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

# Data Pipeline

> 元のデータが、標準化された記録と出典付きの回答になるまで。

対象 mirobody
`1.5.10`
· 元ファイル
[`docs/site/pipeline.md`](https://github.com/thetahealth/mirobody/blob/b37cd7b24af5251a4364aab30905a7497fc69dcb/docs/site/pipeline.md)

> [!NOTE]
>
> このページの日本語版はまだありません。以下は英語の原文です。

Mirobody collects health records, standardizes names and units, and makes the
results available to the web client, agent and MCP clients.

![Collect, standardize and query health records](https://mintcdn.com/thetahealth/AWMM79o21zua_xVu/images/oss/docs/images/collect-standardize-agent.svg?fit=max\&auto=format\&n=AWMM79o21zua_xVu\&q=85\&s=b563ac92a9e4f918eb9fe9cb783833a1)

## From source to answer

| Stage | Result |
| - | - |
| ① Collect | Accept files, device data and journal entries; retain their source. |
| ② Standardize | Resolve recognized indicators to standard codes and convert compatible units. |
| ③ Agent | Query the selected record, compare results and cite the supporting data. |

A file can finish uploading before its readings finish processing. Check the
extraction result before assuming that its measurements are available.
Device records may be combined into daily values; queries state which view
and date window they use.

## Accuracy and corrections

* An unrecognized name remains unresolved; the engine does not invent a code.
* Incompatible units stay separate rather than being forced into one series.
* Query results retain source information so you can inspect the original file.
* Corrected or removed records affect later queries and citation lookups.
* Missing data means no recorded result, not evidence that a condition is absent.

## Data processing

Terminology and unit conversion run offline. Document recognition and chat
use the models selected by the deployment; cloud models receive the inputs
needed for their tasks. See [model selection](https://docs.mirobody.ai/ja/model-choice.md) for local,
cloud and hybrid data flows.

For integration details, see the [full pipeline reference](https://github.com/thetahealth/mirobody/blob/b37cd7b24af5251a4364aab30905a7497fc69dcb/docs/pipeline.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.