2 releases
0.1.1 | Jul 31, 2024 |
---|---|
0.1.0 | Jul 31, 2024 |
#1735 in Web programming
24KB
515 lines
Unofficial api client for Deepseek
Sign up for an account on https://platform.deepseek.com/sign_in to get your API key.
[dependencies]
deepseek-api-client = { path = "https://github.com/acscoder/deepseek-api-client.git" }
Get started
Load your API key in env var or any secret way
use deepseek_api_client::*;
let api_key = std::env::var("DEEPSEEK_API_KEY").expect("$DEEPSEEK_API_KEY is not set");
Call synchronous function
- Get llm by function
chat_completion_sync
let mut llm_completion = chat_completion_sync(api_key) ;
- Then you will get a function
llm_completion
that take input vector of Message and get back Result of Response unwrap response result then take the first choice response text by functionget_response_text
let messages = vec![
Message {
role: "system".to_owned(),
content: "You are a helpful assistant".to_owned(),
},
Message {
role: "user".to_owned(),
content: "Write Hello world in rust".to_owned(),
},
];
let res = llm_completion(messages);
let res_text = get_response_text(&res.unwrap(), 0);
dbg!(res_text);
3 . Do the same with function code_completion_sync
for code generation with deepseek-coder
model and llm_function_call_sync
for function calling
Call asynchronous function
- Get llm by function
chat_completion
let mut llm_completion = chat_completion(api_key) ;
- It the same
chat_completion_sync
but it is async function that we can call with .await, i used tokio crate for async runtime
let rt = Runtime::new().unwrap();
- Input vector of Message and get back Result then take the first choice response text by function
get_response_text
let messages = vec![
Message {
role: "system".to_owned(),
content: "You are a helpful assistant".to_owned(),
},
Message {
role: "user".to_owned(),
content: "Write Hello world in rust".to_owned(),
},
];
let res = llm_completion(messages);
let r = rt.block_on(async { get_response_text(&res.await.unwrap(), 0) });
dbg!(&r);
- Do the same with function
code_completion
for code generation withdeepseek-coder
model andllm_function_call
for function calling
Call asynchronous function stream
- Get llm by
chat_completion_stream
let mut llm_completion = chat_completion_stream(api_key) ;
- We have async function that take input vector of Message and get back stream of Response
let rt = Runtime::new().unwrap();
let messages = vec![
Message {
role: "system".to_owned(),
content: "You are a helpful assistant".to_owned(),
},
Message {
role: "user".to_owned(),
content: "Write Hello world in rust".to_owned(),
},
];
let response_result = llm_completion(messages);
let _ = rt.block_on(async {
let res = response_result.await.unwrap();
let mut stream = res.bytes_stream();
while let Some(item) = stream.next().await {
let item = item.unwrap();
let s = match std::str::from_utf8(&item) {
Ok(v) => v,
Err(e) => panic!("Invalid UTF-8 sequence: {}", e),
};
let data = string_to_ChatResponses(s);
let text = get_response_text(&data, 0).unwrap_or("".to_owned());
println!("{}", text);
}
});
- Do the same with function
code_completion_stream
for code generation withdeepseek-coder
model andllm_function_call_stream
for function calling
Full Example code here https://github.com/acscoder/deepseek-api-example-code
Dependencies
~5–16MB
~199K SLoC