Period
Team
Role
Last 30 days
Messages
—
Checked by the guardrail
Blocked
—
— of messages
Attack attempts
—
Jailbreak, injection and harmful requests
Users
—
Registered accounts
Estimated time saved
—
From automated answers
Messages per day
Click a day to see its messages Last 30 daysNeeds attention
For the selected periodLoading…
Highlights
- Loading…
Usage by segment
Click a row to filter the overview| Department | Users | Messages | Volume | Safe | Blocked | Safe / Blocked |
|---|
Most active users
Ranked by messages| User | Team | Messages | Activity | Safe | Blocked |
|---|
Top intents
Guardrail (safety) readRouting distribution
Which agent handled messagesRouting confidence
Average —Guardrail health
Uptime
—
Checks completed by the classifier
Latency p95 / p99
—
Added before every reply
Tokens per check
—
Prompt tokens, avg
Low-confidence verdicts
—
Below 0.6 confidence, worth a review
Period
Total tokens
—
Input + output, every LLM call
Cost
—
At the prices below, when each call was made
Tokens per turn
—
Average per answered message
Cache hit
—
Input tokens served from Azure's prompt cache
Tokens per day
Input + output tokens per dayBy step
Tokens per pipeline step; hover for cost and latencyBy agent
Whole turns, by the agent that answeredTop users
By tokens; click to open the user's reportModel prices
Per 1M tokens · used to cost each call| Model | Input | Cached | Output | Cur. |
|---|---|---|---|---|
| Loading… | ||||
Pending
0
Awaiting decision
Approved
0
Active accounts
Rejected
0
Denied access
All
0
Total requests
| Employee | Department | Manager | Role | Status | Requested | Last login |
|---|
Registration breakdown
Requested role
Manager vs employeeApproval status
Registration outcomesBy department
By manager
Approval audit trail
Immutable log of registration submissions, admin approve / reject / role actions and password resets.
| When | Action | Performed by | Target user | Request # | Comments |
|---|
Threats
Every blocked / attack message across all users, newest first.
| When | User | Message | Intent | Attack type and framework | Flags |
|---|
Agent Actions
Every tool an agent ran for a user, newest first.
| When | User | Agent | Tool | Arguments | Result | Time |
|---|
Misroutes
SAFE messages where the guardrail verdict ≠ the routing decision. Your routing-tuning list.
| When | User | Message | Guardrail | Routed to |
|---|
Message search
Search message text, or filter by date, user or intent.
| When | User | Message | Routed to | Policy |
|---|