Multiple related tables: register additional tables with $add_table() (or every table from a DBI connection at once with $add_tables()), and the LLM can reason across them — joins, cross-table filters, aggregations. Per-table reactive state ($df(), $sql(), $title()) is available via qc_vals$table("name") on the list returned by $server(). (#195)
qc <- QueryChat$new(orders_df, "orders")
qc$add_table(customers_df, "customers")
qc_vals <- qc$server()
qc_vals$table("orders")$df()
data_dict: annotate tables and columns with plain-English descriptions from a YAML file (following the data-dict spec). This is now the preferred way to give the LLM context about your data, especially with multiple tables — no extra prompting required. (#195)
Chat-first apps: querychat_app() (and QueryChat$app()) now put the chat front and center — the SQL editor and data table live in a drawer that opens automatically when the LLM runs a query. The new $page() method brings the same full-window layout to your own apps:
qc <- QueryChat$new(penguins)
ui <- qc$page("Penguins Explorer")
/handoff slash command: turn selected query and visualization results from a chat session into a downloadable Quarto dashboard or Shiny app, with AI-assisted revision and bundled data.
Persistent conversation history: chats now survive page reloads and browser sessions by default. For a shareable URL or full Shiny bookmark instead, pass history = shinychat::history_options(restore_mode = "url") or "bookmark". Disable with history = FALSE.
File attachments are now enabled by default: users can attach images, PDFs, and text files to their messages. Disable with allow_attachments = FALSE. (#253)
Editable SQL panel: the SQL panel in querychat_app() is now a code editor — tweak the generated SQL and apply it with Ctrl/Cmd+Enter. (#265)
PinSource: chat with datasets pinned to a pins board (parquet, CSV, JSON, RDS); the pin's title, description, and tags serve as the default data description. Multiple pins (and pins mixed with data frames) work in one chat via a shared DuckDB connection. (#246, #312)
Deferred construction is more flexible: table_name is now optional in QueryChat$new(NULL), and $server() gains a table_name parameter so the table can be named per session. (#305)
The $data_source property has been removed. Use qc$table("name")$data_source to read a table's data source, and qc$add_table(df, "name", replace = TRUE) to replace it. (#195)
$app()/$app_obj()'s bookmark_store parameter has been removed; pass history = shinychat::history_options(restore_mode = "bookmark") for equivalent behavior (existing $app() callers get this default automatically). Note the storage mechanism changed: server-side bookmark storage is now required rather than encoding state in the URL, so deployments that relied on $app() being fully stateless should pass history = FALSE.
querychat no longer disconnects a database connection you gave it. $cleanup() now closes only what querychat itself created: DBISource$cleanup() and TblSqlSource$cleanup() no longer disconnect a caller-supplied connection, and neither does the automatic $cleanup() that runs when a Shiny session/app stops (cleanup = NA, the default). If you relied on that, disconnect yourself on shutdown, e.g. shiny::onStop(function() DBI::dbDisconnect(con)). Connections querychat creates for you (for data frames and pins) are still closed automatically.
$server()'s enable_bookmarking parameter is deprecated in favor of history.The "visualize" tool is now included in the default toolset. If the suggested ggsql package is not installed, the tool is dropped with a warning instead of erroring.
Chat greetings render instantly when provided, and generated greetings are now schema-aware — they describe the data at hand — and are preserved across bookmark/restore. Opt additional tables into the greeting with include_in_greeting = TRUE on $add_table()/$add_tables(). (#249, #261)
The system prompt no longer embeds the full schema upfront; the LLM fetches per-table schema on demand, skipping columns already described by a data_dict. (#195)
Special characters like <, >, and & in data_description and extra_instructions are no longer HTML-escaped in the system prompt. (#258)
The close button in $app() is now hidden in deployed (non-interactive) contexts, where stopApp() would crash the session for other users. (#259)
Query results are now expanded in the chat only when the user asks to see the raw table, instead of far more often than intended. (#295)
$server(data_source = ) now registers the table for that session only, instead of modifying the shared QueryChat instance. Concurrent sessions (e.g. per-user database connections on Posit Connect) no longer clobber each other's data source or greeting, and any connection querychat created for the session is cleaned up when the session ends. (#300, #306)
Added a new "visualize" tool that lets querychat render interactive charts inline in the chat. When enabled (via tools = c("filter", "query", "visualize")), the LLM can answer questions with charts by writing ggsql (SQL with a VISUALISE clause) instead of only tables. Charts can be expanded to fullscreen and their underlying query inspected. Requires the ggsql package and bslib >= 0.11.0. (#224)
Added stream cancellation support. A stop button now appears during LLM streaming, allowing users to cancel in-progress responses by clicking it or pressing Escape. Cancellation is enabled by default and can be disabled via enable_cancel = FALSE in the UI. (#241)
Added support for Snowflake Semantic Views. When connected to Snowflake via DBI, querychat automatically discovers available Semantic Views and includes their definitions in the system prompt. This helps the LLM generate correct queries using the SEMANTIC_VIEW() table function with certified business metrics and dimensions. (#200)
QueryChat$new() now supports deferred data source. Pass data_source = NULL at initialization time, then provide the actual data source via the data_source parameter of $server() or by setting the $data_source property. This enables use cases where the data source depends on session-specific authentication or per-user database connections. (#202)
QueryChat$server() now accepts a client parameter for session-scoped chat client overrides. This enables Posit Connect managed OAuth workflows where API credentials are only available inside the Shiny server function. The client spec is stored lazily at construction time and resolved only when needed, so QueryChat$new(NULL, "table") no longer requires an API key. (#205)
The query tool result card now starts collapsed by default. Users can still expand it to see the SQL query and results. Set QUERYCHAT_TOOL_DETAILS=expanded (or options(querychat.tool_details = "expanded")) to restore the previous behavior. (#239)
Query suggestions generated by the LLM now render reliably as clickable cards in the chat. (#236, #238)
The tools parameter now uses "filter" as the preferred name (instead of "update") for the dashboard-filtering tool group. The default is now c("filter", "query"). The legacy name "update" is still accepted everywhere. (#222)
When a custom prompt_template is provided that doesn't contain Mustache references to {{schema}}, the expensive get_schema() call is now skipped entirely. This allows users with large databases to avoid slow startup by providing their own prompt that includes schema information inline (or omits it). (#208)
DBISource now uses database-agnostic SQL for column and type detection, replacing LIMIT syntax with WHERE 1=0 and dbFetch(n=1). This fixes compatibility with SQL Server and other databases that don't support LIMIT. (#112, #197)The update tool now requires that the SQL query returns all columns from the original data source, ensuring that the dashboard can display the complete data frame after filtering or sorting. If the query does not return all columns, an informative error message will be provided. (#180)
Obvious SQL keywords that lead to data modification (e.g., INSERT, UPDATE, DELETE, DROP, etc.) are now prohibited in queries run via the query tool or update tool, to prevent accidental data changes. If such keywords are detected, an informative error message will be provided. (#180)
querychat() and QueryChat$new() now use either {duckdb} or {SQLite} for the in-memory database backend for data frames, depending on which package is installed. If both are installed, {duckdb} will be preferred. You can explicitly choose the engine in DataFrameSource$new() or set querychat.DataFrameSource.engine option to choose a global default. (#178)
QueryChat$sidebar(), QueryChat$ui(), and QueryChat$server() now support an optional id parameter to enable use within Shiny modules. When used in a module UI function, pass id = ns("your_id") where ns is the namespacing function from shiny::NS(). In the corresponding module server function, pass the unwrapped ID to QueryChat$server(id = "your_id"). This enables multiple independent QueryChat instances from the same QueryChat object. (#172)
QueryChat$client() can now create standalone querychat-enabled chat clients with configurable tools and callbacks, enabling use outside of Shiny applications. (#168)
QueryChat$console() was added to launch interactive console-based chat sessions with your data source, with persistent conversation state across invocations. (#168)
The tools used in a QueryChat chatbot are now configurable. Use the new tools parameter of querychat() or QueryChat$new() to select either or both "query" or "update" tools. Choose tools = "update" if you only want QueryChat to be able to update the dashboard (useful when you want to be 100% certain that the LLM will not see any raw data). (#168)
querychat_app() will now only automatically clean up the data source if QueryChat creates the data source internally from a data frame. (#164)
Breaking change: The $sql() method now returns NULL instead of "" (empty string) when no query has been set, aligning with the behavior of $title() for consistency. Most code using isTruthy() or similar falsy checks will continue working without changes. Code that explicitly checks sql() == "" should be updated to use falsy checks (e.g., !isTruthy(sql())) or explicit null checks (is.null(sql())). (#146)
Tool detail cards can now be expanded or collapsed by default when querychat runs a query or updates the dashboard via the querychat.tool_details R option or the QUERYCHAT_TOOL_DETAILS environment variable. Valid values are "expanded", "collapsed", or "default". (#137)
Added bookmarking support to QueryChat$server() and querychat_app(). When bookmarking is enabled (via bookmark_store = "url" or "server" in querychat_app() or $app_obj(), or via enable_bookmarking = TRUE in $server()), the chat state (including current query, title, and chat history) will be saved and restored with Shiny bookmarks. (#107)
Nearly the entire functional API (i.e., querychat_init(), querychat_sidebar(), querychat_server(), etc) has been hard deprecated in favor of a simpler OOP-based API. Namely, the new QueryChat$new() class is now the main entry point (instead of querychat_init()) and has methods to replace old functions (e.g., $sidebar(), $server(), etc). (#109)
querychat_data_source() was renamed to as_querychat_data_source(), and remains exported for a developer extension point, but users no longer have to explicitly create a data source. (#109)Added prompt_template support for querychat_system_prompt(). (Thank you, @oacar! #37, #45)
querychat_init() now accepts a client, replacing the previous create_chat_func argument. (#60)
The client can be:
ellmer::Chat object,ellmer::Chat object,"openai/gpt-4.1", to be passed to ellmer::chat().If client is not provided, querychat will use
querychat.client R option, which can be any of the above options,QUERYCHAT_CLIENT environment variable, which should be a provider-model string,ellmer::chat_openai().querychat_server() now uses a shiny::ExtendedTask for streaming the chat response, which allows the dashboard to update and remain responsive while the chat response is streaming in. (#63)
querychat now requires ellmer version 0.3.0 or later and uses rich tool cards for dashboard updates and database queries. (#65)
New querychat_app() function lets you quickly launch a Shiny app with a querychat chat interface. (#66)
querychat_ui() now adds a .querychat class to the chat container and querychat_sidebar() adds a .querychat-sidebar class to the sidebar, allowing for easier customization via CSS. (#68)
querychat now uses a separate tool to reset the dashboard. (#80)
querychat_greeting() can be used to generate a greeting message for your querychat bot. (#87)
querychat's system prompt and tool descriptions were rewritten for clarity and future extensibility. (#90)