refer to R sessions of interest as "hosts"

DESCRIPTION

@@ -21,7 +21,7 @@ Suggests:
 Config/testthat/edition: 3
 Encoding: UTF-8
 Roxygen: list(markdown = TRUE)
-RoxygenNote: 7.3.2.9000
+RoxygenNote: 7.3.2
 Imports: 
     btw (>= 0.0.1.9000),
     cli,

NAMESPACE

@@ -1,5 +1,5 @@
 # Generated by roxygen2: do not edit by hand
 
-export(mcp_proxy)
-export(mcp_serve)
+export(mcp_host)
+export(mcp_server)
 import(rlang)

R/host.R

@@ -0,0 +1,143 @@
+#' Model context protocol for your R session
+#'
+#' @description
+#' Together, these functions implement a model context protocol server for your
+#' R session.
+#'
+#' @section Configuration:
+#'
+#' [mcp_server()] should be configured with the MCP clients via the `Rscript`
+#' command. For example, to use with Claude Desktop, paste the following in your
+#' Claude Desktop configuration (on macOS, at
+#' `file.edit("~/Library/Application Support/Claude/claude_desktop_config.json")`):
+#'
+#' ```json
+#' {
+#'   "mcpServers": {
+#'     "r-acquaint": {
+#'       "command": "Rscript",
+#'       "args": ["-e", "acquaint::mcp_server()"]
+#'     }
+#'   }
+#' }
+#' ```
+#'
+#' Or, to use with Claude Code, you might type in a terminal:
+#'
+#' ```bash
+#' claude mcp add -s "user" r-acquaint Rscript -e "acquaint::mcp_server()"
+#' ```
+#'
+#' **mcp_server() is not intended for interactive use.**
+#'
+#' The server interfaces with the MCP client on behalf of the host in
+#' your R session. **Use [mcp_host()] to start the host in your R session.**
+#' Place a call to `acquaint::mcp_host()` in your `.Rprofile`, perhaps with
+#' `usethis::edit_r_profile()`, to start a host for every interactive R session
+#' you start.
+#'
+#' @examples
+#' if (interactive()) {
+#' mcp_host()
+#' }
+#'
+#' @name mcp
+#' @export
+mcp_host <- function() {
+  # HACK: If a host is already running in one session via `.Rprofile`,
+  # `mcp_host()` will be called again when the client runs the command
+  # Rscript -e "acquaint::mcp_server()" and the existing host will be wiped.
+  # Returning early in this case allows for the desired R session host to be
+  # running already before the client initiates the server.
+  if (!interactive()) {
+    return(invisible())
+  }
+
+  the$host_socket <- nanonext::socket("poly")
+  i <- 1L
+  suppressWarnings(
+    while (i < 1024L) {
+      # prevent indefinite loop
+      nanonext::listen(
+        the$host_socket,
+        url = sprintf("%s%d", acquaint_socket, i)
+      ) ||
+        break
+      i <- i + 1L
+    }
+  )
+
+  schedule_handle_message_from_server()
+}
+
+handle_message_from_server <- function(msg) {
+  pipe <- nanonext::pipe_id(the$raio)
+  schedule_handle_message_from_server()
+
+  # cat("RECV :", msg, "\n", sep = "", file = stderr())
+  if (!nzchar(msg)) {
+    return(nanonext::send_aio(the$host_socket, commandArgs(), pipe = pipe))
+  }
+  data <- jsonlite::parse_json(msg)
+
+  if (data$method == "tools/call") {
+    name <- data$params$name
+    fn <- getNamespace("btw")[[name]]
+    args <- data$params$arguments
+
+    # HACK for btw_tool_env_describe_environment. In the JSON, it will have
+    # `"items": []`, and that translates to an empty list, but we want NULL.
+    if (name == "btw_tool_env_describe_environment") {
+      if (identical(args$items, list())) {
+        args$items <- NULL
+      }
+    }
+
+    tool_call_result <- do.call(fn, args)
+    # cat(paste(capture.output(str(body)), collapse="\n"), file=stderr())
+
+    body <- jsonrpc_response(
+      data$id,
+      list(
+        content = list(
+          list(
+            type = "text",
+            text = paste(tool_call_result, collapse = "\n")
+          )
+        ),
+        isError = FALSE
+      )
+    )
+  } else {
+    body <- jsonrpc_response(
+      data$id,
+      error = list(code = -32601, message = "Method not found")
+    )
+  }
+  # cat("SEND:", to_json(body), "\n", sep = "", file = stderr())
+
+  nanonext::send_aio(
+    the$host_socket,
+    to_json(body),
+    mode = "raw",
+    pipe = pipe
+  )
+}
+
+schedule_handle_message_from_server <- function() {
+  the$raio <- nanonext::recv_aio(the$host_socket, mode = "string")
+  promises::as.promise(the$raio)$then(handle_message_from_server)$catch(
+    function(
+      e
+    ) {
+      print(e)
+    }
+  )
+}
+
+# Create a jsonrpc-structured response object.
+
+# Given a vector or list, drop all the NULL items in it
+drop_nulls <- function(x) {
+  x[!vapply(x, is.null, FUN.VALUE = logical(1))]
+}