Tool Responses: AI Tool Execution Logs

Response log overview

The Tool Responses page shows a complete table of all tool invocations made by AI assistants with the corresponding responses received.

Interface Components

📅

Date filter

Two date fields (Date from / Date to) allow filtering responses by specific time range. Use the "SEARCH" button to apply the filter.

🔍

Tool filter

Text field "Filter by Tool" to search for responses from a specific tool (e.g., "calendar_list", "send_sms", "atlesa").

🆔

UUID filter

Field to filter all responses of a specific conversation using the unique call UUID.

🏷️

Quick filters

Quick filter buttons: TOOLRESP_ALL (all), RAG (only RAG), TOOL (only standard tools).

📊

Total counter

Displays the total number of responses found with applied filters (e.g., "Total: 234").

Table Columns

Date

Complete timestamp of the response received from the tool in format DD/MM/YYYY HH:MM:SS.

Example: 12/12/2025 10:18:33

UUID

Unique conversation identifier. All tool calls from the same conversation share the same UUID.

Example: ee2843c2-e927-46d7-a89d-300a07f880d4e

Tool

Name of the tool invoked by the AI assistant.

Examples: atlesa, calendar_list, calendar_delete, calendar_create_if_free, send_sms

Type

Badge identifying the tool category. Currently always shows TOOL for standard tools.

Response

Truncated preview of the JSON payload received from the tool. Click the ⓘ icon at the end of the row to see the complete response.

Example: {"success":true,"message":"OK YOU CAN PROCEED","waited_seconds":5}

Detail Action

Clickable ⓘ (info) icon that opens a modal with the complete JSON of the response formatted in a readable way.

Response detail view

Tool Response Detail

Clicking the ⓘ icon opens a "Tool Response Detail" modal showing the complete JSON of the tool response with syntax highlighting.

Success Response

For tools that executed correctly, the modal shows the complete structured payload.

Error Response

For tools that failed, the modal shows the error payload.

Use cases

🔍 Debug tool errors

Scenario: a customer reports that the assistant cannot book appointments.

Action:

Filter by UUID of the problematic call
Identify which tools were called
Open detail of calendar_create_if_free tool response
Check error payload: "Invalid or expired AI user token"
Diagnosis: Google Calendar token expired → renew authorization

📊 Tool performance analysis

Scenario: verify how many times a specific tool is called.

Action:

Filter by tool send_sms
Set date range for last week
View total count and analyze temporal patterns
Open details to verify success rate

🎯 Specific conversation tracking

Scenario: analyze all tool calls of a single conversation.

Action:

Copy UUID from conversations page
Paste in UUID filter
View complete sequence of called tools
Verify chronological order and payload of each

⏱️ Temporal audit

Scenario: verify tool activity in a specific time window.

Action:

Set Date from: 12/12/2025, Date to: 14/12/2025
Click SEARCH
Analyze total count and tool distribution
Filter by type (RAG vs TOOL) for granular analysis

Response examples

✅

Success with data

Payload with "success": true and structured data (events, messages, query results).

{"success": true, "events": [...], "summary": "..."}

✅

Operation success

Execution confirmation without additional data.

{"success": true, "message": "Calendar event deleted successfully"}

⏳

Wait completed

Delay/wait tool returning waited time.

{"success": true, "message": "OK YOU CAN PROCEED", "waited_seconds": 5}

❌

Authentication error

Failure due to expired or invalid token.

{"error": "Invalid or expired AI user token"}

📧

SMS response

Send confirmation with recipient details.

{"status": "sent", "response": "200", "alias": "default", "destination": ...}

📅

Calendar data

Array of events with timestamps, location, Google Calendar link.

{"success": true, "events": [{...}], "htmlLink": "https://..."}

? What exactly does a row in the Tool Responses table represent?

Each row is a single response received from a tool after the AI assistant invoked it during a conversation. It includes timestamp, conversation identifier (UUID), name of called tool and complete JSON payload of the response.

? How can I see the complete JSON of a response instead of the truncated preview?

Click the ⓘ (info) icon at the end of the row. A "Tool Response Detail" modal will open with the complete formatted and scrollable JSON.

? What does the "TOOL" badge in the Type column mean?

Identifies that the response comes from a standard tool (not RAG). It is a visual category indicator that allows quick distinction of call types.

? How do I filter all responses from a specific conversation?

Copy the conversation UUID (visible in the UUID column or from the conversations page) and paste it in the "UUID" field above the table. All tool calls from that conversation share the same UUID.

? What does "Total: 234" above the table indicate?

It is the counter of the total number of responses that match the currently applied filters. If you have no active filters, it shows all responses recorded in the system.

? How can I search for responses from a specific tool?

Use the "Filter by Tool" field and type the tool name (e.g., "send_sms", "calendar_list", "atlesa"). Then click SEARCH to apply the filter.

? What are the TOOLRESP_ALL, RAG and TOOL buttons?

They are quick filters for response type. TOOLRESP_ALL shows all responses, RAG filters only responses from knowledge base/RAG, TOOL filters only standard operational tools.

? Why do I see "Invalid or expired AI user token" in some responses?

This error indicates that the authentication token for an external service (e.g., Google Calendar) has expired or is invalid. You need to renew the authorization in the tool configuration.

? How do I interpret a response with "success": true?

It means the tool successfully executed the requested operation. The payload may contain additional data (events, messages, results) or simply operation confirmation.

? Can I see the chronological order of tool calls in a conversation?

Yes. Filter by conversation UUID and sort by Date column. You will see the temporal sequence of all tools called during that conversation.

? How do I identify if a tool is failing frequently?

Filter by specific tool name and scroll through responses. If you see many payloads with "error" field instead of "success": true, you have a recurring issue to investigate.

? Can I export table data for external analysis?

Currently there is no direct export function in the shown interface. For bulk analysis or data export, contact support or use AI4CALL API if available.

? How long are tool responses kept in the system?

Retention policy depends on the plan. Generally, tool responses are kept for debugging and audit purposes. Verify the policies of your specific plan.

? Can I see which AI assistant called a tool?

From the conversation UUID you can trace back to the assistant. The UUID allows you to correlate with the conversations page where you find assistant details and configuration.

? How do I use this page for production debugging?

When a customer reports a problem: (1) Find timestamp/UUID of problematic call, (2) Filter here by UUID, (3) Identify called tools, (4) Open error payload details, (5) Diagnose cause (expired token, wrong parameters, service down).