AI: OpenAI-compatible provider now uses /v1/chat/completions (not the legacy /v1/completions) so Nvidia NIM and similar work; tolerate a trailing /v1 in the base URL (#96)

This commit is contained in:
Yuri Karamian
2026-06-11 22:36:03 +02:00
parent d9cb615f1f
commit fb4c5748a8
2 changed files with 12 additions and 166 deletions
+11 -165
View File
@@ -55,8 +55,8 @@ pub fn ai_request(
if url.is_empty() {
Err("No base URL configured for OpenAI Compatible provider".to_string())
} else {
let url = format!("{}/v1/completions", url.trim_end_matches('/'));
stream_completions(
let url = format!("{}/v1/chat/completions", normalize_openai_base(url));
stream_openai(
&app,
&url,
key_opt,
@@ -94,6 +94,13 @@ pub fn ai_request(
});
}
/// Normalize an OpenAI-compatible base URL: drop a trailing slash and a trailing `/v1`,
/// so both `https://host` and `https://host/v1` work (we append `/v1/chat/completions`).
fn normalize_openai_base(base: &str) -> String {
let b = base.trim().trim_end_matches('/');
b.strip_suffix("/v1").unwrap_or(b).trim_end_matches('/').to_string()
}
async fn stream_anthropic(
app: &AppHandle,
api_key: &str,
@@ -386,8 +393,8 @@ pub async fn test_connection(
if url.is_empty() {
return Err("No base URL configured for OpenAI Compatible provider".to_string());
}
let url = format!("{}/v1/completions", url.trim_end_matches('/'));
test_completions(&url, key_opt, model).await
let url = format!("{}/v1/chat/completions", normalize_openai_base(url));
test_openai(&url, key_opt, model).await
}
_ => test_anthropic(api_key, model).await,
}
@@ -426,167 +433,6 @@ async fn test_anthropic(api_key: &str, model: &str) -> Result<String, String> {
}
}
async fn stream_completions(
app: &AppHandle,
url: &str,
api_key: Option<&str>,
model: &str,
system_prompt: &str,
user_message: &str,
_request_id: &str,
) -> Result<(), String> {
let client = Client::new();
let prompt = format!("{}\n\nUser: {}\nAssistant:", system_prompt, user_message);
let body = json!({
"model": model,
"prompt": prompt,
"max_tokens": 4096,
"stream": true,
"temperature": 0.7
});
let mut req = client
.post(url)
.header("content-type", "application/json");
if let Some(key) = api_key {
req = req.header("Authorization", format!("Bearer {}", key));
}
let response = req
.json(&body)
.send()
.await
.map_err(|e| format!("Request failed: {}", e))?;
if !response.status().is_success() {
let status = response.status();
let body_text = response.text().await.unwrap_or_default();
return Err(format!("API error {}: {}", status, body_text));
}
// Parse SSE stream
use futures::StreamExt;
let mut stream = response.bytes_stream();
let mut buffer = String::new();
while let Some(chunk) = stream.next().await {
let chunk = chunk.map_err(|e| format!("Stream error: {}", e))?;
buffer.push_str(&String::from_utf8_lossy(&chunk));
// Process complete SSE events from buffer
while let Some(event_end) = buffer.find("\n\n") {
let event_str = buffer[..event_end].to_string();
buffer = buffer[event_end + 2..].to_string();
for line in event_str.lines() {
if let Some(data) = line.strip_prefix("data: ") {
if data == "[DONE]" {
let _ = app.emit(
"ai-stream",
AiStreamEvent {
event_type: "done".to_string(),
text: None,
error: None,
},
);
return Ok(());
}
if let Ok(parsed) = serde_json::from_str::<serde_json::Value>(data) {
// Completions streaming: choices[0].text
if let Some(text) = parsed["choices"][0]["text"].as_str() {
if !text.is_empty() {
let _ = app.emit(
"ai-stream",
AiStreamEvent {
event_type: "text".to_string(),
text: Some(text.to_string()),
error: None,
},
);
}
}
// Check finish_reason
if let Some(reason) = parsed["choices"][0]["finish_reason"].as_str() {
if reason == "stop" || reason == "length" {
let _ = app.emit(
"ai-stream",
AiStreamEvent {
event_type: "done".to_string(),
text: None,
error: None,
},
);
return Ok(());
}
}
// Check for error in stream
if let Some(err) = parsed["error"]["message"].as_str() {
let _ = app.emit(
"ai-stream",
AiStreamEvent {
event_type: "error".to_string(),
text: None,
error: Some(err.to_string()),
},
);
return Err(err.to_string());
}
}
}
}
}
}
let _ = app.emit(
"ai-stream",
AiStreamEvent {
event_type: "done".to_string(),
text: None,
error: None,
},
);
Ok(())
}
async fn test_completions(url: &str, api_key: Option<&str>, model: &str) -> Result<String, String> {
let client = Client::new();
let body = json!({
"model": model,
"prompt": "Hi",
"max_tokens": 10,
});
let mut req = client
.post(url)
.header("content-type", "application/json");
if let Some(key) = api_key {
req = req.header("Authorization", format!("Bearer {}", key));
}
let response = req
.json(&body)
.send()
.await
.map_err(|e| format!("Connection failed: {}", e))?;
if response.status().is_success() {
Ok("Connection successful".to_string())
} else {
let status = response.status();
let body_text = response.text().await.unwrap_or_default();
Err(format!("API error {}: {}", status, body_text))
}
}
async fn test_openai(url: &str, api_key: Option<&str>, model: &str) -> Result<String, String> {
let client = Client::new();