Callbacks and chat ownership | Schemas and tool results | Documents and uploaded files | Structured streams | Stream events and recovery | Request limits and observability | Tools followed by structured extraction | Async, parallel and batch boundaries

Choose a reading method | Discover the data source once | Read with SQL | Read one table without writing a query | Read Eventhouse data | Query a semantic model with DAX | Read a OneLake file | Read through GraphQL or Spark | Scale up with Arrow streams

Find and connect to a Lakehouse | List and read ordinary files | Write or upload a file | Discover and read managed tables | Write an R object as a managed table | Turn existing files into a table | Read a large or historical Delta table

Find and connect to a Warehouse | Prepare a staging Lakehouse for writes | Add rows to an existing table | Replace table data | Scale up with Arrow | More information

Choose a destination | Prepare a small R data frame | Write a Lakehouse table | Write an ordinary OneLake file | Write a Warehouse table | Write an Eventhouse table | Load a file that is already in a Lakehouse | Scale up with Arrow

The Fabric objects you will see | Sign in | Find a workspace and an item | Complete a first read | Choose the next guide

Find a KQL database | Write a small data frame in one call | Queue a tracked batch | Wait and inspect the outcome | Scale up with Arrow | Export a large KQL result to OneLake

Before the first call | Run one piece of Spark code | Choose the language that matches the code | Reuse a session for several statements | Submit a complete application file | Use an Environment for repeatable configuration

Query the model with DAX | Check access before refreshing | Refresh after an upstream update | Use enhanced refresh controls | Diagnose attempts and failures | Inspect history | More information

Run once and inspect history | Build schedule configurations | Create, list, update, and disable | Delete a schedule

Run a first query | Inspect the available fields | Read more than one page | Build an analysis-ready tibble | Treat incomplete pagination as partial data

Lifecycle | Prepare the function in Fabric | Make a first call | Send structured parameters | Retry only safe functions | Limits | More information

Quick start: sign in as yourself | 1. Find your tenant ID | 2. Set the tenant ID in R | 3. Test the connection | 4. Save the settings for future R sessions | Make sure Fabric access has been granted | First-login troubleshooting | How authentication works | Treat endpoint URLs as credential boundaries | Choosing an advanced authentication method | Advanced: authentication for automation | Service principal with a client secret | Managed identities and custom token providers | Set Fabric permissions by workload | Signing out (token cache)

Setup an LLM provider | Basic prompting | Prompt wraps | answer_as: Retrieving output in a specific format | JSON output | answer_by: Adding a reasoning mode to the LLM | answer_using: Have the LLM work with tools and code | Tools (function-calling) | Code generation and evaluation | Creating custom prompt wraps | Provider-level prompt wraps

Example

Breaking out of the evaluation loop | Extraction versus validation functions | Prompt wrap types and order of application | Configuring a LLM provider with a prompt wrap | Configuring a prompt wrap based on the LLM provider or HTTP responses | Configuring a prompt wrap based on other prompt wraps | Handler functions