subx-cli 1.7.4

AI subtitle processing CLI tool, which automatically matches, renames, and converts subtitle files.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220

use crate::Result;
use crate::error::SubXError;
use crate::services::ai::{AnalysisRequest, ConfidenceScore, MatchResult, VerificationRequest};
use serde_json;

/// Prompt builder trait for AI providers.
pub trait PromptBuilder {
    /// Build analysis prompt.
    fn build_analysis_prompt(&self, request: &AnalysisRequest) -> String {
        build_analysis_prompt_base(request)
    }

    /// Build verification prompt.
    fn build_verification_prompt(&self, request: &VerificationRequest) -> String {
        build_verification_prompt_base(request)
    }

    /// System message for analysis prompt.
    fn get_analysis_system_message() -> &'static str {
        "You are a professional subtitle matching assistant that can analyze the correspondence between video and subtitle files."
    }

    /// System message for verification prompt.
    fn get_verification_system_message() -> &'static str {
        "Please evaluate the confidence level of subtitle matching and provide a score between 0-1."
    }
}

/// Response parsing trait for AI providers.
pub trait ResponseParser {
    /// Parse match result.
    fn parse_match_result(&self, response: &str) -> Result<MatchResult> {
        parse_match_result_base(response)
    }

    /// Parse confidence score.
    fn parse_confidence_score(&self, response: &str) -> Result<ConfidenceScore> {
        parse_confidence_score_base(response)
    }
}

/// Build analysis prompt for AI providers.
pub fn build_analysis_prompt_base(request: &AnalysisRequest) -> String {
    let mut prompt = String::new();
    prompt.push_str(
        "Please analyze the matching relationship between the following video and subtitle files. Each file has a unique ID that you must use in your response.\n\n",
    );
    prompt.push_str("Video files:\n");
    for video in &request.video_files {
        prompt.push_str(&format!("- {}\n", video));
    }
    prompt.push_str("\nSubtitle files:\n");
    for subtitle in &request.subtitle_files {
        prompt.push_str(&format!("- {}\n", subtitle));
    }
    if !request.content_samples.is_empty() {
        prompt.push_str("\nSubtitle content preview:\n");
        for sample in &request.content_samples {
            prompt.push_str(&format!("File: {}\n", sample.filename));
            prompt.push_str(&format!("Content: {}\n\n", sample.content_preview));
        }
    }
    prompt.push_str(
        "Please provide matching suggestions based on filename patterns, content similarity, and other factors.\n\
Response format must be JSON using the file IDs:\n\
{\n\
  \"matches\": [\n\
    {\n\
      \"video_file_id\": \"file_019dcc51-f7da-74e3-9e0d-f75d40fc569c\",\n\
      \"subtitle_file_id\": \"file_019dcc51-f7d5-7640-8bb1-d2bbbc127a23\",\n\
      \"confidence\": 0.95,\n\
      \"match_factors\": [\"filename_similarity\", \"content_correlation\"]\n\
    }\n\
  ],\n\
  \"confidence\": 0.9,\n\
  \"reasoning\": \"Explanation for the matching decisions\"\n\
}",
    );
    prompt
}

/// Parse matching results from AI response.
pub fn parse_match_result_base(response: &str) -> Result<MatchResult> {
    let json_start = response.find('{').unwrap_or(0);
    let json_end = response.rfind('}').map(|i| i + 1).unwrap_or(response.len());
    let json_str = &response[json_start..json_end];
    serde_json::from_str(json_str)
        .map_err(|e| SubXError::AiService(format!("AI response parsing failed: {}", e)))
}

/// Build verification prompt for AI providers.
pub fn build_verification_prompt_base(request: &VerificationRequest) -> String {
    let mut prompt = String::new();
    prompt.push_str(
        "Please evaluate the confidence level based on the following matching information:\n",
    );
    prompt.push_str(&format!("Video file: {}\n", request.video_file));
    prompt.push_str(&format!("Subtitle file: {}\n", request.subtitle_file));
    prompt.push_str("Matching factors:\n");
    for factor in &request.match_factors {
        prompt.push_str(&format!("- {}\n", factor));
    }
    prompt.push_str(
        "\nPlease respond in JSON format as follows:\n{\"score\": 0.9,\"factors\": [\"...\"]}",
    );
    prompt
}

/// Parse confidence score from AI response.
pub fn parse_confidence_score_base(response: &str) -> Result<ConfidenceScore> {
    let json_start = response.find('{').unwrap_or(0);
    let json_end = response.rfind('}').map(|i| i + 1).unwrap_or(response.len());
    let json_str = &response[json_start..json_end];
    serde_json::from_str(json_str)
        .map_err(|e| SubXError::AiService(format!("AI confidence parsing failed: {}", e)))
}

#[cfg(test)]
mod tests {

    use crate::services::ai::prompts::{PromptBuilder, ResponseParser};
    use crate::services::ai::{AnalysisRequest, OpenAIClient};

    #[test]
    fn test_ai_prompt_with_file_ids_english() {
        let client = OpenAIClient::new("test_key".into(), "gpt-4.1".into(), 0.1, 1000, 0, 0);
        let video_id = "file_019dcc51-f7da-74e3-9e0d-f75d40fc569c";
        let subtitle_id = "file_019dcc51-f7d5-7640-8bb1-d2bbbc127a23";
        let request = AnalysisRequest {
            video_files: vec![format!("ID:{video_id} | Name:movie.mkv | Path:movie.mkv")],
            subtitle_files: vec![format!(
                "ID:{subtitle_id} | Name:movie.srt | Path:movie.srt"
            )],
            content_samples: vec![],
        };

        let prompt = client.build_analysis_prompt(&request);

        assert!(prompt.contains(&format!("ID:{video_id}")));
        assert!(prompt.contains("video_file_id"));
        assert!(prompt.contains("subtitle_file_id"));
        assert!(prompt.contains("Please analyze the matching"));
        assert!(prompt.contains("unique ID"));
        assert!(prompt.contains("Response format must be JSON"));
        assert!(!prompt.contains("請分析"));
        assert!(!prompt.contains("影片檔案"));
        assert!(!prompt.contains("字幕檔案"));
    }

    #[test]
    fn test_parse_match_result_with_ids() {
        let client = OpenAIClient::new("test_key".into(), "gpt-4.1".into(), 0.1, 1000, 0, 0);
        let video_id = "file_019dcc51-f7da-74e3-9e0d-f75d40fc569c";
        let subtitle_id = "file_019dcc51-f7d5-7640-8bb1-d2bbbc127a23";
        let json_resp = format!(
            r#"{{
            "matches": [{{
                "video_file_id": "{video_id}",
                "subtitle_file_id": "{subtitle_id}",
                "confidence": 0.95,
                "match_factors": ["filename_similarity"]
            }}],
            "confidence": 0.9,
            "reasoning": "Strong match based on filename patterns"
        }}"#
        );

        let result = client.parse_match_result(&json_resp).unwrap();
        assert_eq!(result.matches.len(), 1);
        assert_eq!(result.matches[0].video_file_id, video_id);
        assert_eq!(result.matches[0].subtitle_file_id, subtitle_id);
        assert_eq!(result.matches[0].confidence, 0.95);
        assert_eq!(result.matches[0].match_factors[0], "filename_similarity");
    }

    #[test]
    fn test_ai_prompt_structure_consistency() {
        let client = OpenAIClient::new("test_key".into(), "gpt-4.1".into(), 0.1, 1000, 0, 0);
        let request = AnalysisRequest {
            video_files: vec![
                "ID:file_video1 | Name:video1.mkv | Path:season1/video1.mkv".into(),
                "ID:file_video2 | Name:video2.mkv | Path:season1/video2.mkv".into(),
            ],
            subtitle_files: vec![
                "ID:file_sub1 | Name:sub1.srt | Path:season1/sub1.srt".into(),
                "ID:file_sub2 | Name:sub2.srt | Path:season1/sub2.srt".into(),
            ],
            content_samples: vec![],
        };

        let prompt = client.build_analysis_prompt(&request);

        assert!(prompt.contains("ID:file_video1"));
        assert!(prompt.contains("ID:file_video2"));
        assert!(prompt.contains("ID:file_sub1"));
        assert!(prompt.contains("ID:file_sub2"));
        assert!(prompt.contains("Video files:"));
        assert!(prompt.contains("Subtitle files:"));
        assert!(prompt.contains("Response format must be JSON"));
    }

    #[test]
    fn test_parse_confidence_score() {
        let client = OpenAIClient::new("test_key".into(), "gpt-4.1".into(), 0.1, 1000, 0, 0);
        let json_resp = r#"{
            "score": 0.88,
            "factors": ["filename_similarity", "content_correlation"]
        }"#;

        let result = client.parse_confidence_score(json_resp).unwrap();
        assert_eq!(result.score, 0.88);
        assert_eq!(
            result.factors,
            vec![
                "filename_similarity".to_string(),
                "content_correlation".to_string()
            ]
        );
    }
}