cost limit
Do not assume a rerun reproduces the original output byte for byte. After a rerun, re-check the output’s data-source and provenance list to confirm which sources were actually used before relying on it.
Deleting artifacts
Delete artifacts you no longer need from the artifact detail or list view when the UI exposes the delete action.
Deletion removes the saved artifact record. It does not rewrite old chat messages that mentioned the artifact candidate.
Source and provenance metadata
Artifacts can carry data-source metadata. This helps you understand where the output came from and what should be verified before relying on it.
Look for:
- source names
- source freshness
- source permissions
- generated time
- assumptions and limitations
If source provenance is missing or unclear, ask Mahout to restate the analysis with source details.
Practical tips
- Save final charts, tables, and research notes rather than every intermediate attempt.
- Keep separate sessions for separate research questions so artifact history stays readable.
- Rerun code-backed artifacts before relying on older outputs.
- Treat artifacts as research records, not trading instructions.