\u5927\u5bb6\u597d，\u6211\u662f\u4e00\u94ed\u3002\u505a\u8fc7\u540e\u7aef\u5f00\u53d1\u7684\u540c\u5b66\u5e94\u8be5\u90fd\u6709\u8fd9\u6837\u7684\u7ecf\u5386：\u751f\u4ea7\u73af\u5883\u6302\u4e86，\u6253\u5f00\u65e5\u5fd7\u4e00\u770b，\u51e0\u767e\u5146\u7684\u975e\u7ed3\u6784\u5316\u6587\u672c，\u683c\u5f0f\u4e94\u82b1\u516b\u95e8\u2014\u2014\u6709\u7684\u662f Nginx \u7684 access log，\u6709\u7684\u662f\u5e94\u7528\u81ea\u5df1 println! \u6253\u51fa\u6765\u7684，\u8fd8\u6709\u5404\u79cd\u7b2c\u4e09\u65b9 SDK \u7684 debug \u8f93\u51fa\u3002\u60f3\u4ece\u4e2d\u63d0\u53d6\u5173\u952e\u4fe1\u606f？\u53ea\u80fd\u4eba\u773c\u4e00\u884c\u884c\u770b，\u7136\u540e\u7528 grep\u3001awk\u3001sed \u624b\u5de5\u62fc\u51d1\u3002<\/p>\n
\u6709\u6ca1\u6709\u4e00\u79cd\u65b9\u6cd5，\u80fd\u628a\u8fd9\u4e9b\u4e71\u4e03\u516b\u7cdf\u7684\u65e5\u5fd7，\u81ea\u52a8\u5730\u3001\u51c6\u786e\u5730\u8f6c\u6210\u7ed3\u6784\u5316\u7684 JSON，\u65b9\u4fbf\u540e\u7eed\u505a\u76d1\u63a7\u3001\u544a\u8b66\u3001\u5206\u6790？<\/p>\n
\u7b54\u6848\u662f：\u7528 AI \u6a21\u578b\u505a\u6570\u636e\u6e05\u6d17\u3002<\/p>\n
\u6574\u4e2a\u5de5\u5177\u5206\u4e3a\u4e09\u5c42：<\/p>\n
\u5148\u6765\u770b\u770b CLI \u5165\u53e3\u7684\u7ed3\u6784\u5b9a\u4e49：<\/p>\n
use clap::Parser;<\/p>\n
\/\/\/ AI \u9a71\u52a8\u7684\u65e5\u5fd7\u6e05\u6d17 CLI \u5de5\u5177
\n\/\/\/ \u5c06\u975e\u7ed3\u6784\u5316\u65e5\u5fd7\u81ea\u52a8\u8f6c\u4e3a\u7ed3\u6784\u5316 JSON
\n#[derive(Parser, Debug)]
\n#[command(name = "log-cleaner")]
\n#[command(version = "0.1.0")]
\n#[command(about = "\u7528 AI \u6a21\u578b\u628a\u65e5\u5fd7\u8f6c\u6210\u7ed3\u6784\u5316 JSON", long_about = None)]
\nstruct Cli {
\n \/\/\/ \u8f93\u5165\u65e5\u5fd7\u6587\u4ef6\u8def\u5f84（\u4e0d\u4f20\u5219\u4ece\u6807\u51c6\u8f93\u5165\u8bfb\u53d6）
\n #[arg(short, long)]
\n input: Option<String>,<\/p>\n
\/\/\/ \u8f93\u51fa JSON \u6587\u4ef6\u8def\u5f84（\u4e0d\u4f20\u5219\u8f93\u51fa\u5230\u6807\u51c6\u8f93\u51fa）
\n #[arg(short, long)]
\n output: Option<String>,<\/p>\n
\/\/\/ \u672c\u5730\u6a21\u578b API \u5730\u5740
\n #[arg(short = 'm', long, default_value = "http:\/\/localhost:11434")]
\n model_url: String,<\/p>\n
\/\/\/ \u6a21\u578b\u540d\u79f0
\n #[arg(short = 'n', long, default_value = "qwen2.5:7b")]
\n model_name: String,<\/p>\n
\/\/\/ \u6bcf\u6b21\u9001\u6a21\u578b\u7684\u6700\u5927\u884c\u6570
\n #[arg(short = 'b', long, default_value_t = 50)]
\n batch_size: usize,
\n}<\/p>\n
\u4e0b\u9762\u662f\u6838\u5fc3\u7684\u6e05\u6d17\u903b\u8f91\u2014\u2014\u628a\u6bcf\u6279\u65e5\u5fd7\u9001\u7ed9\u672c\u5730\u6a21\u578b，\u8ba9\u5b83\u8fd4\u56de\u7ed3\u6784\u5316\u7684 JSON：<\/p>\n
use serde_json::Value;
\nuse reqwest::Client;<\/p>\n
\/\/\/ \u7528\u672c\u5730 AI \u6a21\u578b\u6e05\u6d17\u4e00\u6279\u65e5\u5fd7\u884c
\nasync fn clean_batch(
\n client: &Client,
\n model_url: &str,
\n model_name: &str,
\n lines: &[String],
\n) -> Result<Vec<Value>, Box<dyn std::error::Error>> {
\n \/\/ \u62fc\u63a5\u65e5\u5fd7\u6587\u672c，\u4f5c\u4e3a prompt \u7684\u4e00\u90e8\u5206
\n let log_text = lines.join("\\\\n");<\/p>\n
\/\/ \u6784\u5efa\u53d1\u9001\u7ed9\u6a21\u578b\u7684\u63d0\u793a\u8bcd
\n \/\/ \u8981\u6c42\u6a21\u578b\u8fd4\u56de\u4e25\u683c\u7684 JSON \u6570\u7ec4\u683c\u5f0f
\n let prompt = format!(
\n r#"\u4f60\u662f\u4e00\u4e2a\u65e5\u5fd7\u89e3\u6790\u5f15\u64ce\u3002\u8bf7\u628a\u4e0b\u9762\u7684\u539f\u59cb\u65e5\u5fd7\u89e3\u6790\u6210 JSON \u6570\u7ec4\u3002
\n\u6bcf\u6761\u65e5\u5fd7\u8f93\u51fa\u4e00\u4e2a JSON \u5bf9\u8c61，\u5305\u542b\u4ee5\u4e0b\u5b57\u6bb5：
\n– timestamp: \u65f6\u95f4\u6233（ISO 8601 \u683c\u5f0f）
\n– level: \u65e5\u5fd7\u7ea7\u522b（INFO\/WARN\/ERROR\/DEBUG）
\n– message: \u65e5\u5fd7\u5185\u5bb9
\n– source: \u6765\u6e90（\u5982\u679c\u80fd\u8bc6\u522b）
\n– extra: \u5176\u4ed6\u989d\u5916\u4fe1\u606f<\/p>\n
\u53ea\u8fd4\u56de JSON \u6570\u7ec4，\u4e0d\u8981\u52a0\u4efb\u4f55\u89e3\u91ca\u6216 markdown \u6807\u8bb0\u3002
\n\u539f\u59cb\u65e5\u5fd7：
\n{}"#,
\n log_text
\n );<\/p>\n
\/\/ \u8c03\u7528\u672c\u5730 Ollama API
\n let resp = client
\n .post(format!("{}\/api\/generate", model_url))
\n .json(&serde_json::json!({
\n "model": model_name,
\n "prompt": prompt,
\n "stream": false,
\n "format": "json", \/\/ \u5f3a\u5236\u6a21\u578b\u8fd4\u56de\u5408\u6cd5 JSON
\n "options": {
\n "temperature": 0.0 \/\/ \u6e29\u5ea6\u8bbe\u4e3a0，\u786e\u4fdd\u8f93\u51fa\u7a33\u5b9a
\n }
\n }))
\n .send()
\n .await?;<\/p>\n
let body: Value = resp.json().await?;<\/p>\n
\/\/ \u4ece\u6a21\u578b\u54cd\u5e94\u4e2d\u63d0\u53d6 JSON
\n let response_text = body["response"].as_str().unwrap_or("[]");
\n let parsed: Vec<Value> = serde_json::from_str(response_text)?;<\/p>\n
Ok(parsed)
\n}<\/p>\n
\u6709\u4e86\u5355\u6279\u6b21\u7684\u6e05\u6d17\u903b\u8f91，\u63a5\u4e0b\u6765\u7528 rayon \u5e76\u884c\u5904\u7406\u6574\u4e2a\u6587\u4ef6\u3002\u6ce8\u610f\u8fd9\u91cc\u7528 tokio \u7684 spawn_blocking \u6765\u907f\u514d\u963b\u585e\u5f02\u6b65\u8fd0\u884c\u65f6：<\/p>\n
use rayon::prelude::*;<\/p>\n
\/\/\/ \u5e76\u884c\u5904\u7406\u6574\u4e2a\u65e5\u5fd7\u6587\u4ef6
\nasync fn process_file(input_path: &str, batch_size: usize) -> Vec<Value> {
\n \/\/ \u8bfb\u53d6\u5168\u90e8\u65e5\u5fd7\u5185\u5bb9（\u5bf9\u4e8e\u5927\u6587\u4ef6\u53ef\u4ee5\u8003\u8651 memory-map）
\n let content = tokio::fs::read_to_string(input_path)
\n .await
\n .expect("\u65e0\u6cd5\u8bfb\u53d6\u65e5\u5fd7\u6587\u4ef6");<\/p>\n
\/\/ \u6309\u884c\u5207\u5206，\u518d\u6309 batch_size \u5206\u5757
\n let lines: Vec<&str> = content.lines().collect();
\n let chunks: Vec<Vec<String>> = lines
\n .chunks(batch_size)
\n .map(|chunk| chunk.iter().map(|s| s.to_string()).collect())
\n .collect();<\/p>\n
let client = reqwest::Client::new();<\/p>\n
\/\/ \u5e76\u884c\u8c03\u7528\u6a21\u578b\u5904\u7406\u6bcf\u4e2a\u5757
\n let results: Vec<Vec<Value>> = futures::future::join_all(
\n chunks.iter().map(|chunk| {
\n clean_batch(&client, "http:\/\/localhost:11434", "qwen2.5:7b", chunk)
\n })
\n )
\n .await
\n .into_iter()
\n .filter_map(|r| r.ok()) \/\/ \u8df3\u8fc7\u5904\u7406\u5931\u8d25\u7684\u6279\u6b21
\n .collect();<\/p>\n
\/\/ \u5c55\u5e73\u6240\u6709\u7ed3\u679c
\n results.into_iter().flatten().collect()
\n}<\/p>\n
\u4e0b\u9762\u662f\u5b8c\u6574\u7684\u5904\u7406\u6d41\u7a0b\u56fe，\u5c55\u793a\u4ece\u8f93\u5165\u5230\u8f93\u51fa\u7684\u6bcf\u4e2a\u73af\u8282：<\/p>\n
\u672c\u5730\u8dd1\u63a8\u8350 qwen2.5:7b \u6216 llama3.1:8b\u3002\u4e24\u8005\u90fd\u80fd\u5728 16GB \u5185\u5b58\u7684\u673a\u5668\u4e0a\u6d41\u7545\u8fd0\u884c\u3002\u5982\u679c\u7528\u4e91\u7aef API，\u53ef\u4ee5\u8003\u8651 gpt-4o-mini，\u6210\u672c\u6781\u4f4e\u3002<\/p>\n
\u8fd9\u662f\u6574\u4e2a\u65b9\u6848\u6700\u5173\u952e\u7684\u4e00\u73af\u3002\u6211\u7684\u7ecf\u9a8c\u662f：<\/p>\n
\u5904\u7406 GB \u7ea7\u65e5\u5fd7\u65f6，\u4e0d\u8981 read_to_string，\u6539\u7528 memory-mapped file：<\/p>\n
use memmap2::Mmap;
\nuse std::fs::File;<\/p>\n
\/\/\/ \u9ad8\u6548\u8bfb\u53d6\u5927\u6587\u4ef6：\u4f7f\u7528\u5185\u5b58\u6620\u5c04
\nfn read_large_file(path: &str) -> Result<Mmap, std::io::Error> {
\n let file = File::open(path)?;
\n \/\/ \u5b89\u5168：\u53ea\u8bfb\u6620\u5c04，\u4e0d\u4f1a\u4fee\u6539\u6e90\u6587\u4ef6
\n let mmap = unsafe { Mmap::map(&file)? };
\n Ok(mmap)
\n}<\/p>\n
memmap \u8ba9\u64cd\u4f5c\u7cfb\u7edf\u6309\u9700\u52a0\u8f7d\u9875\u9762，\u5373\u4f7f\u65e5\u5fd7\u6587\u4ef6\u5927\u5230 10GB，\u5b9e\u9645\u5185\u5b58\u5360\u7528\u4e5f\u53ea\u6709\u51e0\u767e MB\u3002<\/p>\n
\u5b9e\u9645\u9879\u76ee\u91cc\u8e29\u8fc7\u4e00\u4e2a\u5751：\u7528 Ollama \u7684 format: json \u53c2\u6570\u5f3a\u5236\u7ed3\u6784\u5316\u8f93\u51fa，\u6a21\u578b\u6709\u65f6\u8fd4\u56de\u7684\u4e0d\u662f\u7eaf JSON \u6570\u7ec4，\u800c\u662f\u5e26\u8bf4\u660e\u6587\u5b57\u7684 JSON\u3002\u7ed9 prompt \u52a0\u4e0a "\u53ea\u8fd4\u56de JSON \u6570\u7ec4，\u4e0d\u8981\u52a0\u4efb\u4f55\u89e3\u91ca" \u540e\u597d\u4e86\u4e00\u4e9b，\u4f46\u6211\u53d1\u73b0\u6279\u91cf\u5904\u7406 500 \u884c\u65e5\u5fd7\u65f6，\u6700\u540e\u4e00\u6279\u5076\u5c14\u8fd8\u662f\u4f1a\u591a\u51fa\u4e00\u53e5 "\u5df2\u5b8c\u6210\u5904\u7406"\u3002\u6700\u7ec8\u52a0\u4e86\u4e00\u5c42 serde_json::from_str \u7684\u9519\u8bef\u91cd\u8bd5\u673a\u5236\u89e3\u51b3\u3002<\/p>\n
Rust + AI \u662f\u4e00\u4e2a\u975e\u5e38\u6709\u6f5c\u529b\u7684\u7ec4\u5408\u3002Rust \u8d1f\u8d23\u9ad8\u6027\u80fd\u3001\u4f4e\u5185\u5b58\u7684\u5de5\u7a0b\u5e95\u5ea7，AI \u8d1f\u8d23\u5904\u7406\u90a3\u4e9b\u4f20\u7edf\u89c4\u5219\u5f15\u64ce\u641e\u4e0d\u5b9a\u7684"\u810f\u6d3b\u7d2f\u6d3b"\u3002\u4e24\u8005\u7ed3\u5408，\u80fd\u505a\u5f88\u591a\u4ee5\u524d\u60f3\u90fd\u4e0d\u6562\u60f3\u7684\u4e8b\u60c5\u3002<\/p>\n
\u5b8c\u6574\u4ee3\u7801\u6211\u653e\u5230\u4e86 GitHub \u4e0a（log-cleaner \u4ed3\u5e93），\u611f\u5174\u8da3\u7684\u540c\u5b66\u53ef\u4ee5 clone \u4e0b\u6765\u8dd1\u4e00\u8dd1\u3002\u5982\u679c\u6709\u4efb\u4f55\u95ee\u9898，\u6b22\u8fce\u5728\u8bc4\u8bba\u533a\u4ea4\u6d41！<\/p>\n","protected":false},"excerpt":{"rendered":"
Rust AI \u6570\u636e\u6e05\u6d17 CLI：\u628a\u975e\u7ed3\u6784\u5316\u65e5\u5fd7\u7528\u6a21\u578b\u8f6c\u6210\u7ed3\u6784\u5316 JSON \u7684\u65b9\u6848
\n\u4e00\u3001\u95ee\u9898\u573a\u666f：\u65e5\u5fd7\u6e05\u6d17\u7684\u75db\u70b9
\n\u5927\u5bb6\u597d，\u6211\u662f\u4e00\u94ed\u3002\u505a\u8fc7\u540e\u7aef\u5f00\u53d1\u7684\u540c\u5b66\u5e94\u8be5\u90fd\u6709\u8fd9\u6837\u7684\u7ecf\u5386：\u751f\u4ea7\u73af\u5883\u6302\u4e86，\u6253\u5f00\u65e5\u5fd7\u4e00\u770b，\u51e0\u767e\u5146\u7684\u975e\u7ed3\u6784\u5316\u6587\u672c，\u683c\u5f0f\u4e94\u82b1\u516b\u95e8\u2014\u2014\u6709\u7684\u662f Nginx \u7684 access log，\u6709\u7684\u662f\u5e94\u7528\u81ea\u5df1 println! \u6253\u51fa\u6765\u7684，\u8fd8\u6709\u5404\u79cd\u7b2c\u4e09\u65b9 SDK \u7684 <\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[92,81,206],"topic":[],"class_list":["post-99403","post","type-post","status-publish","format-standard","hentry","category-server","tag-github","tag-python","tag-rust"],"yoast_head":"\n