> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mithunai.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Usage reports and unanswered questions

> Measure how much your assistants are used and how often they answer, then read the questions they declined: each one names a gap your content could close.

MITHUNAI reports two things: how much your assistants are used and how well they answer, and which questions they could not answer. Find both in the console under **Operations**.

## Usage

**Operations → Usage** shows, for a time window:

| Metric                   | Meaning                                                                                     |
| ------------------------ | ------------------------------------------------------------------------------------------- |
| **Conversations**        | Conversations started                                                                       |
| **Questions**            | Questions asked                                                                             |
| **Grounded answers**     | Answers supported by your content, with citations                                           |
| **Declined to answer**   | Abstentions: questions your content did not cover                                           |
| **Failed**               | Answers that could not be produced, for example because a provider failed                   |
| **Rates**                | Grounded, abstention and failure rates as a share of answers                                |
| **Answers by model**     | Which models wrote the answers                                                              |
| **Questions by channel** | `user` for the console, `service` for API keys and MCP clients, `widget` for website embeds |

Filter the report to one assistant to compare assistants. The report contains counts only, never question or answer text.

With the API, use [Get usage](/api-reference/analytics/get-usage) with `days` for a trailing window, or `start` and `end` for an exact range of up to 366 days.

Owners, Administrators and Editors can read usage.

## Unanswered questions

**Operations → Unanswered** lists the questions your assistants declined to answer, newest first, for the last 7, 30 or 90 days. Each row shows the question, when it was asked, the assistant and the channel. It never shows who asked.

This is your content backlog. For each question:

1. Decide whether your assistant *should* answer it. Off-topic questions are correctly declined.
2. If it should, find or write the content that answers it.
3. Re-sync the source, and ask the question again.

Because it shows questions other people asked, the unanswered report is limited to Owners and Administrators. With the API, use [List unanswered questions](/api-reference/analytics/list-unanswered).

## Reading the numbers

* **A high abstention rate** usually means a content gap, not a product fault. Check the unanswered report before changing anything else.
* **A sudden jump in abstentions** after a change often means a source failed to sync or an assistant lost its knowledge. Check the [ingestion jobs](/knowledge/ingestion).
* **Failures** point at the model provider or the deployment. If they persist, contact your MITHUNAI administrator.

## Not yet available

Time-series charts, export, per-source breakdowns and cost reporting are planned. See [Product status](/resources/product-status).