tidyverse · s-spavound · Mar 7, 2025 · Mar 7, 2025 · Mar 7, 2025 · hadley
diff --git a/DESCRIPTION b/DESCRIPTION
@@ -84,6 +84,7 @@ Collate:
     'provider-gemini.R'
     'provider-github.R'
     'provider-groq.R'
+    'provider-huggingface.r'
     'provider-ollama.R'
     'provider-openrouter.R'
     'provider-perplexity.R'

diff --git a/NAMESPACE b/NAMESPACE
@@ -25,6 +25,7 @@ export(chat_deepseek)
 export(chat_gemini)
 export(chat_github)
 export(chat_groq)
+export(chat_hf)
 export(chat_ollama)
 export(chat_openai)
 export(chat_openrouter)

diff --git a/R/provider-huggingface.r b/R/provider-huggingface.r
@@ -0,0 +1,58 @@
+#' Chat with a model hosted on Hugging Face Serverless Inference API
+#'
+#' @description
+#' [Hugging Face](https://huggingface.co/) hosts a variety of open-source
+#' and proprietary AI models available via their Inference API.
+#' To use the Hugging Face API, you must have an Access Token, which you can obtain
+#' from your [Hugging Face account](https://huggingface.co/settings/tokens).
+#'
+#' This function is a lightweight wrapper around [chat_openai()], with
+#' the defaults adjusted for Hugging Face. Model defaults to `meta-llama/Llama-3.1-8B-Instruct`.
+#'
+#' ## Known limitations
+#'
+#' * Some models do not support the chat interface or parts of it, for example
+#' `google/gemma-2-2b-it` does not support a system prompt. You will need to
+#'  carefully choose the model.
-#' `google/gemma-2-2b-it` does not support a system prompt. You will need to
-#'  carefully choose the model.
+#'   `google/gemma-2-2b-it` does not support a system prompt. You will need to
+#'    carefully choose the model.
-#' `google/gemma-2-2b-it` does not support a system prompt. You will need to
-#'  carefully choose the model.
+#'   `google/gemma-2-2b-it` does not support a system prompt. You will need to
+#'    carefully choose the model.
+#'
+#' @family chatbots
+#' @param api_key The API key to use for authentication. You should not
+#'   supply this directly; instead, store your Hugging Face API key as an
+#'   environment variable (`HUGGINGFACE_API_KEY`) in your `.Renviron` file.
+#'   Use `usethis::edit_r_environ()` to modify it.
+#' @export
+#' @inheritParams chat_openai
+#' @inherit chat_openai return
+#' @examples
+#' \dontrun{
+#' chat <- chat_hf()
+#' chat$chat("Tell me three jokes about statisticians")
+#' }
+chat_hf <- function(system_prompt = NULL,
+                    turns = NULL,
+                    base_url = "https://api-inference.huggingface.co/models/",
+                    api_key = hf_key(),
+                    model = NULL,
+                    seed = NULL,
+                    api_args = list(),
+                    echo = NULL) {
+
+  model <- set_default(model, "meta-llama/Llama-3.1-8B-Instruct")
+  echo <- check_echo(echo)
+
+  chat_openai(
+    system_prompt = system_prompt,
+    turns = turns,
+    # modify base_url for hugging face compatibility with openai
+    base_url = paste0(base_url, model, "/v1"),
+    api_key = api_key,
+    model = model,
+    seed = seed,
+    api_args = api_args,
+    echo = echo
+  )
+}
+
+hf_key <- function() {
+  key_get("HUGGINGFACE_API_KEY")
+}
diff --git a/R/provider-openai.R b/R/provider-openai.R
@@ -50,7 +50,7 @@ NULL
 #'   Answer with a bulleted list
 #' ")
 #'
-#' chat$chat("Tell me three funny jokes about statistcians")
+#' chat$chat("Tell me three funny jokes about statisticians")
 chat_openai <- function(system_prompt = NULL,
                             turns = NULL,
                             base_url = "https://api.openai.com/v1",

diff --git a/ellmer.Rproj b/ellmer.Rproj
@@ -1,4 +1,5 @@
 Version: 1.0
+ProjectId: cd6d3e25-62f1-4041-a95a-6833286990e5
 
 RestoreWorkspace: No
 SaveWorkspace: No

diff --git a/man/chat_bedrock.Rd b/man/chat_bedrock.Rd
diff --git a/man/chat_claude.Rd b/man/chat_claude.Rd
diff --git a/man/chat_cortex_analyst.Rd b/man/chat_cortex_analyst.Rd
diff --git a/man/chat_databricks.Rd b/man/chat_databricks.Rd
diff --git a/man/chat_deepseek.Rd b/man/chat_deepseek.Rd
diff --git a/man/chat_gemini.Rd b/man/chat_gemini.Rd
diff --git a/man/chat_github.Rd b/man/chat_github.Rd
diff --git a/man/chat_groq.Rd b/man/chat_groq.Rd
diff --git a/man/chat_hf.Rd b/man/chat_hf.Rd
diff --git a/man/chat_ollama.Rd b/man/chat_ollama.Rd
diff --git a/man/chat_openai.Rd b/man/chat_openai.Rd
diff --git a/man/chat_openrouter.Rd b/man/chat_openrouter.Rd
diff --git a/man/chat_perplexity.Rd b/man/chat_perplexity.Rd
diff --git a/tests/testthat/test-provider-huggingface.R b/tests/testthat/test-provider-huggingface.R
@@ -0,0 +1,6 @@
+test_that("can make simple request", {
+  chat <- chat_hf("Be as terse as possible; no punctuation", model = "meta-llama/Llama-3.1-8B-Instruct")
+  resp <- chat$chat("What is 1 + 1?", echo = FALSE)
+  expect_match(resp, "2")
+  expect_equal(chat$last_turn()@tokens > 0, c(TRUE, TRUE))
+})