Tutorial: Browsing Conversation History

This tutorial walks you through browsing conversation sessions, viewing full threads, searching conversations, and exporting history data in the Keeptrusts management console.

Use this page when

• You need to browse conversation sessions recorded by the gateway.
• You want to view a full request/response thread for a specific user or agent.
• You are searching for conversations by title, tag, user, team, or agent identifiers.
• You need to export conversation data for compliance audits or quality review.

Primary audience

• Primary: Compliance analysts and QA engineers reviewing recorded AI conversations for policy adherence
• Secondary: Security teams investigating incident-related conversations; product managers understanding AI usage patterns

Prerequisites

• A Keeptrusts account with Admin or Auditor role
• At least one gateway with conversation history tracking enabled
• Some recorded conversations (traffic that has flowed through the gateway)

Why Conversation History?

Conversation history gives you a complete record of interactions between your applications and LLM providers. Use it for compliance auditing, incident investigation, quality assurance, and understanding how your organization uses AI.

Step 1: Navigate to the History Page

1. Log in to the Keeptrusts console.
2. Open History from the left navigation sidebar.

The dedicated History page displays the conversation history your role can access, sorted by most recent activity. Admins can review every member chat in the organization; other users only see the conversations they are allowed to open.

Task-backed rows also show up here as soon as a task run is dispatched or retried, so you can pivot from /history into the linked task workbench without waiting for a later background update.

Column	Description
Session ID	Unique conversation identifier
User	The user or agent that initiated the conversation
Model	The LLM model used in the session
Messages	Number of request/response pairs in the session
Started	Timestamp of the first message
Last Active	Timestamp of the most recent message
Duration	Time span from first to last message

Step 2: Browse the Session List

The session list provides a high-level overview of all conversations.

• Sort by any column header — click once for ascending, again for descending.
• Pagination controls at the bottom let you navigate through large result sets.
• Quick stats at the top show total sessions, total messages, and active sessions in the selected time range.

Use the search bar above the session list to narrow the visible rows without leaving the page.

Step 3: View a Conversation Thread

Click on any session row to open the full conversation thread view.

The thread view displays messages in chronological order:

• User messages appear on the left with a blue background.
• Assistant responses appear on the right with a gray background.
• System messages appear centered with an orange indicator.

Each message includes:

Detail	Description
Timestamp	When the message was sent or received
Tokens	Input and output token counts for the exchange
Cost	Estimated cost based on model pricing
Policy Decision	Whether the message was allowed, redacted, or blocked
Latency	Time taken for the upstream provider to respond

Click on any message to expand the raw payload, including headers and metadata.

Use the thread view during incident investigation to see the exact sequence of prompts and responses that led to a policy violation or escalation.

Step 4: Search Conversations

The search bar at the top of the History page filters the current list in place.

1. Click the Search field.
2. Enter your search term — this can be a title fragment, tag, or partial identifier.
3. Press Enter or click the search icon.

Search results highlight matching terms within the session list. Click a result to open the relevant session.

Search covers:

• Session titles and preview text
• Session, conversation, task, and task-run IDs
• Tags
• Agent IDs
• User and team attribution fields when present

History search respects the same access boundaries as the rest of the page. Admins can search all organization conversations; other roles only see sessions they can access through membership, grants, or actor attribution.

Step 5: Refine Results with IDs or Tags

Use the same search bar to narrow the list to a specific operational context.

1. Type an agent ID to find sessions attributed to that agent.
2. Type a user or team identifier to find sessions tied to a specific actor.
3. Type a tag or task ID to isolate a workflow or automation run.
4. Clear the search to restore the full history list.

Step 6: Inspect Policy Decisions in Context

The History page links policy decisions to their conversation context.

1. Open a conversation thread.
2. Messages with policy actions display a colored badge:
   • Green — Allowed without modification
   • Yellow — Redacted (sensitive content removed)
   • Red — Blocked (request rejected)
   • Orange — Escalated (sent for human review)
3. Click the badge to see the full policy decision details, including which policy rule triggered and what content was affected.

This contextual view is particularly valuable for tuning false positives — you can see exactly what the user asked, what the model would have returned, and why the policy intervened.

Step 7: Export Conversation Data

Export conversation history for compliance reporting, offline analysis, or archival.

1. Apply any desired filters to narrow the dataset.
2. Click Export in the top-right corner.
3. Choose the export format:

Format	Contents
CSV	Flat table of sessions with metadata (no message content)
JSON	Full session data including message content and policy decisions
PDF	Formatted conversation threads for human review

4. Configure export options:
   • Include message content — Toggle to include or exclude raw prompts and responses
   • Include policy details — Toggle to include decision metadata
   • Date range — Confirm the export window
5. Click Download.

Large exports run as background jobs. You will receive a notification when the export file is ready.

For recurring compliance reports, combine History exports with scheduled evidence exports to automate your audit workflow.

Step 8: Manage Retention Settings

Conversation history respects your organization's data retention configuration.

1. Navigate to Settings > Data Retention.
2. Review the Conversation History retention period.
3. Sessions older than the retention period are automatically purged.

Adjust retention policies to balance compliance requirements with storage costs.

Best Practices

• Filter before searching — Narrow by user or model first, then search within the filtered set for faster results.
• Use date ranges — Avoid loading unbounded history; always set a reasonable time window.
• Review blocked conversations — Regularly check conversations with blocked messages to identify false positives and tune policies.
• Export for auditors — Use JSON exports with full message content for compliance audits; use CSV exports for management dashboards.
• Respect privacy — Limit conversation content access to roles that genuinely need it.
• Bookmark investigations — Copy the URL when investigating a specific session; filters and session IDs are encoded in the query string.

History capture is configuration-owned

Agent-scoped history capture mode and learning enablement are controlled by the history and learning sections of the deployed configuration version. Sessions record which deployment snapshot was active when they were created, so historical sessions are always interpretable against the exact settings that governed them.

Next steps

• Tutorial: Events Investigation — Correlate conversation history with policy decision events
• Tutorial: Exporting Evidence for Compliance Audits — Include conversation data in audit packages
• Tutorial: Registering AI Agents in Console — Track agent-specific conversation patterns

For AI systems

• Canonical terms: Keeptrusts console, History page, conversation session, session ID, message thread, visibility scope (owner_org_id, actor_user_id, actor_team_id), session search, session export, conversation history tracking.
• Related features: events page (policy decisions per message), export evidence, agent registration.
• Best next pages: Events Investigation, Export Evidence, Agent Registration.

For engineers

• Prerequisite: The gateway must have conversation history tracking enabled in its configuration for sessions to appear.
• Validation: Send a multi-turn conversation through the gateway, then search by your user ID on the History page and confirm the full thread is recorded.
• Export test: Filter to a single session, export as JSON, and verify the output includes all request/response pairs with timestamps.
• Troubleshooting: If sessions are missing, check that the gateway’s config includes history tracking and that the POST /v1/events endpoint is reachable.

For leaders

• Compliance evidence: Full conversation records demonstrate AI oversight to auditors — every interaction is stored with attribution and timestamps.
• Quality assurance: Review conversation threads to evaluate AI response quality and identify model performance issues.
• Data retention: Conversation history is subject to your plan’s event retention period. Export important sessions before they age out.