Configure Data Sources

Under Settings → Data & Structure → Org Intelligence (/signals) you manage the data sources from which interaction signals are collected.

Available Collector Types

Type	Data Source	Required Credentials
msgraph	Microsoft Graph / Teams / Exchange	Azure Tenant ID, Client ID, Client Secret
confluence	Atlassian Confluence	Confluence URL, Email, API Token
sharepoint	SharePoint Online	Azure Tenant ID, Client ID, Client Secret + Site IDs
jira	Jira / Jira Service Management	Jira URL, Email, API Token
csv_import	Any CSV export	CSV content (paste directly or file path)
filesystem	Network share / local sync	Directory path
ical	iCalendar files (.ics)	File URL or path
ldap	LDAP / Active Directory	LDAP URL, Bind DN, Password
chrome_crawler	Intranet web pages (headless Chrome)	Optional login (username/password)

The system is openly extensible — new connector types can be added without UI changes.

---

Microsoft Graph / Teams / Exchange (msgraph)

Collects interaction signals from Microsoft 365 — Teams channels, calls, and Exchange calendar.

What is collected?

Signal	Meaning
teams_channel_coactivity	People active together in a Teams channel
teams_call_coparticipation	Participants of the same call
exchange_recurring_meeting	Organizer and attendees of recurring meetings
exchange_meeting_response_latency	Response behavior to meeting invitations

Setting up credentials

You need an App Registration in Azure Entra ID with the following API permissions (Application permissions, not delegated):

• Team.ReadBasic.All
• CallRecords.Read.All
• Calendars.Read
• User.Read.All

Steps:

1. Azure Portal → Entra ID → App Registrations → New registration
2. Certificates & Secrets → Create new client secret
3. API permissions → Add the above → Grant admin consent
4. Enter Azure Tenant ID, Client ID, and Client Secret in roleALPHA

Configuration

Parameter	Meaning	Default
Max age (days)	How far back to retrieve data	90
Min interactions	Minimum count to keep an edge	see Scope of analysis
Auto-map persons	Link email addresses directly to rA persons	no

Notes

• Only metadata is collected (no content, no audio/video data)
• Graph API throttling is automatically handled with exponential backoff

---

Atlassian Confluence (confluence)

Analyzes collaboration structure in Confluence wikis based on page edits and comments.

What is collected?

Signal	Meaning
confluence_coauthorship	Multiple authors editing the same page
confluence_mention	Comment authors connected to page authors
confluence_cross_space_link	Links between different spaces

Setting up credentials

1. Atlassian → My account → Security → Create API token
2. Enter Confluence URL (e.g. https://company.atlassian.net), email address, and token in roleALPHA

The user needs read access to all spaces to be analyzed.

Configuration

Parameter	Meaning	Default
Max age (days)	How far back version history is evaluated	90
Min interactions	Minimum number of shared edits	see Scope of analysis

---

SharePoint Online (sharepoint)

Evaluates file changes on SharePoint sites and detects collaborative document work.

What is collected?

Signal	Meaning
sharepoint_coauthorship	Creator and last editor of a file
sharepoint_site_coactivity	Users active on the same site

Setting up credentials

Same Azure App Registration as MS Graph (Client Credentials Flow), additional permission:

• Sites.Read.All

Find Site IDs via: https://company.sharepoint.com/sites/SITENAME/_api/site/id

Configuration

Parameter	Meaning	Default
Site IDs	Comma-separated list of sites to evaluate	—
Max age (days)	File changes within this period	90
Min interactions	Minimum number of shared edits	see Scope of analysis

Notes

• sharepoint_site_coactivity requires SharePoint Analytics, not included in all licenses
• Without site analytics, only sharepoint_coauthorship signals are generated

---

Jira / Jira Service Management (jira)

Detects collaboration from issue assignments, updates, and comments.

What is collected?

Signal	Meaning
jira_co_assignee	Reporter and assignee of the same ticket
jira_cocomment	All pairs of people who commented on the same ticket

Setting up credentials

1. Atlassian → My account → Security → Create API token
2. Enter Jira URL (e.g. https://company.atlassian.net), email address, and token in roleALPHA

Configuration

Parameter	Meaning	Default
Project keys	Restrict to specific projects (empty = all)	all
Max age (days)	Tickets within this period (by last update)	90
Min interactions	Minimum shared activities	see Scope of analysis

---

CSV Import (csv_import)

The fastest entry point without API access — any interaction data from a CSV file.

What is collected?

Any interactions: the signal type is freely configurable (default: csv_interaction).

Format

fromPerson,toPerson,count
alice@company.com,bob@company.com,12
carol@company.com,alice@company.com,7

• fromPerson / toPerson: External IDs (email, name, any identifier)
• count: Optional — interaction count (default: 1 per row)
• Column names are freely configurable

Configuration

Parameter	Meaning	Default
From column	Column name for the source actor	fromPerson
To column	Column name for the target actor	toPerson
Count column	Column name for interaction count (optional)	—
Signal type	Label for this signal source	csv_interaction

Notes

• CSV content can be pasted directly into the text field or given as a server file path
• Identical pairs across multiple rows are automatically summed
• Ideal for initial data import or one-time analyses

---

Filesystem / Network Share (filesystem)

Reads Office documents from a local or network path and extracts co-authors from metadata.

What is collected?

Signal	Meaning
filesystem_coauthorship	Creator and last editor of a document

Supported formats

.docx, .xlsx, .pptx (OOXML format — metadata from core.xml)

Configuration

Parameter	Meaning	Default
Path	Directory scanned recursively	—
File extensions	Filtered formats	.docx, .xlsx, .pptx
Max age (days)	Files by modification date	90
Min interactions	Minimum co-authorships	see Scope of analysis

Notes

• The server must have read access to the specified path
• Only metadata is evaluated (no document content)
• Creator and last editor must be different people — pure self-edits are ignored
• Network shares only work if the ra-signals service has the path mounted

---

iCalendar (.ics) (ical)

Evaluates calendar files in iCalendar format and identifies shared meeting participants.

What is collected?

Signal	Meaning
ical_coapplicant	All attendees and organizers of the same event

Configuration

Parameter	Meaning	Default
Calendar URL	HTTP(S) URL to the .ics file	—
Calendar path	Alternative server file path	—
Max age (days)	Events within this period	90
Min interactions	Minimum shared events	see Scope of analysis

Notes

• Either URL or path must be provided
• Supports public calendar URLs (e.g. shared Outlook calendars) and local .ics exports
• Creates connections between all participants of all captured events

---

LDAP / Active Directory (ldap)

Derives collaboration signals from group memberships — who is in which groups?

What is collected?

Signal	Meaning
ldap_group_coactivity	All members of the same LDAP group

Setting up credentials

A read-only bind user is sufficient. Required:

• ldap:// or ldaps:// URL (e.g. ldap://dc.company.com:389)
• Bind DN (e.g. cn=readonly,dc=company,dc=com)
• Bind password

Configuration

Parameter	Meaning	Default
Search base	Base DN for group search	—
Group filter	LDAP filter for groups	(objectClass=group)
Max age (days)	(used for threshold)	90
Min interactions	Minimum shared groups	see Scope of analysis

Notes

• Supports member, uniqueMember, and memberUid attributes
• Connection is closed immediately after retrieval
• For very large directories (>10,000 groups) a narrower search base is recommended

---

Chrome Web Crawler (chrome_crawler)

Crawls internal web pages (intranet, wikis, web apps) via headless Chrome and extracts mention and co-authorship signals.

Special note: The crawler runs as a CLI tool on the client machine — not as a server process. Data is sent to roleALPHA via API after crawling. This source entry documents the start URLs and configuration as a reference.

What is collected?

Signal	Meaning
web_cocitation	Two people mentioned on the same page
web_person_entity	Person and entity (role, value stream, OKR …) on the same page
web_entity_cooccurrence	Two entities mentioned together
web_coauthorship	Detected co-authors of a page
web_cocomment	Comments from multiple people on a page

Prerequisites

• Google Chrome on the executing machine (auto-detected)
• Node.js + tsx to run the CLI
• Network access to start URLs
• No root privileges required

CLI usage

tsx services/ra-signals/chrome-crawler/index.ts \
  --url https://wiki.internal/home \
  --pattern "https://wiki\\.internal/.*" \
  --depth 3 \
  --max-pages 100 \
  --endpoint https://ra.example.com \
  --token <jwt-token> \
  --tenant <tenant-uuid>

All parameters

Parameter	Meaning	Default
--url	Start URL(s) — can be specified multiple times	—
--pattern	Regex to scope the crawl area	Prefix of first URL
--depth	Max crawl depth	3
--max-pages	Safety limit on pages	100
--endpoint	roleALPHA URL	—
--token	JWT token (from Settings → Personal → Profile)	—
--tenant	Tenant UUID	—
--login-url	Login form URL (optional)	—
--login-user	Username for login	—
--login-pass	Password for login	—
--author-selector	CSS selector for author elements	—
--chrome-path	Path to Chrome binary	auto-detected

Live progress

[  1/100] https://wiki.internal/home — 3 entities found (Alice Müller, Value stream X, OKR Y)
[  2/100] https://wiki.internal/team-a — 5 entities found
...
✔ Ingest successful: 234 edges, 18 actors

Source traceability

Every discovered connection carries the concrete source URLs. These appear in the Org Intelligence Reports gap analysis as clickable links — making every insight fully traceable.

---

Adding a Source

1. Click New Source.
2. Select the collector type from the dropdown.
3. Fill in the credentials and configuration (the form adapts dynamically to the type).
4. Enable Automatic collection for hourly scheduled runs.
5. Save.

Troubleshooting: collector type list is empty

If the collector type dropdown stays empty (and existing sources are not shown either), the Org Intelligence service (ra-signals) is not switched on, unreachable or not yet rolled out for this tenant.

In that case the dialog shows an explicit notice instead of an empty selection. Resolution:

1. The Org Intelligence data node (ra-signals) must be switched on under Settings › Data Nodes (Data Nodes) — the tenant admin does that themselves.
2. By the platform admin: the ra-signals service must be running on the tenant’s deployment target. If the service was added later, use the Re-Sync button on the deployment target in Platform Admin — it triggers a fresh rollout (the service starts, routes are set up, and it registers in the service registry).

Troubleshooting: a source reports an error and “Last collected” stays empty

If an enabled source shows an error and the “Last collected” timestamp stays empty, the collect run started but failed while writing. The error is shown on the source; the analyses (org gaps, signal reports) stay empty because no edges arrive.

Important to understand: an empty analysis here does not mean “no signals”, it means “nothing ingested”. So always check the error on the source first, before concluding from an empty report that there is no collaboration.

The service repairs the most common cause itself: if the uniqueness index that ingestion relies on is missing, ra-signals creates it on the next run (removing duplicate edges from the existing data first). If the error persists after that, the database role most likely lacks the permission to create an index — then the platform admin can help.

Credentials

All sensitive fields (passwords, API tokens, client secrets) are stored AES-256-GCM encrypted and never returned in the API.

Manual trigger & connection test

• Heart icon: Tests the connection to the external API without collecting data.
• Play icon: Starts an immediate collect run (runs in the background).

Configuration parameters (global)

Parameter	Meaning	Default
Max age (days)	How far back interactions are collected	90
Min interactions	Threshold: edges with fewer interactions are discarded	see Scope of analysis

Automatic Collection Schedule

Enabled sources are collected hourly automatically. The “Last collected” timestamp shows the last successful run.

Related

• Org Intelligence — Feature overview
• Person Mapping — Link collected actors to persons
• Org Intelligence Reports — Gap analysis and reports