GeneralBots/botserver
2
1
Fork 0
Code Activity Actions 4

fix: kimi max_tokens 131072, add stream traces, content chars tracking
All checks were successful
BotServer CI/CD / build (push) Successful in 4m12s
Details

Browse source
This commit is contained in:
Rodrigo Rodriguez (Pragmatismo) 2026-04-13 19:42:31 -03:00
parent 4d9d38ffda
commit 79997fc3b3
1 changed files with 17 additions and 11 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

28
src/llm/kimi.rs
View file

@ -157,7 +157,7 @@ impl LLMProvider for KimiClient {
model: model_name.to_string(),
messages,
stream: Some(false),
max_tokens: None,
max_tokens: Some(131072),
temperature: Some(1.0),
top_p: Some(1.0),
tools: None,
@ -246,7 +246,7 @@ impl LLMProvider for KimiClient {
model: model_name.to_string(),
messages,
stream: Some(true),
max_tokens: None,
max_tokens: Some(131072),
temperature: Some(1.0),
top_p: Some(1.0),
tools: tools.cloned(),
@ -257,7 +257,7 @@ impl LLMProvider for KimiClient {
};
let url = self.build_url();
info!("Kimi streaming request to: {}", url);
info!("[Kimi] Streaming request to: {} model={} max_tokens=131072", url, model_name);
let response = self
.client
@ -271,13 +271,17 @@ impl LLMProvider for KimiClient {
if !response.status().is_success() {
let error_text = response.text().await.unwrap_or_default();
error!("Kimi streaming error: {}", error_text);
error!("[Kimi] Streaming error: {}", error_text);
return Err(format!("Kimi streaming error: {}", error_text).into());
}
let mut stream = response.bytes_stream();
info!("[Kimi] Connection established, starting stream");
let mut stream = response.bytes_stream();
let mut total_content_chars: usize = 0;
let mut chunk_count: usize = 0;
let mut buffer = Vec::new();
while let Some(chunk_result) = stream.next().await {
let chunk = chunk_result.map_err(|e| format!("Stream error: {}", e))?;
@ -292,7 +296,8 @@ impl LLMProvider for KimiClient {
}
if line == "data: [DONE]" {
std::mem::drop(tx.send(String::new()));
info!("[Kimi] Stream done: {} chunks, {} content chars sent", chunk_count, total_content_chars);
let _ = tx.send(String::new()).await;
return Ok(());
}
@ -302,6 +307,8 @@ impl LLMProvider for KimiClient {
if let Some(choices) = chunk_data.get("choices").and_then(|c| c.as_array()) {
for choice in choices {
if let Some(delta) = choice.get("delta") {
chunk_count += 1;
if let Some(tool_calls) = delta.get("tool_calls").and_then(|t| t.as_array()) {
for tool_call in tool_calls {
let tool_call_json = serde_json::json!({
@ -312,11 +319,10 @@ impl LLMProvider for KimiClient {
}
}
// Kimi K2.5 sends thinking via reasoning_content
// The actual user-facing response is in content field
// We ONLY send content — never reasoning_content (internal thinking)
// Kimi K2.5: content has the answer, reasoning/reasoning_content is thinking
if let Some(text) = delta.get("content").and_then(|c| c.as_str()) {
if !text.is_empty() {
total_content_chars += text.len();
let _ = tx.send(text.to_string()).await;
}
}
@ -324,8 +330,8 @@ impl LLMProvider for KimiClient {
if let Some(reason) = choice.get("finish_reason").and_then(|r| r.as_str()) {
if !reason.is_empty() {
info!("Kimi stream finished: {}", reason);
std::mem::drop(tx.send(String::new()));
info!("[Kimi] Stream finished: {}, {} content chars", reason, total_content_chars);
let _ = tx.send(String::new()).await;
return Ok(());
}
}