How can ellmer integrate DeepSeek provided by other manufacturers?

[trafficfan]

DeepseekR1 is open-source and available from many manufacturers, but the price of the tokens varies. I found a cheaper option, but there were problems with the integration.
The provider‘s web here
百度云文档
here's the python code

from openai import OpenAI

client = OpenAI(
   api_key="bce-v3/ALTAK-KZke********/f1d6ee*************",  # bearer token
   base_url="https://qianfan.baidubce.com/v2"
)

completion = client.chat.completions.create(
    model="deepseek-r1",   
    messages=[{'role': 'system', 'content': 'You are a helpful assistant.'},
              {'role': 'user', 'content': '你是谁？'}]
)

print(completion.choices[0].message)

I've tried ellmer, maybe it should be like

library(ellmer)
chat <- chat_openai(
  system_prompt = NULL,
  turns = NULL,
  base_url = "https://qianfan.baidubce.com/v2/chat/completions",
  api_key = "bce-v3/ALTAK-KZke********/f1d6ee*************",
  model = "deepseek-r1",
  seed = NULL,
  api_args = list(),
  echo = c("none", "text", "all")
)
chat$chat("
  What is the difference between a tibble and a data frame?
  Answer with a bulleted list
")

chat$chat("Tell me three funny jokes about statistcians")

however it returns

Error in `httr2::req_perform_connection()`:
! HTTP 404 Not Found.
Run `rlang::last_trace()` to see where the error occurred.

I want to know if it's my code that's wrong or if Ellmer doesn't support it temporarily? I think I used the correct API key.

[hadley]

ellmer automatically adds the /chat/completions so you might try base_url = "https://qianfan.baidubce.com/v2"

[hadley]

I'll also add base_url to chat_deepseek() since there are a few customisations needed.

[trafficfan]

This didn't solve the problem, and the current error message is

Error in `httr2::req_perform_connection()`:
! HTTP 400 Bad Request.
• Message index[1] content should be string
Run `rlang::last_trace()` to see where the error occurred.

The python code can work however

PS C:\Users\admin> & C:/Users/admin/AppData/Local/Programs/Python/Python312/python.exe "c:/Users/admin/Desktop/from openai import OpenAI.py"
ChatCompletionMessage(content='<think>\n\n</think>\n\n您好！我是由中国的深度求索（DeepSeek）公司开发的智能助手DeepSeek-R1。如您有任何任何问题，我会尽我所能为您提供帮助。', refusal=None, role='assistant', audio=None, function_call=None, tool_calls=None)

ellmer automatically adds the /chat/completions so you might try base_url = "https://qianfan.baidubce.com/v2"

[trafficfan]

This R code can work however

library(httr2)
api_key <- "***********"
base_url <- "https://qianfan.baidubce.com/v2/chat/completions"
request_body <- list(
  model = "deepseek-r1",
  messages = list(
    list(role = "system", content = "You are a helpful assistant."),
    list(role = "user", content = "你是谁？")
  )
)
response <- request(base_url) |>
  req_headers(
    `Authorization` = paste("Bearer", api_key),
    `Content-Type` = "application/json"
  ) |>
  req_body_json(request_body) |>
  req_method("POST") |>
  req_perform()
response_content <- response |> resp_body_json()
print(response_content$choices[[1]]$message$content)

Maybe the message I provided could help the development : )

[hadley]

Yes, that's why chat_deepseek() needs a base_url parameter.