S3 AutoSync

S3 AutoSync ingests files from an S3-compatible bucket (OTC Object Storage, OBS) into your Smartchat Space on a schedule. New, changed and removed files in the bucket are picked up and become searchable context for every member of the Space.

Prerequisites

You need the Space Owner role in the Smartchat Space. See Roles for details.
An S3 bucket, its endpoint URL and an access key pair with read access to it. Files are uploaded to the bucket with your own tooling (for example boto3, the OBS Browser or the Huawei SDK).
S3 AutoSync must be enabled for your Smartchat environment. If the S3 AutoSync page is missing, ask your Application Owner.

OTC OBS does not support the S3 request-checksum / integrity signing that recent boto3 versions (> 1.36) enable by default. Uploading with it enabled corrupts the stored files: they lose metadata and can no longer be ingested.

When uploading to the bucket with boto3, disable it by setting both request_checksum_calculation and response_checksum_validation to when_required on the client config:

from botocore.config import Config

s3 = boto3.client(
    "s3",
    config=Config(
        request_checksum_calculation="when_required",
        response_checksum_validation="when_required",
    ),
)

Alternatively, upload via the OBS Browser or the Huawei SDK.

Step 1 — Open S3 AutoSync

Click your profile image, choose Administration, and pick the Space in Selected Space under Space Owner. Then open S3 AutoSync. The table lists the S3 connections of this Space and refreshes itself every minute.

S3 AutoSync page with one active connection

Step 2 — Add a connection

Click Add S3 connection and fill in the form:

Add S3 connection form with folder rules

Source

Field	Description
Name	A name for the connection, shown in the table.
Endpoint URL	The S3 endpoint, e.g. `https://obs.eu-de.otc.t-systems.com`.
Bucket name	The bucket to watch: lowercase letters, digits, dots and hyphens.
Region	The bucket’s region, `eu-de` for OTC.

Credentials

Field	Description
Access key ID / Secret access key	A key pair with read access to the bucket. The values are stored encrypted and never shown again; leave both blank when editing to keep them.

Synchronization

Field	Description
Sync interval	How often the bucket is checked: Every 15 minutes, Every hour (default), Every 6 hours, Every 12 hours, Daily or Weekly.
Ingest pipeline	Only shown when the Space has more than one ingestion configuration. Choose which one processes the synced files.
Target collection	Where files land that no folder rule covers. General collection is the Space’s default collection. Changing it later moves already synced files on the next sync.
Folder rules	Route folders of the bucket into other collections, see Folder rules.
Active	Leave on. Paused connections keep their settings, but the bucket is not checked.

Click Save. Smartchat checks the bucket with the given credentials right away:

On create, an unreachable bucket is rejected. Nothing is stored until endpoint, bucket and keys are correct.
On edit, the connection is saved either way; if the bucket is not reachable the sync is paused (toast Connection saved, but the bucket is not reachable. The sync stays paused until the settings are fixed.).

The first run starts with the next interval. Files already in the bucket are ingested on that run; from then on only additions, changes and deletions are processed.

Folder rules

By default every file of the bucket lands in the Target collection of the connection. If you funnel several data sources into one bucket, you can instead route each folder into its own collection with folder rules.

In the form, click Add folder rule and enter the folder path inside the bucket (e.g. reports/2026) and the Target collection for it. Add one row per folder; a folder may appear only once.

How folders are matched

A rule applies to every object inside the folder, at any depth. reports/2026 matches reports/2026/q1.pdf and reports/2026/archive/old.pdf.
Matching stops at folder boundaries: reports/2026 does not match reports/2026-old/x.pdf, and a rule reports does not match reportsarchive/x.pdf.
If there is a rule for a folder and another rule for one of its subfolders, the rule for the subfolder takes precedence. The order of the rules does not matter.
Files that no rule covers use the connection’s Target collection.
Leading and trailing slashes are ignored.

Example with the connection’s Target collection set to Misc:

Rule	Files	Collection
`reports/2026`	`reports/2026/q1.pdf`	Finance 2026
`reports`	`reports/2025/q4.pdf`, `reports/summary.pdf`	Finance
(none)	`reports/2026-old/x.pdf`, `hr/leave.pdf`, `readme.txt`	Misc

Changing the rules

Changed rules take effect on the next scheduled sync. Files that were already synced under the previous rules are moved: on the next run AutoSync re-ingests each affected file into its new collection and removes it from the old one, exactly as if the file had changed in the bucket. The same happens when you change the connection’s Target collection.

Keep in mind:

Moving is a re-ingest. A rule that covers many files causes correspondingly many ingestions, processed in batches over consecutive runs.
A file whose move fails shows up with a failed status in the collection, like any other failed ingestion, and is retried on the next change of that file.
Only files that AutoSync itself uploaded are moved. A copy of the same file that someone uploaded manually keeps its collection.

Monitoring a connection

Column	Meaning
Status	Active: scheduled and running. Paused: settings kept, bucket not checked. No schedule: saved without a schedule yet.
Bucket	The bucket name; hover for the endpoint URL.
Collection	The target collection and the number of folder rules.
Schedule	The interval and Next: the time of the next run.
Last sync	When the last run finished, with counts of Success, Failed and Pending files.

Click the connection’s Name to copy it to the clipboard.

Sync statistics

Sync statistics opens a summary of the connection: All time totals for Files processed, Added, Updated and Removed, and a table of the Last sync runs with the same counts per run.

Sync statistics dialog

Failed files are listed with their error in the target collection under Workspace → Knowledge; see Ingestion Errors for the meaning of each status.

Managing connections

Button	Action
Test connection	Checks whether the bucket is reachable with the stored credentials.
Pause sync / Resume sync	Stops checking the bucket until you resume. Already synced files stay in the Space.
Sync statistics	Opens the statistics dialog described above.
Edit S3 connection	Change endpoint, bucket, credentials, interval, target collection or folder rules.
Remove S3 connection	Deletes the connection and its schedule. Content already synced into the Space stays.