1 view

JustVugg/colibri

Run frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦

C★ 35,849+873 that day#11 on trending

Analyzed September 18, 2026 at a8f2ca6 (main) with jscpd 5.4.0 in 15.5s. Code duplication went from 6.54% on Sep 11 to 6.37% on Sep 18.

6.37%duplicated code · 7,644 of 120,085 lines
606clones in code
227code files of 674 scanned
5.23%duplicated over all files · 900 clones

The code number leaves out tests, fixtures, docs, examples, data files, styles, translations and vendored code, and 2 generated files: c/qwen38_nfc_tables.h, c/tok_unicode.h.

Largest clones in the code

  • 401 lines in 2 placesc4,187 tokens
    2204/* ---- begin include deepseek_v4_attention.c ---- */
    2205#include "deepseek_v4_internal.h"
    2206
    2207#include <math.h>
    2208#include <stdarg.h>
    2209#include <stdint.h>
    2210#include <stdio.h>
    2211#include <stdlib.h>
    2212#include <string.h>
    2213
    

    First 10 of 401 lines.

  • 294 lines in 2 placesc2,898 tokens
    3130#include "deepseek_v4_internal.h"
    3131
    3132#include <math.h>
    3133#include <stdarg.h>
    3134#include <stdint.h>
    3135#include <stdio.h>
    3136#include <stdlib.h>
    3137#include <string.h>
    3138
    3139#include "deepseek_v4_internal.h"
    

    First 10 of 294 lines.

  • 185 lines in 2 placesc1,910 tokens
    601#include <stdarg.h>
    602#include <stdio.h>
    603#include <stdlib.h>
    604#include <string.h>
    605
    606static int set_error(char *error, size_t size, const char *format, ...) {
    607    if (error && size) {
    608        va_list arguments;
    609        va_start(arguments, format);
    610        vsnprintf(error, size, format, arguments);
    

    First 10 of 185 lines.

7 more
Largest clones over all files, tests and data included
By format, code only
FormatFilesLinesClonesDuplicated linesDuplication
c2361,5334576,26110.18%
python10636,341677422.04%
c-header6317,231635072.94%
objectivec12,039181286.28%
tsx101,017160.59%
bash161,082000%
typescript6522000%
perl1174000%
powershell1146000%
By format, all files
FormatFilesLinesClonesDuplicated linesDuplication
c17986,6065987,6058.78%
python20461,9511701,9963.22%
c-header6736,149645141.42%
objectivec33,150282016.38%
yaml81,774121146.43%
json292,9701050517%
markup7731717924.49%
bash521,6516412.48%
markdown6315,0492120.08%
text1831523310.48%

Trending appearances

DayRankStarsFilesLinesClonesAll filesCodeHealthCommit
Sep 18, 2026#1135,849 (+873)674214,2739005.23%6.37%—a8f2ca6
Sep 17, 2026#335,169 (+1,546)674214,2739005.23%6.37%—a8f2ca6
Sep 16, 2026#234,107 (+2,026)674214,2739005.23%6.37%—a8f2ca6
Sep 15, 2026#132,433 (+2,173)674214,2739005.23%6.37%—f028d26
Sep 14, 2026#130,246 (+868)674214,2739005.23%6.37%—f028d26
Sep 11, 2026#927,535 (+98)626198,1198125.18%6.54%—fd93c41

Want this for your own project? Install jscpd and run jscpd ., or measure it the way the baselines were measured.