Enabling conversation history
Saved conversation history (previous chats) enables Spotter 3 to evolve from a session-based search tool into a persistent analytical assistant. By enabling this feature, you allow users to preserve the context of their insights, allowing them to return to complex analyses days or weeks later without losing progress.
Governance and administration
ThoughtSpot ensures that saved conversations remain as secure as a live search.
Saved conversation history is disabled by default. Additionally, no data caching is used in the conversation history; instead, Spotter preserves the analytical logic (tokens) used to generate insights and reruns a live query as requested by the user.
This approach ensures that every time a user resumes a saved chat, their data remains secure and accurate.
- Live data fetch in history
-
Results are refetched from the cloud data warehouse, rather than being cached. As a side effect, the data fetched might be out of sync with the query data.
- Integrated governance
-
The chat history feature is fully integrated with your existing security model. If access to a data model is revoked, associated conversations are automatically protected from unauthorized viewing.
- Permission sync
-
Spotter automatically re-verifies row-level security (RLS) and column-level security (CLS) in real-time. Upon data fetch, it applies the latest privileges the user has.
| If a column or row is redacted after the original conversation in which the row or column was used, the text analysis generated by Spotter may still contain references to this data, but the user will not be able to query this data from the conversation history, nor will Spotter have access to this redacted data during a follow-up. |
Data retention policy
ThoughtSpot retains the conversation history for 180 days if the time to live (TTL) countdown begins from the updated_at timestamp of the last message in a thread.
| Retention policies are applied in batches. If you enable and disable conversation history before the maintenance cycle finishes, some old conversation might still be visible. Please give the system 48 hours to complete the maintenance job. |
Dynamic schema handling
Spotter is built to handle schema changes gracefully. If an underlying data model is updated or columns are renamed, Spotter will fail, but it then attempts to resolve the saved query using the remaining valid context to ensure continuity.
Handling impermanent artifacts
Spotter does not store impermanent artifacts directly, but remembers the context in which they were fetched or generated. If this context is sufficient to address any follow-up questions, Spotter relies on it. Otherwise, it will attempt to refetch and regenerate any impermanent artifacts such as coding output, generated files, or data fetched from the cloud data warehouse to answer the user’s query.
Feature enablement (opt-in)
To provide your users with access to past conversations, follow these steps:
-
Navigate to the Admin panel, select All Orgs, and select ThoughtSpot AI.
-
Locate the Enable chat history toggle under Spotter 3 capabilities and switch it to Enabled.
Upon toggling the feature on, all user conversations with Spotter will be stored in conversation history, and available in the conversation history left rail.
