srcwalk 0.1.5

Tree-sitter indexed lookups — smart code reading for AI agents
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

#!/usr/bin/env python3
"""
Analyze exploration patterns from benchmark results.
Extract tool call sequences and exploration efficiency metrics.
"""

import json
import os
from pathlib import Path
from collections import defaultdict
from typing import Dict, List, Tuple

def parse_jsonl(file_path: str) -> List[Dict]:
    """Parse JSONL file and return list of run results."""
    results = []
    with open(file_path, 'r') as f:
        for line in f:
            line = line.strip()
            if line:
                results.append(json.loads(line))
    return results

def calculate_ratios(run: Dict) -> Dict:
    """Calculate exploration efficiency metrics."""
    tool_calls = run.get('tool_calls', {})
    num_turns = run.get('num_turns', 0)
    num_tool_calls = run.get('num_tool_calls', 0)

    # Extract counts for different tool categories
    reads = 0
    searches = 0
    greps = 0
    maps = 0
    globs = 0

    for tool, count in tool_calls.items():
        if 'read' in tool.lower():
            reads += count
        if 'search' in tool.lower():
            searches += count
        if 'grep' in tool.lower():
            greps += count
        if 'map' in tool.lower():
            maps += count
        if 'glob' in tool.lower():
            globs += count

    # Calculate ratios
    ratios = {
        'reads_per_turn': reads / num_turns if num_turns > 0 else 0,
        'searches_per_turn': searches / num_turns if num_turns > 0 else 0,
        'greps_per_turn': greps / num_turns if num_turns > 0 else 0,
        'tools_per_turn': num_tool_calls / num_turns if num_turns > 0 else 0,
        'total_reads': reads,
        'total_searches': searches,
        'total_greps': greps,
        'total_maps': maps,
        'total_globs': globs,
    }

    # Infer likely first tool (if map=1 in srcwalk mode, it's likely first)
    first_tool = None
    if maps == 1 and run.get('mode') == 'srcwalk':
        first_tool = 'srcwalk_map'
    elif globs >= 1 and run.get('mode') == 'baseline':
        first_tool = 'Glob (likely)'

    ratios['likely_first_tool'] = first_tool

    return ratios

def analyze_file(file_path: str) -> Tuple[str, List[Dict]]:
    """Analyze a single benchmark file."""
    runs = parse_jsonl(file_path)
    filename = os.path.basename(file_path)

    analyzed_runs = []
    for run in runs:
        if 'error' in run:
            continue  # Skip errors/timeouts

        analysis = {
            'task': run.get('task'),
            'repo': run.get('repo'),
            'mode': run.get('mode'),
            'model': run.get('model'),
            'num_turns': run.get('num_turns'),
            'num_tool_calls': run.get('num_tool_calls'),
            'tool_calls': run.get('tool_calls', {}),
            'correct': run.get('correct'),
            'srcwalk_version': run.get('srcwalk_version'),
        }

        analysis.update(calculate_ratios(run))
        analyzed_runs.append(analysis)

    return filename, analyzed_runs

def compare_modes(runs: List[Dict]):
    """Compare baseline vs srcwalk exploration patterns."""
    baseline_runs = [r for r in runs if r['mode'] == 'baseline' and r['model'] == 'sonnet']
    srcwalk_runs = [r for r in runs if r['mode'] == 'srcwalk' and r['model'] == 'sonnet']

    def calc_avg(runs, key):
        values = [r[key] for r in runs if key in r]
        return sum(values) / len(values) if values else 0

    print("\n" + "="*80)
    print("BASELINE vs SRCWALK COMPARISON (Sonnet only)")
    print("="*80)

    print(f"\nBaseline runs: {len(baseline_runs)}")
    print(f"Srcwalk runs: {len(srcwalk_runs)}")

    print("\n--- Average Exploration Metrics ---")
    print(f"{'Metric':<30} {'Baseline':>15} {'Srcwalk':>15} {'Delta':>15}")
    print("-" * 80)

    metrics = [
        'tools_per_turn',
        'reads_per_turn',
        'searches_per_turn',
        'greps_per_turn',
        'num_turns',
        'num_tool_calls',
    ]

    for metric in metrics:
        baseline_avg = calc_avg(baseline_runs, metric)
        srcwalk_avg = calc_avg(srcwalk_runs, metric)
        delta = srcwalk_avg - baseline_avg
        print(f"{metric:<30} {baseline_avg:>15.2f} {srcwalk_avg:>15.2f} {delta:>15.2f}")

    # Tool preference analysis
    print("\n--- Tool Preference (Average per run) ---")
    print(f"{'Tool':<30} {'Baseline':>15} {'Srcwalk':>15}")
    print("-" * 80)

    baseline_reads = calc_avg(baseline_runs, 'total_reads')
    baseline_greps = calc_avg(baseline_runs, 'total_greps')
    baseline_globs = calc_avg(baseline_runs, 'total_globs')

    srcwalk_reads = calc_avg(srcwalk_runs, 'total_reads')
    srcwalk_searches = calc_avg(srcwalk_runs, 'total_searches')
    srcwalk_maps = calc_avg(srcwalk_runs, 'total_maps')

    print(f"{'Read operations':<30} {baseline_reads:>15.2f} {srcwalk_reads:>15.2f}")
    print(f"{'Search operations (Grep/Search)':<30} {baseline_greps:>15.2f} {srcwalk_searches:>15.2f}")
    print(f"{'Discovery (Glob/Map)':<30} {baseline_globs:>15.2f} {srcwalk_maps:>15.2f}")

    # Srcwalk map usage
    print("\n--- Srcwalk Map Usage ---")
    srcwalk_with_map = sum(1 for r in srcwalk_runs if r['total_maps'] > 0)
    print(f"Srcwalk runs starting with map: {srcwalk_with_map}/{len(srcwalk_runs)} ({100*srcwalk_with_map/len(srcwalk_runs):.1f}%)")

    # Search to read ratio
    print("\n--- Exploration Patterns ---")
    baseline_search_read_ratio = baseline_greps / baseline_reads if baseline_reads > 0 else 0
    srcwalk_search_read_ratio = srcwalk_searches / srcwalk_reads if srcwalk_reads > 0 else 0

    print(f"Baseline Grep:Read ratio: {baseline_search_read_ratio:.2f}:1")
    print(f"Srcwalk Search:Read ratio: {srcwalk_search_read_ratio:.2f}:1")

    return baseline_runs, srcwalk_runs

def print_detailed_runs(runs: List[Dict], title: str, limit: int = 5):
    """Print detailed information for individual runs."""
    print(f"\n{title}")
    print("="*80)

    for i, run in enumerate(runs[:limit], 1):
        print(f"\n{i}. {run['task']} ({run['repo']}) - {'✓' if run['correct'] else '✗'}")
        print(f"   Turns: {run['num_turns']}, Tool calls: {run['num_tool_calls']}")
        print(f"   Tools: {run['tool_calls']}")
        print(f"   Ratios: tools/turn={run['tools_per_turn']:.2f}, reads/turn={run['reads_per_turn']:.2f}, searches/turn={run['searches_per_turn']:.2f}")
        if run['likely_first_tool']:
            print(f"   Likely first tool: {run['likely_first_tool']}")

def main():
    """Main analysis function."""
    results_dir = Path("/Users/flysikring/conductor/workspaces/srcwalk/almaty/benchmark/results")

    # Focus on the two newest files
    target_files = [
        "benchmark_20260213_131246.jsonl",  # Previous srcwalk run
        "benchmark_20260213_135039.jsonl",  # New srcwalk run (MCP-only)
    ]

    all_runs = []

    for filename in target_files:
        file_path = results_dir / filename
        if file_path.exists():
            print(f"\nProcessing: {filename}")
            _, runs = analyze_file(str(file_path))
            all_runs.extend(runs)
            print(f"  Found {len(runs)} valid runs")

    # Compare modes
    baseline_runs, srcwalk_runs = compare_modes(all_runs)

    # Show detailed examples
    print_detailed_runs(baseline_runs, "\nDETAILED BASELINE EXAMPLES", limit=3)
    print_detailed_runs(srcwalk_runs, "\nDETAILED SRCWALK EXAMPLES", limit=3)

    # Task-specific comparison
    print("\n" + "="*80)
    print("TASK-SPECIFIC COMPARISON")
    print("="*80)

    tasks = set(r['task'] for r in all_runs)
    for task in sorted(tasks):
        task_baseline = [r for r in baseline_runs if r['task'] == task]
        task_srcwalk = [r for r in srcwalk_runs if r['task'] == task]

        if task_baseline and task_srcwalk:
            print(f"\n{task}:")

            # Compare both files for this task
            for b_run, t_run in zip(task_baseline, task_srcwalk):
                file_marker = "OLD" if "131246" in str(b_run.get('srcwalk_version', '')) else "NEW"
                print(f"  Baseline: turns={b_run['num_turns']}, tools={b_run['num_tool_calls']}, {b_run['tool_calls']}")
                print(f"  Srcwalk ({file_marker}): turns={t_run['num_turns']}, tools={t_run['num_tool_calls']}, {t_run['tool_calls']}")
                print(f"  Efficiency: Baseline={b_run['tools_per_turn']:.2f} tools/turn, Srcwalk={t_run['tools_per_turn']:.2f} tools/turn")

if __name__ == "__main__":
    main()