selfware 0.6.1

Your personal AI workshop — software you own, software that lasts
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

//! L1: Terminal State & Aesthetic Recognition (Easy)
//!
//! Input: Screenshots of selfware's ratatui TUI in various states.
//! Tasks: Identify active panel, read status bar text, detect color theme,
//!        count visible widgets.
//! Scoring: Keyword match, structured JSON parse.
//! Pass threshold: 80% keyword accuracy.

use async_trait::async_trait;
use std::path::PathBuf;

use crate::vlm_bench::scoring::{self, LevelScore, Rating};
use crate::vlm_bench::{BenchScenario, Difficulty, ExpectedAnswer, VlmBenchLevel};

const PASS_THRESHOLD: f64 = 0.80;

/// Level 1: Terminal State & Aesthetic Recognition.
pub struct L1TuiState {
    fixtures_dir: PathBuf,
}

impl Default for L1TuiState {
    fn default() -> Self {
        Self::new()
    }
}

impl L1TuiState {
    pub fn new() -> Self {
        Self {
            fixtures_dir: PathBuf::from("vlm_fixtures/l1_tui_state"),
        }
    }
}

#[async_trait]
impl VlmBenchLevel for L1TuiState {
    fn name(&self) -> &str {
        "L1 TUI State"
    }

    fn difficulty(&self) -> Difficulty {
        Difficulty::Easy
    }

    fn description(&self) -> &str {
        "Terminal state recognition: identify panels, read status text, \
         detect color themes, and count visible widgets in TUI screenshots."
    }

    fn scenarios(&self) -> Vec<BenchScenario> {
        vec![
            BenchScenario {
                id: "l1_dashboard_normal".into(),
                description: "Identify the active panel and status in a normal dashboard view"
                    .into(),
                image_path: self.fixtures_dir.join("dashboard_normal.png"),
                prompt: "Analyze this terminal UI screenshot. Respond with a JSON object containing: \
                         {\"active_panel\": \"<panel name>\", \"status\": \"<status bar text>\", \
                         \"widget_count\": <number of visible widgets>, \"theme\": \"<dark or light>\"}".into(),
                expected: ExpectedAnswer::JsonFields(serde_json::json!({
                    "active_panel": "dashboard",
                    "theme": "dark"
                })),
            },
            BenchScenario {
                id: "l1_dashboard_error".into(),
                description: "Detect error state in a dashboard with error indicators".into(),
                image_path: self.fixtures_dir.join("dashboard_error.png"),
                prompt: "Analyze this terminal UI screenshot showing an error state. \
                         What error or warning is visible? Respond with JSON: \
                         {\"has_error\": true/false, \"error_type\": \"<type>\", \
                         \"active_panel\": \"<panel name>\"}".into(),
                expected: ExpectedAnswer::Keywords(vec![
                    "error".into(),
                    "true".into(),
                ]),
            },
            BenchScenario {
                id: "l1_help_panel".into(),
                description: "Read content from a help/documentation panel".into(),
                image_path: self.fixtures_dir.join("help_panel.png"),
                prompt: "This screenshot shows a help panel in a terminal UI. \
                         List the keyboard shortcuts or commands visible. \
                         Respond with JSON: {\"panel_type\": \"help\", \
                         \"shortcuts\": [\"<shortcut1>\", ...]}".into(),
                expected: ExpectedAnswer::Keywords(vec![
                    "help".into(),
                    "shortcut".into(),
                ]),
            },
            BenchScenario {
                id: "l1_loading_state".into(),
                description: "Identify a loading/spinner state".into(),
                image_path: self.fixtures_dir.join("loading_state.png"),
                prompt: "Analyze this terminal UI screenshot. Is the interface in a loading state? \
                         What visual indicators suggest loading? Respond with JSON: \
                         {\"is_loading\": true/false, \"indicators\": [\"<indicator>\", ...], \
                         \"active_panel\": \"<panel name>\"}".into(),
                expected: ExpectedAnswer::Keywords(vec![
                    "loading".into(),
                    "true".into(),
                ]),
            },
        ]
    }

    fn evaluate(&self, scenario: &BenchScenario, response: &str) -> LevelScore {
        crate::vlm_bench::score_response(scenario, response, PASS_THRESHOLD)
    }
}

#[cfg(test)]
#[path = "../../../tests/unit/vlm_bench/levels/l1_tui_state/l1_tui_state_test.rs"]
mod tests;