From 2d2775b26e645c15aa540b8130c5c9256c7a7563 Mon Sep 17 00:00:00 2001 From: rewine Date: Wed, 6 Aug 2025 15:57:37 +0800 Subject: [PATCH 1/2] feat: add max_tokens configuration for AI translation services 1. Added max_tokens field to all translator structs (OpenAI, Claude, Gemini, Grok, Qwen) 2. Integrated max_tokens configuration from config file with default fallback (2048) 3. Added max_tokens parameter to API requests for all translation services 4. Replaced hardcoded max_tokens value in Qwen implementation with configurable value This change allows better control over token usage and response length across all AI translation services. The configurable max_tokens parameter enables: - Cost optimization by limiting token consumption - Better response time management - Consistent behavior across different translation models - More flexible configuration based on use case requirements --- src/ai_service.rs | 39 +++++++++++++++++++++++++++++---------- 1 file changed, 29 insertions(+), 10 deletions(-) diff --git a/src/ai_service.rs b/src/ai_service.rs index 1baf854..d340939 100644 --- a/src/ai_service.rs +++ b/src/ai_service.rs @@ -37,6 +37,7 @@ pub struct OpenAITranslator { endpoint: String, model: String, timeout_seconds: u64, + max_tokens: u64, } pub struct ClaudeTranslator { @@ -44,6 +45,7 @@ pub struct ClaudeTranslator { endpoint: String, model: String, timeout_seconds: u64, + max_tokens: u64, } pub struct CopilotTranslator { @@ -56,6 +58,7 @@ pub struct GeminiTranslator { endpoint: String, model: String, timeout_seconds: u64, + max_tokens: u64, } pub struct GrokTranslator { @@ -63,6 +66,7 @@ pub struct GrokTranslator { endpoint: String, model: String, timeout_seconds: u64, + max_tokens: u64, } pub struct QwenTranslator { @@ -70,6 +74,7 @@ pub struct QwenTranslator { endpoint: String, model: String, timeout_seconds: u64, + max_tokens: u64, } impl DeepSeekTranslator { @@ -101,6 +106,9 @@ impl OpenAITranslator { timeout_seconds: crate::config::Config::load() .map(|c| c.timeout_seconds) .unwrap_or(20), + max_tokens: crate::config::Config::load() + .map(|c| c.max_tokens) + .unwrap_or(2048), } } } @@ -116,6 +124,9 @@ impl ClaudeTranslator { timeout_seconds: crate::config::Config::load() .map(|c| c.timeout_seconds) .unwrap_or(20), + max_tokens: crate::config::Config::load() + .map(|c| c.max_tokens) + .unwrap_or(2048), } } } @@ -137,6 +148,9 @@ impl GeminiTranslator { timeout_seconds: crate::config::Config::load() .map(|c| c.timeout_seconds) .unwrap_or(20), + max_tokens: crate::config::Config::load() + .map(|c| c.max_tokens) + .unwrap_or(2048), } } } @@ -152,6 +166,9 @@ impl GrokTranslator { timeout_seconds: crate::config::Config::load() .map(|c| c.timeout_seconds) .unwrap_or(20), + max_tokens: crate::config::Config::load() + .map(|c| c.max_tokens) + .unwrap_or(2048), } } } @@ -167,6 +184,9 @@ impl QwenTranslator { timeout_seconds: crate::config::Config::load() .map(|c| c.timeout_seconds) .unwrap_or(20), + max_tokens: crate::config::Config::load() + .map(|c| c.max_tokens) + .unwrap_or(2048), } } } @@ -320,7 +340,8 @@ impl AiService for OpenAITranslator { debug!("发送给 OpenAI 的消息:\n{}", serde_json::to_string_pretty(&messages)?); let body = serde_json::json!({ "model": self.model, - "messages": messages + "messages": messages, + "max_tokens": self.max_tokens }); let ai_host = match url.split('/').nth(2) { @@ -394,7 +415,8 @@ impl AiService for ClaudeTranslator { debug!("发送给 Claude 的消息:\n{}", serde_json::to_string_pretty(&messages)?); let body = serde_json::json!({ "model": self.model, - "messages": messages + "messages": messages, + "max_tokens": self.max_tokens }); let ai_host = match url.split('/').nth(2) { @@ -490,7 +512,8 @@ impl AiService for GeminiTranslator { "parts": [{ "text": prompt }] - }] + }], + "max_tokens": self.max_tokens }); let ai_host = match url.split('/').nth(2) { @@ -563,7 +586,8 @@ impl AiService for GrokTranslator { debug!("发送给 Grok 的消息:\n{}", serde_json::to_string_pretty(&messages)?); let body = serde_json::json!({ "model": self.model, - "messages": messages + "messages": messages, + "max_tokens": self.max_tokens }); let ai_host = match url.split('/').nth(2) { @@ -638,13 +662,8 @@ impl AiService for QwenTranslator { let body = serde_json::json!({ "model": self.model, "messages": messages, - // temperature 控制输出的随机性，范围 0-1 - // 设置为 0.1 使输出更加确定和集中，减少不必要的发散 "temperature": 0.1, - // max_tokens 限制生成文本的最大长度 - // 对于提交信息翻译来说，256 tokens 足够用了 - // 限制长度可以显著提升响应速度 - "max_tokens": 256 + "max_tokens": self.max_tokens }); let ai_host = match url.split('/').nth(2) { From 6566bf8b171fee692e38188388bcbbaf10e795d1 Mon Sep 17 00:00:00 2001 From: rewine Date: Mon, 11 Aug 2025 11:47:45 +0800 Subject: [PATCH 2/2] Apply suggestion from @Copilot Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> --- src/ai_service.rs | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/src/ai_service.rs b/src/ai_service.rs index d340939..bf145a4 100644 --- a/src/ai_service.rs +++ b/src/ai_service.rs @@ -513,7 +513,9 @@ impl AiService for GeminiTranslator { "text": prompt }] }], - "max_tokens": self.max_tokens + "generationConfig": { + "maxOutputTokens": self.max_tokens + } }); let ai_host = match url.split('/').nth(2) {