lean-ctx 3.9.1

Context Runtime for AI Agents with CCP. 71 MCP tools, 10 read modes, 95+ compression patterns, cross-session memory (CCP), persistent AI knowledge with temporal facts + contradiction detection, multi-agent context sharing, LITM-aware positioning, AAAK compact format, adaptive compression with Thompson Sampling bandits. Supports 24+ AI tools. Reduces LLM token consumption by up to 99%.
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
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
use std::collections::HashMap;

use crate::core::stats::StatsStore;
use crate::core::tokens::count_tokens;
use crate::shell::output_policy::{OutputPolicy, classify};

/// Command families with a dedicated lean-ctx compressor, used to *recognise*
/// compressible commands in shell history (the savings numbers themselves come
/// from real measured `core::stats`, never from this table). `base` is the
/// `normalize_command` base name; keep roughly in sync with the dispatch in
/// `core::patterns::try_specific_pattern`.
const COMPRESSIBLE_FAMILIES: &[(&str, &str)] = &[
    ("git", "git status/diff/log/commit/push"),
    ("gh", "GitHub CLI"),
    ("glab", "GitLab CLI"),
    ("cargo", "cargo build/test/clippy"),
    ("npm", "npm install/run/test"),
    ("pnpm", "pnpm install/run/test"),
    ("yarn", "yarn install/run/test"),
    ("bun", "Bun runtime"),
    ("deno", "Deno runtime"),
    ("docker", "docker ps/images/logs/build"),
    ("kubectl", "kubectl get/describe/logs"),
    ("helm", "Kubernetes Helm"),
    ("pip", "pip install/list/freeze"),
    ("poetry", "Poetry"),
    ("uv", "uv add/lock/sync"),
    ("conda", "conda/mamba env"),
    ("pipx", "pipx install"),
    ("go", "go test/build/vet"),
    ("mypy", "mypy type check"),
    ("pyright", "pyright type check"),
    ("ruff", "ruff check/format"),
    ("eslint", "eslint/biome lint"),
    ("prettier", "prettier --check"),
    ("tsc", "TypeScript compiler"),
    ("pytest", "Python tests"),
    ("jest", "Jest tests"),
    ("vitest", "Vitest tests"),
    ("mocha", "Mocha tests"),
    ("playwright", "Playwright tests"),
    ("rspec", "Ruby tests"),
    ("rubocop", "RuboCop lint"),
    ("bundle", "Bundler"),
    ("rake", "Rake tasks"),
    ("curl", "HTTP requests"),
    ("wget", "HTTP downloads"),
    ("grep", "grep/rg search"),
    ("rg", "ripgrep search"),
    ("find", "find files"),
    ("fd", "fd file search"),
    ("ls", "directory listing"),
    ("jq", "JSON processing"),
    ("aws", "AWS CLI"),
    ("terraform", "Terraform"),
    ("tofu", "OpenTofu"),
    ("pulumi", "Pulumi IaC"),
    ("ansible", "Ansible"),
    ("prisma", "Prisma ORM"),
    ("psql", "PostgreSQL"),
    ("mysql", "MySQL/MariaDB"),
    ("cmake", "CMake build"),
    ("ninja", "Ninja build"),
    ("bazel", "Bazel build"),
    ("make", "Make targets"),
    ("just", "just recipes"),
    ("mvn", "Maven build"),
    ("gradle", "Gradle build"),
    ("dotnet", "dotnet build/test"),
    ("flutter", "Flutter build"),
    ("swift", "Swift build/test"),
    ("zig", "Zig build/test"),
    ("composer", "PHP Composer"),
    ("mix", "Elixir Mix"),
    ("next", "Next.js build"),
    ("vite", "Vite build"),
    ("turbo", "Turborepo"),
    ("nx", "Nx monorepo"),
    ("systemctl", "systemd units"),
    ("journalctl", "systemd logs"),
    ("dbt", "dbt models/tests"),
    ("alembic", "Alembic migrations"),
    ("flyway", "Flyway migrations"),
    ("ollama", "Ollama models"),
    ("mlflow", "MLflow runs"),
    ("semgrep", "Semgrep scan"),
    ("trivy", "Trivy scan"),
    ("grype", "Grype scan"),
    ("syft", "Syft SBOM"),
    ("cosign", "Cosign verify"),
    ("swiftlint", "SwiftLint"),
    ("jj", "Jujutsu VCS"),
    ("mise", "mise toolchain"),
    ("buf", "Protobuf buf"),
    ("gem", "RubyGems"),
    ("linkerd", "Linkerd check"),
    ("argocd", "Argo CD"),
    ("vercel", "Vercel deploy"),
    ("fly", "Fly.io deploy"),
    ("wrangler", "Cloudflare deploy"),
    ("skaffold", "Skaffold"),
    ("supabase", "Supabase"),
];

pub struct DiscoverResult {
    pub total_commands: u32,
    pub already_optimized: u32,
    pub missed_commands: Vec<MissedCommand>,
    pub potential_tokens: usize,
    pub potential_usd: f64,
    /// True when at least one missed command had real measured savings backing
    /// its token estimate. When false, `potential_tokens` is 0 and callers
    /// should fall back to a frequency-based framing instead of a $ figure.
    pub has_measured_data: bool,
}

pub struct MissedCommand {
    pub prefix: String,
    pub description: String,
    /// Real measured savings for this family (e.g. "84%"), or "—" when the
    /// user has not yet run this family through lean-ctx.
    pub savings_range: String,
    pub count: u32,
    pub estimated_tokens: usize,
    pub measured: bool,
}

/// First whitespace token's file name — the `normalize_command` base.
fn base_of(command: &str) -> &str {
    let first = command.split_whitespace().next().unwrap_or(command);
    std::path::Path::new(first)
        .file_name()
        .and_then(|n| n.to_str())
        .unwrap_or(first)
}

fn describe(base: &str) -> Option<&'static str> {
    COMPRESSIBLE_FAMILIES
        .iter()
        .find(|(b, _)| *b == base)
        .map(|(_, d)| *d)
}

/// Aggregated real measurements per command family from `core::stats`:
/// base → (input_tokens, output_tokens, count).
fn measured_by_family(store: &StatsStore) -> HashMap<String, (u64, u64, u64)> {
    let mut by_base: HashMap<String, (u64, u64, u64)> = HashMap::new();
    for (key, s) in &store.commands {
        let entry = by_base.entry(base_of(key).to_string()).or_default();
        entry.0 = entry.0.saturating_add(s.input_tokens);
        entry.1 = entry.1.saturating_add(s.output_tokens);
        entry.2 = entry.2.saturating_add(s.count);
    }
    by_base
}

pub fn analyze_history(history: &[String], limit: usize) -> DiscoverResult {
    let store = crate::core::stats::load();
    analyze_history_with_stats(history, limit, &store)
}

/// Core analysis with the measured-stats source injected (testable, pure aside
/// from the policy classifier). `analyze_history` is the disk-backed wrapper.
fn analyze_history_with_stats(
    history: &[String],
    limit: usize,
    store: &StatsStore,
) -> DiscoverResult {
    let mut missed: HashMap<String, u32> = HashMap::new();
    let mut already_optimized = 0u32;
    let mut total_commands = 0u32;

    let measured = measured_by_family(store);

    for cmd in history {
        let trimmed = cmd.trim();
        if trimmed.is_empty() {
            continue;
        }
        total_commands += 1;

        if trimmed.starts_with("lean-ctx ") || trimmed.starts_with("lean-ctx\t") {
            already_optimized += 1;
            continue;
        }

        let base = base_of(trimmed);
        // A command is a missed save only if lean-ctx has a compressor for it
        // (known family or already measured) AND the policy engine would
        // actually compress it (not a protected verbatim/passthrough command).
        let known = describe(base).is_some() || measured.contains_key(base);
        if known && classify(trimmed, &[]) == OutputPolicy::Compressible {
            *missed.entry(base.to_string()).or_insert(0) += 1;
        }
    }

    let mut sorted: Vec<_> = missed.into_iter().collect();
    sorted.sort_by_key(|x| std::cmp::Reverse(x.1));

    let price_per_tok = crate::core::stats::DEFAULT_INPUT_PRICE_PER_M / 1_000_000.0;
    let mut potential_tokens = 0usize;
    let mut has_measured_data = false;

    let missed_commands: Vec<MissedCommand> = sorted
        .into_iter()
        .take(limit)
        .map(|(base, count)| {
            let description = describe(&base).unwrap_or("compressible output").to_string();
            // Project the family's *real* measured savings onto its plain-shell
            // frequency. Families without measurements contribute nothing — we
            // never invent a token figure.
            let (savings_range, estimated_tokens, is_measured) = match measured.get(&base) {
                Some(&(input, output, cnt)) if input > 0 && cnt > 0 => {
                    let rate = 1.0 - (output as f64 / input as f64);
                    let avg_input = input as f64 / cnt as f64;
                    let est = (count as f64 * avg_input * rate).max(0.0) as usize;
                    has_measured_data = true;
                    potential_tokens += est;
                    (format!("{:.0}%", (rate * 100.0).max(0.0)), est, true)
                }
                _ => ("".to_string(), 0, false),
            };
            MissedCommand {
                prefix: base,
                description,
                savings_range,
                count,
                estimated_tokens,
                measured: is_measured,
            }
        })
        .collect();

    let potential_usd = potential_tokens as f64 * price_per_tok;

    DiscoverResult {
        total_commands,
        already_optimized,
        missed_commands,
        potential_tokens,
        potential_usd,
        has_measured_data,
    }
}

pub fn discover_from_history(history: &[String], limit: usize) -> String {
    let result = analyze_history(history, limit);

    if result.missed_commands.is_empty() {
        return format!(
            "No missed savings found in last {} commands. \
            {} already optimized.",
            result.total_commands, result.already_optimized
        );
    }

    let mut lines = Vec::new();
    lines.push(format!(
        "Analyzed {} commands ({} already optimized):",
        result.total_commands, result.already_optimized
    ));
    lines.push(String::new());

    let total_missed: u32 = result.missed_commands.iter().map(|m| m.count).sum();
    lines.push(format!(
        "{total_missed} commands could benefit from lean-ctx:"
    ));
    lines.push(String::new());

    for m in &result.missed_commands {
        lines.push(format!(
            "  {:>4}x  {:<12} {} ({})",
            m.count, m.prefix, m.description, m.savings_range
        ));
    }

    lines.push(String::new());
    if result.has_measured_data {
        lines.push(format!(
            "Estimated potential: ~{} tokens saved (~${:.2}), projected from your measured savings",
            result.potential_tokens, result.potential_usd
        ));
    } else {
        lines.push(
            "No measured savings yet — run these via lean-ctx, then re-run discover for real numbers."
                .to_string(),
        );
    }
    lines.push(String::new());
    lines.push("Fix: run 'lean-ctx init --global' to auto-compress all commands.".to_string());
    lines.push("Or:  run 'lean-ctx init --agent <tool>' for AI tool hooks.".to_string());

    let output = lines.join("\n");
    let tokens = count_tokens(&output);
    format!("{output}\n\n[{tokens} tok]")
}

pub fn format_cli_output(result: &DiscoverResult) -> String {
    if result.missed_commands.is_empty() {
        return format!(
            "All compressible commands are already using lean-ctx!\n\
             ({} commands analyzed, {} via lean-ctx)",
            result.total_commands, result.already_optimized
        );
    }

    let mut lines = Vec::new();
    let total_missed: u32 = result.missed_commands.iter().map(|m| m.count).sum();

    lines.push(format!(
        "Found {total_missed} compressible commands not using lean-ctx:\n"
    ));
    lines.push(format!(
        "  {:<14} {:>5}  {:>10}  {:<30} {}",
        "COMMAND", "COUNT", "SAVINGS", "DESCRIPTION", "EST. TOKENS"
    ));
    lines.push(format!("  {}", "-".repeat(80)));

    for m in &result.missed_commands {
        let est = if m.measured {
            format!("~{}", m.estimated_tokens)
        } else {
            "".to_string()
        };
        lines.push(format!(
            "  {:<14} {:>5}x {:>10}  {:<30} {}",
            m.prefix, m.count, m.savings_range, m.description, est
        ));
    }

    lines.push(String::new());
    if result.has_measured_data {
        lines.push(format!(
            "Estimated missed savings: ~{} tokens (~${:.2}/month), projected from your measured rate",
            result.potential_tokens,
            result.potential_usd * 30.0
        ));
    } else {
        lines.push(
            "No measured savings yet — route these through lean-ctx, then re-run discover."
                .to_string(),
        );
    }
    lines.push(format!(
        "Already using lean-ctx: {} commands",
        result.already_optimized
    ));
    lines.push(String::new());
    lines.push("Run 'lean-ctx init --global' to enable compression for all commands.".to_string());

    lines.join("\n")
}

/// Renders a shareable "before lean-ctx" SVG card from a discover analysis — the
/// "ghost tokens you're leaving on the table" framing that drives the first-run share
/// loop. Same 1200x630 social-card dimensions and visual language as the Wrapped card,
/// but in an amber/red "leak" palette. Pure string building; all data-derived text is
/// XML-escaped. Aggregate estimates only — never command contents or arguments.
pub fn render_before_card(result: &DiscoverResult) -> String {
    let saved = crate::core::wrapped::format_tokens(result.potential_tokens as u64);
    let monthly_usd = result.potential_usd * 30.0;
    let total_missed: u32 = result.missed_commands.iter().map(|m| m.count).sum();
    let top = before_card_top_commands(result);
    format!(
        r##"<svg xmlns="http://www.w3.org/2000/svg" width="1200" height="630" viewBox="0 0 1200 630" font-family="Inter, system-ui, -apple-system, Segoe UI, Roboto, sans-serif">
  <defs>
    <linearGradient id="bg" x1="0" y1="0" x2="1" y2="1">
      <stop offset="0" stop-color="#0b1020"/>
      <stop offset="1" stop-color="#131a2e"/>
    </linearGradient>
    <linearGradient id="accent" x1="0" y1="0" x2="1" y2="0">
      <stop offset="0" stop-color="#f59e0b"/>
      <stop offset="1" stop-color="#ef4444"/>
    </linearGradient>
  </defs>
  <rect width="1200" height="630" fill="url(#bg)"/>
  <rect x="0" y="0" width="1200" height="8" fill="url(#accent)"/>
  <text x="70" y="92" fill="#e5e7eb" font-size="34" font-weight="700">lean-ctx <tspan fill="#f59e0b">Ghost Tokens</tspan></text>
  <text x="70" y="130" fill="#94a3b8" font-size="24">before lean-ctx — estimated from my shell history</text>
  <text x="70" y="300" fill="#f59e0b" font-size="120" font-weight="800" font-family="ui-monospace, SFMono-Regular, Menlo, monospace">{saved}</text>
  <text x="76" y="346" fill="#94a3b8" font-size="26">tokens/month left on the table</text>
  <text x="70" y="430" fill="#e5e7eb" font-size="60" font-weight="800" font-family="ui-monospace, SFMono-Regular, Menlo, monospace">${monthly_usd:.0}</text>
  <text x="74" y="462" fill="#94a3b8" font-size="22">potential monthly savings</text>
  <text x="70" y="512" fill="#cbd5e1" font-size="22">{total_missed} uncompressed commands · {already} already via lean-ctx</text>
{top}
  <text x="70" y="600" fill="#475569" font-size="17">Estimate from local shell history · run `lean-ctx onboard` to stop the leak</text>
  <text x="1130" y="600" text-anchor="end" fill="#f59e0b" font-size="26" font-weight="700">leanctx.com</text>
</svg>"##,
        already = result.already_optimized,
    )
}

/// The top three missed commands as a single muted line. Empty when none.
fn before_card_top_commands(result: &DiscoverResult) -> String {
    if result.missed_commands.is_empty() {
        return String::new();
    }
    let joined = result
        .missed_commands
        .iter()
        .take(3)
        .map(|m| format!("{} {}x", m.prefix, m.count))
        .collect::<Vec<_>>()
        .join("    ·    ");
    format!(
        "  <text x=\"70\" y=\"556\" fill=\"#cbd5e1\" font-size=\"22\">top missed  {}</text>",
        xml_escape(&joined)
    )
}

/// Minimal XML text escaping for data-derived strings in the SVG card.
fn xml_escape(s: &str) -> String {
    s.replace('&', "&amp;")
        .replace('<', "&lt;")
        .replace('>', "&gt;")
        .replace('"', "&quot;")
        .replace('\'', "&apos;")
}

#[cfg(test)]
mod tests {
    use super::{analyze_history, analyze_history_with_stats, render_before_card};
    use crate::core::stats::{CommandStats, StatsStore};

    fn history() -> Vec<String> {
        vec![
            "git status".into(),
            "git diff".into(),
            "cargo build".into(),
            "cargo test".into(),
            "lean-ctx gain".into(),
            "vim notes.txt".into(),
        ]
    }

    fn stats_with(entries: &[(&str, u64, u64, u64)]) -> StatsStore {
        let mut s = StatsStore::default();
        for (key, count, input, output) in entries {
            s.commands.insert(
                (*key).to_string(),
                CommandStats {
                    count: *count,
                    input_tokens: *input,
                    output_tokens: *output,
                },
            );
            s.total_commands += count;
            s.total_input_tokens += input;
            s.total_output_tokens += output;
        }
        s
    }

    #[test]
    fn no_measured_data_yields_zero_potential_and_no_fabrication() {
        // Empty stats: detection still works, but NO token figure is invented.
        let r = analyze_history_with_stats(&history(), 20, &StatsStore::default());
        assert!(!r.has_measured_data, "no stats => no measured data");
        assert_eq!(r.potential_tokens, 0, "must not fabricate tokens");
        assert_eq!(r.potential_usd, 0.0, "must not fabricate dollars");
        assert!(
            r.missed_commands
                .iter()
                .all(|m| !m.measured && m.savings_range == ""),
            "unmeasured families show em dash, not a fake range"
        );
        // git (2) + cargo (2) recognised; vim ignored; lean-ctx already-optimized.
        assert_eq!(r.already_optimized, 1);
        assert!(
            r.missed_commands
                .iter()
                .any(|m| m.prefix == "git" && m.count == 2)
        );
    }

    #[test]
    fn measured_family_projects_real_savings() {
        // git measured at 90% savings (1000 in -> 100 out over 2 runs => 500 avg in).
        let store = stats_with(&[("git status", 2, 1000, 100)]);
        let r = analyze_history_with_stats(&history(), 20, &store);
        assert!(r.has_measured_data);
        let git = r
            .missed_commands
            .iter()
            .find(|m| m.prefix == "git")
            .expect("git present");
        assert!(git.measured);
        assert_eq!(git.savings_range, "90%", "real measured rate, not a guess");
        // 2 plain-shell git cmds * 500 avg_input * 0.9 = 900 projected.
        assert_eq!(git.estimated_tokens, 900);
        assert!(git.savings_range.ends_with('%'));
        // cargo has no measurement => contributes nothing.
        let cargo = r
            .missed_commands
            .iter()
            .find(|m| m.prefix == "cargo")
            .unwrap();
        assert_eq!(cargo.estimated_tokens, 0);
        assert_eq!(r.potential_tokens, 900);
    }

    #[test]
    fn newly_shipped_families_are_recognised() {
        // Regression guard: families added in #657-#661 must be discoverable.
        let hist: Vec<String> = ["dbt run", "trivy image nginx", "pulumi up", "jj log"]
            .into_iter()
            .map(String::from)
            .collect();
        let r = analyze_history_with_stats(&hist, 20, &StatsStore::default());
        for fam in ["dbt", "trivy", "pulumi", "jj"] {
            assert!(
                r.missed_commands.iter().any(|m| m.prefix == fam),
                "{fam} should be recognised as compressible"
            );
        }
    }

    #[test]
    fn before_card_is_well_formed_and_branded() {
        let result = analyze_history(&history(), 20);
        let svg = render_before_card(&result);
        assert!(svg.starts_with("<svg"), "must be an SVG document");
        assert!(svg.trim_end().ends_with("</svg>"), "must close the svg tag");
        assert!(svg.contains("leanctx.com"), "must carry the brand footer");
        assert!(svg.contains("Ghost Tokens"), "must frame the leak");
        assert!(
            svg.contains("tokens/month left on the table"),
            "headline label present"
        );
    }

    #[test]
    fn xml_escape_neutralizes_markup() {
        assert_eq!(super::xml_escape("a<b>&\"'"), "a&lt;b&gt;&amp;&quot;&apos;");
    }
}