Conversations and Activity

The local smartsite_logs table records successful, failed, and blocked turns with full prompt/response or error plus operational metadata.

Chat History is the detailed operational record stored by WordPress. Each row can include the full question and answer or error, model and token counts, duration, page and device context, channel, knowledge-routing information, tool diagnostics, and security flags. It is the most useful troubleshooting screen and also the most privacy-sensitive administrator screen.

The practical starting point is to set a narrow date range. Continue through the workflow until you use cleanup for logs older than 30, 90, 180, or 365 days, then use the field notes below to understand which choices affected AI output and which only changed delivery, access, or presentation.

What this feature does and when to use it

Use filtered details for troubleshooting, moderation, QA, and privacy operations.

Use this feature in the following situations:

You need to investigate one wrong answer, error, slow response, tool call, or blocked message.
You need to compare web and WhatsApp activity or isolate a specific assistant and date range.
You are carrying out an approved local retention cleanup or conversation deletion.

Where to find it

WordPress locationWordPress Dashboard → AI Website Chat → Analytics → Chat History

Before you begin

SmartSite Assistant is installed and activated.
You are signed in with an account that can manage WordPress options.

Set it up step by step

Set a narrow date range.
Filter by search text, assistant, status, device, or channel as needed.
Optionally filter token ranges and sort by date, duration, or token metrics.
Open one conversation/turn and inspect routing, performance, knowledge, tools, and security details.
Record sanitized evidence before deleting anything.
Delete a turn/conversation only under an approved retention or privacy process.
Use cleanup for logs older than 30, 90, 180, or 365 days.

Fields, controls, and important values

These filters help you find the small set of conversations that can explain a problem without browsing unrelated visitor data. The details may show the model, timing, knowledge search, tool activity, errors, and security flags behind a response. Reviewing that evidence can lead to better sources or instructions, but sorting and filtering alone never changes future AI behavior.

Fields, controls, and important values
Field, control, or status	What SmartSite Assistant does with it	How to use it and why it matters
Search	Finds matching text in locally stored visitor prompts and assistant responses.	Enter the smallest distinctive word or phrase needed to find the incident. Search examines stored prompt and response content, so avoid exposing the results and clear the term before another administrator uses the screen.
Date from / Date to	Restricts results to a chosen local-log date range so unrelated conversations are not reviewed.	Start with the shortest period that includes the event. The dates reduce the local rows being reviewed; they do not delete anything or change the date range used by OpenAI cost analytics.
Assistant filter	Shows rows recorded for one saved assistant ID/name.	Choose the local assistant that handled the turn when several profiles share the log table. The filter narrows the evidence displayed and does not activate, edit, or rerun that assistant.
Status filter	Narrows rows by recorded outcome such as success, error, or blocked.	Use the recorded outcome to separate completed turns from failures or security blocks. Open a matching row to find the actual error or flag; the status alone does not identify the cause.
Device filter	Narrows rows using the recorded desktop/mobile device classification.	Use this to compare rows classified as desktop or mobile from their stored request context. It is useful for reproducing interface issues but is not a precise identity or device fingerprint.
Channel filter	Separates website (`web`) activity from WhatsApp activity.	Choose web or WhatsApp to isolate the delivery path that needs investigation. This is important because the channels share parts of the AI and security pipeline but have different transport and setup failures.
User-token filter	Finds rows associated with an exact browser/channel token; treat the token as sensitive pseudonymous data.	Paste an exact browser or channel token only when investigating an authorized case. It groups related activity but is pseudonymous rather than a verified person, so protect it like other visitor data.
Token minimum / maximum	Restricts results by recorded token counts to find unusually small or large turns.	Use these boundaries to find unusually small or expensive turns based on recorded token counts. They filter existing logs only; they do not impose a runtime token limit or change OpenAI billing.
Sort field and direction	Orders results by date, duration, or token measures in ascending or descending order.	Order the current result set by date, duration, or token measure to bring the relevant outliers to the top. Sorting changes presentation only and should be combined with a narrow filter scope.
Results per page	Shows 10, 25, or 50 rows per page without changing stored data.	Choose 10, 25, or 50 rows to balance scanning with the amount of sensitive data shown at once. This pagination setting neither changes retention nor limits how many conversations are recorded.
Delete selected conversation/turn	Permanently removes the selected local log scope after confirmation.	Use this only for the exact row or conversation approved for removal. Confirm the scope before accepting the prompt; the action permanently removes local log data but does not prove deletion from external services.
Database cleanup	Deletes local logs older than 30, 90, 180, or 365 days, or deletes all logs. No automatic schedule was found.	Use an approved retention period to remove local logs older than the selected age, or all logs only when explicitly required. Take needed sanitized evidence first because the plugin has no automatic cleanup schedule or undo.

How to confirm it is working

Judge Conversations and Activity where its effect is actually consumed—by the administrator, visitor, model, or connected service. The expected result above is more useful than a green badge because it describes the behavior the configuration was meant to produce.

Practical example

Filter Status=blocked and Channel=WhatsApp for the last seven days before reviewing a reported moderation issue.

Recommended practice

Change one part of Conversations and Activity at a time and keep a short record of the previous value and test result.
Turn one observed pattern into one controlled configuration change, then compare new conversations with the earlier evidence.

Important warnings

Common problems and focused checks

Common problems and focused checks
Problem	What to check and what to do next
Conversations and Activity is missing or does not match this guide.	Confirm the plugin is active and the account can manage WordPress options. Narrow the date and filter scope, then open source records instead of relying on an empty chart or summary alone.
A change on Conversations and Activity does not produce the expected result.	Keep the exact notice and test case, then review the browser console and WordPress/PHP log. Narrow the date and filter scope, then open source records instead of relying on an empty chart or summary alone.

Screen reference

Screenshot capture pending /images/screenshots/analytics/conversations.webp

Conversations and Activity

Capture: Show Chat History filtered to fictional web conversations with one detail drawer open and all content/token/IP values redacted.
Show: Filters, results, status/channel badges, detail sections, cleanup control
Viewport: Desktop, 1440 × 900
Annotate: Use numbered callouts only for controls referenced in the procedure.
Redact: OpenAI keys, tokens, secrets, personal information, private URLs, IP addresses, and conversation text