# Sources

The **Sources** section defines which external data is made available to the AI agent as contextual information.

Sources are **read-only** and are automatically added to the agent’s context during execution.\
They do **not replace input parameters**, but **augment** them with additional data the agent can reason over.

***

### Source types

<table><thead><tr><th width="188">Source type</th><th>Description</th></tr></thead><tbody><tr><td><strong>Dataclips</strong></td><td>Datasets created within the Any2Info platform.</td></tr><tr><td><strong>Dataflows</strong></td><td>Dataflows within the Any2Info platform, used as tools.</td></tr><tr><td><strong>Files</strong></td><td>Uploaded documents such as PDF, DOCX, TXT, or other supported formats.</td></tr><tr><td><strong>Vector Stores</strong></td><td>Indexed data stored in vector storage, optimized for semantic search and retrieval.</td></tr></tbody></table>

***

### Tool calling

**Tool calling** is the mechanism by which an AI agent can actively invoke an external function or system during a conversation or task — not just generate text.

#### Supported tools

* **Dataclips**
* **Dataflows**
* **Files (search)**
* **Web (search)**

#### Importance of descriptions

For correct tool usage, **clear descriptions are critical**.\
The AI relies heavily on descriptions to understand **when** and **how** to use tools and how to map inputs and outputs correctly.

This applies to:

* **Dataclips** – Each dataclip has a *Description* field explaining its content and purpose
* **Dataflows** – Each dataflow has a *Description* field explaining what it does
* **Event parameters** – Each parameter includes a *Description* field describing its meaning and usage

Well-written descriptions significantly improve the agent’s ability to:

* Select the correct tool
* Populate parameters correctly
* Execute actions reliably

***

### Dataclips

#### What is it?

A (live) dataset available within the Any2Info platform (ERP, CRM, SQL, API, etc.).

#### What does it provide to the agent?

It supplies the AI with **up-to-date, structured data** to reason with.

#### Examples

* Relationships (AccountID, name, address)
* Articles (ArticleID, price, stock)
* Projects, orders, employees

#### Role in relation to agents

Dataclips answer the question:\
\&#xNAN;**“What exists in my systems?”**

Agents use dataclips to:

* Perform fuzzy matching
* Find IDs
* Validate user choices
* Prevent hallucinated input

***

### Dataflows

#### What is it?

An action or operation within the Any2Info platform, such as:

* Create
* Update
* Sync
* Export
* Trigger

#### What does it provide to the agent?

Dataflows are **“the buttons the AI can press to make things happen.”**

#### Examples

* Create a Synergy document
* Create a Synergy request
* Create a Synergy request with a document
* Send an email
* Download a PDF

#### Role in relation to agents

Dataflows answer the question:\
\&#xNAN;**“What is the agent allowed to DO?”**

***

### Files

#### What is it?

Concrete documents or media passed into an AI call.

#### Examples

* PDFs
* Word documents
* Excel files
* Images
* Scans
* Transcripts

#### What does it provide to the agent?

Files are the **primary sources** that must be read, interpreted, or analyzed.

From files, the AI can:

* Extract text
* Read amounts and values
* Interpret rules and clauses
* Understand document context

#### Ideal use cases

* Quotations
* Invoices
* Inspection reports
* Contracts

#### Role in relation to agents

Files represent:\
\&#xNAN;**“Reality captured in documents.”**

***

### Vector stores

#### What is it?

An indexed (external) knowledge base built on stored data using embeddings.

#### Typical contents

* Manuals
* Contracts
* Product documentation
* Procedures
* Historical cases
* Knowledge articles

#### What does it provide to the agent?

Vector stores give the AI **semantic memory**.

This enables:

* Meaning-based search instead of exact keyword matching
* Retrieval Augmented Generation (RAG)
* Policy enforcement
* Explanations and guidance
* Consistent answers

#### Role in relation to agents

Vector stores answer the question:\
\&#xNAN;**“What does the organization know?”**