Transparency & Audit
In regulated and high-stakes environments, knowing that an AI gave an answer is not enough: you need to know why it gave that answer, what data it used, and how it arrived at the conclusion. Squad is built for full traceability from question to answer.
Retrieval Provenance
Every answer Squad produces carries a chain of evidence back to its source material.
When AIM retrieves information from the knowledge graph, the response includes:
- Source episodes: The specific document sections or conversation turns that contributed to the answer
- Entity references: Which knowledge graph entities were consulted
- Retrieval method: Whether the answer came from a direct graph lookup, a semantic search, a stored workflow, or an external source
- Confidence indicators: How strongly the retrieved information matched the query
This means a user can always ask “where did this come from?” and trace the answer back to a specific paragraph in a specific document.
Workflow Data Lineage
When Squad executes a multi-step workflow, every step in the chain is recorded with its inputs, outputs, and the tool that produced them.
For a completed workflow run, you can inspect:
- Step-by-step execution log: What each step did, in order
- Input/output data: What data each step received and what it produced
- Tool identification: Which tool was invoked at each step
- Timing and status: How long each step took and whether it succeeded or required recovery
Each workflow execution is isolated with its own artifact space, ensuring that runs do not interfere with each other and that all intermediate artifacts are preserved for inspection.
Immutable Audit Trail
Every interaction with the platform generates an immutable, append-only audit record.
| Audit Record | What It Captures |
|---|---|
| Correlation ID | Unique identifier per execution, flows through all logs and events |
| User identity | Authenticated user who initiated the action |
| Session context | Which session and conversation the action belongs to |
| Event stream | Real-time SSE events with per-node granularity as AIM processes a query |
| Decision points | Classification results, disambiguation choices, risk assessments |
The audit trail is append-only by design: there is no mechanism to delete or modify historical records. This provides a tamper-evident log suitable for regulatory compliance and incident investigation.
Graph Exploration
Squad provides interactive tools for browsing the knowledge graph directly, allowing users and administrators to:
- Explore entities, relationships, and episodes visually
- Verify that ingested data was correctly extracted and linked
- Inspect the provenance chain for any entity or fact
- Understand the graph structure that underpins the platform’s answers
For details on the graph exploration interface, see Graph Visualization.
Related
- Guardrails & Safety: the safety controls that generate audit events
- Security & Governance: authentication, data protection, and compliance
- Administration & Access: who can access what, and how access is controlled
- Graph Visualization: interactive knowledge graph exploration