ANTE User ManualAI Usage Logs
AI Usage Logs give you a complete, request-by-request record of every AI call made across the system, so you can review what was sent, what came back, how many tokens it used, and whether it succeeded.
⚠️ Developer Access Only
This page lives in the Developer portal and is only available to users with developer privileges.
Overview
Whenever the system uses AI — for example to chat, extract information from a document, read an image, or summarize text — a log entry is created. The AI Usage Logs page lets you browse all of these entries across companies, search and filter them, and open any single call to see its full details.
This is useful for:
- Reviewing how AI is being used across companies and users
- Checking token consumption
- Investigating failed AI calls and reading their error messages
- Confirming exactly what was sent to and returned from the AI
AI Usage Logs vs. AI Statistic — Use AI Usage Logs (this page) to drill into individual AI calls, one row at a time. Use the AI Statistic page for the analytics view: totals, credential-source breakdowns, and the heaviest companies and users.
Where to Find It
In the Developer portal sidebar, open the System group and click AI Usage Logs.
Summary Cards
At the top of the page, four summary cards show totals across all logged AI activity:
| Card | Description |
|---|---|
| Total Calls | Total number of AI calls recorded |
| Total Tokens | Total tokens used across all calls |
| Prompt Tokens | Total tokens used for the input (prompt) side |
| Errors | Number of calls that ended in an error (highlighted when greater than zero) |
The Logs Table
Below the summary cards, every AI call is listed in a table, newest first. Each row shows:
| Column | Description |
|---|---|
| Date | When the call was made |
| Company | The company the call belongs to (shows "-" if none) |
| User | The user who triggered the call (shows "-" if none) |
| Operation | The kind of AI action — shown as a colored tag |
| Provider | The AI provider used |
| Model | The specific AI model used |
| Tokens | Total tokens used by the call |
| Prompt | Tokens used for the input (hidden by default) |
| Completion | Tokens used for the response (hidden by default) |
| Duration | How long the call took (shown in milliseconds or seconds; "-" if not recorded) |
| Status | Whether the call succeeded or failed |
The Prompt and Completion columns are hidden by default. You can turn them on using the table's column options.
Operation Types
The Operation column uses colored tags to show what kind of AI action was performed:
| Operation | Meaning |
|---|---|
| Chat | A conversational AI request |
| Extract | Pulling structured information out of content |
| Vision | Reading or interpreting an image |
| Summarize | Condensing text into a summary |
| Embeddings | Generating numeric representations of text |
| Stream | A streamed AI response |
Status
The Status column shows one of:
- Success — the call completed normally
- Error — the call failed (shown in red)
Searching and Filtering
Search
Use the search box to look up logs by keyword. You can search by:
- Operation
- Provider
- Model
Filters
You can narrow the list using these filters:
| Filter | Options |
|---|---|
| Operation | Chat, Extract, Vision, Summarize, Embeddings, Stream |
| Provider | OpenAI, Gemini, Claude, Grok |
| Status | Success, Error |
You can also sort the table by Date, Operation, Provider, Model, Tokens, or Duration.
Viewing a Single Log
Click any row in the table to open its detail page. The detail page is titled AI Log # followed by the log number, with a status badge in the top right. Use the back arrow to return to the list.
Request Details
| Field | Description |
|---|---|
| Operation | The kind of AI action |
| Provider | The AI provider used |
| Model | The specific model used |
| Status | Success or Error |
| Date | The full date and time of the call |
| Duration | How long the call took |
Company & User
| Field | Description |
|---|---|
| Company | The company associated with the call |
| User | The name of the user who made the call |
| User Email | The email address of that user |
Token Usage
| Field | Description |
|---|---|
| Prompt Tokens | Tokens used for the input |
| Completion Tokens | Tokens used for the response |
| Total Tokens | Combined token total |
Error Message
If the call failed, an Error Message section appears showing the full error text returned by the AI.
Input Data
The Input Data section shows what was sent to the AI — for example, the chat messages or prompt. Long text is shown in readable blocks, and chat messages are grouped by their role. If an image was sent, it is shown as a placeholder rather than the raw image data. A Truncated tag appears if the input was shortened to save space.
Output Data
The Output Data section shows what the AI returned. A Truncated tag appears if the output was shortened to save space.
Metadata
If the call recorded any extra information, it appears in a Metadata section at the bottom.
Note: Images sent to the AI are not stored in full. When a log includes an image, you will see a placeholder noting that the image data was not kept to save space.
Common Tasks
Reviewing Recent AI Activity
- Go to Developer → AI Usage Logs
- Check the summary cards for overall totals
- Browse the table, which lists the most recent calls first
Investigating a Failed AI Call
- Go to Developer → AI Usage Logs
- Set the Status filter to Error
- Click the failed call to open its detail page
- Read the Error Message section
- Review the Input Data to see what was sent
Checking Token Usage by Operation or Provider
- Go to Developer → AI Usage Logs
- Use the Operation or Provider filter to narrow the list
- Sort by the Tokens column to see the heaviest calls
Inspecting What Was Sent and Returned
- Open a log from the table
- Review the Input Data section to see the request
- Review the Output Data section to see the AI's response
Tips
- Errors first — Use the Errors summary card and the Error status filter to quickly surface problems.
- Watch token use — Sort by Tokens to find unusually large calls.
- Use the detail view — The list gives a quick overview, but the detail page shows the full request, response, and any error.
Related Pages
- AI Statistic — Analytics dashboard for AI usage totals, credential-source breakdown, and top companies/users
- User Management — System users
- Company Management — Companies in the system
💡 Note: AI Usage Logs are recorded automatically for every AI call. Truncated input or output indicates the content was shortened to save storage space.