search for standard sites
pub-search.waow.tech
search
zig
blog
atproto
improve search: OR ranking, URL decoding, lower default limit
- use OR between search terms with BM25 ranking for better recall
- decode + as space in query params (form-urlencoded)
- lower MCP search default limit from 20 to 5
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
+68
-13
+62
-11
backend/src/db/mod.zig
+62
-11
backend/src/db/mod.zig
···
410
410
return try output.toOwnedSlice();
411
411
}
412
412
413
-
/// Build FTS5 query with prefix on last word only: "cat dog" -> "cat dog*"
413
+
/// Build FTS5 query with OR between terms: "cat dog" -> "cat OR dog*"
414
+
/// Uses OR for better recall with BM25 ranking (more matches = higher score)
414
415
fn buildFtsQuery(alloc: Allocator, query: []const u8) ![]const u8 {
415
416
if (query.len == 0) return "";
416
417
417
-
// find actual content bounds (trim whitespace)
418
+
// normalize: trim whitespace
418
419
var start: usize = 0;
419
420
var end: usize = query.len;
420
421
while (start < end and query[start] == ' ') start += 1;
421
422
while (end > start and query[end - 1] == ' ') end -= 1;
423
+
if (start >= end) return "";
422
424
423
-
if (start >= end) return "";
425
+
const trimmed = query[start..end];
426
+
427
+
// count words and total length
428
+
var word_count: usize = 0;
429
+
var total_word_len: usize = 0;
430
+
var in_word = false;
431
+
for (trimmed) |c| {
432
+
const is_sep = (c == ' ' or c == '.');
433
+
if (is_sep) {
434
+
in_word = false;
435
+
} else {
436
+
if (!in_word) word_count += 1;
437
+
in_word = true;
438
+
total_word_len += 1;
439
+
}
440
+
}
424
441
425
-
// allocate: trimmed length + 1 for '*' at end
426
-
const trimmed_len = end - start;
427
-
const buf = try alloc.alloc(u8, trimmed_len + 1);
442
+
if (word_count == 0) return "";
428
443
429
-
// copy and normalize dots to spaces
430
-
for (query[start..end], 0..) |c, i| {
431
-
buf[i] = if (c == '.') ' ' else c;
444
+
// single word: just add prefix wildcard
445
+
if (word_count == 1) {
446
+
const buf = try alloc.alloc(u8, total_word_len + 1);
447
+
var pos: usize = 0;
448
+
for (trimmed) |c| {
449
+
if (c != ' ' and c != '.') {
450
+
buf[pos] = c;
451
+
pos += 1;
452
+
}
453
+
}
454
+
buf[pos] = '*';
455
+
return buf;
432
456
}
433
-
buf[trimmed_len] = '*';
434
457
435
-
return buf[0 .. trimmed_len + 1];
458
+
// multiple words: join with " OR ", prefix on last
459
+
// size = word chars + (n-1) * 4 for " OR " + 1 for "*"
460
+
const buf_len = total_word_len + (word_count - 1) * 4 + 1;
461
+
const buf = try alloc.alloc(u8, buf_len);
462
+
463
+
var pos: usize = 0;
464
+
var current_word: usize = 0;
465
+
in_word = false;
466
+
467
+
for (trimmed) |c| {
468
+
const is_sep = (c == ' ' or c == '.');
469
+
if (is_sep) {
470
+
if (in_word) {
471
+
// end of word - add " OR " if not last
472
+
current_word += 1;
473
+
if (current_word < word_count) {
474
+
@memcpy(buf[pos .. pos + 4], " OR ");
475
+
pos += 4;
476
+
}
477
+
}
478
+
in_word = false;
479
+
} else {
480
+
buf[pos] = c;
481
+
pos += 1;
482
+
in_word = true;
483
+
}
484
+
}
485
+
buf[pos] = '*';
486
+
return buf;
436
487
}
437
488
438
489
/// Find documents similar to a given document using vector similarity
+4
backend/src/server.zig
+4
backend/src/server.zig
···
117
117
const end = mem.indexOf(u8, encoded, "&") orelse encoded.len;
118
118
const query_encoded = encoded[0..end];
119
119
const buf = try alloc.dupe(u8, query_encoded);
120
+
// decode + as space (form-urlencoded), then percent-decode
121
+
for (buf) |*c| {
122
+
if (c.* == '+') c.* = ' ';
123
+
}
120
124
return std.Uri.percentDecodeInPlace(buf);
121
125
}
122
126
}
+2
-2
mcp/src/leaflet_mcp/server.py
+2
-2
mcp/src/leaflet_mcp/server.py
···
93
93
async def search(
94
94
query: str = "",
95
95
tag: str | None = None,
96
-
limit: int = 20,
96
+
limit: int = 5,
97
97
) -> list[SearchResult]:
98
98
"""search leaflet documents and publications.
99
99
···
103
103
args:
104
104
query: search query (searches titles and content)
105
105
tag: optional tag to filter by (only applies to documents)
106
-
limit: max results to return (default 20, max 40)
106
+
limit: max results to return (default 5, max 40)
107
107
108
108
returns:
109
109
list of search results with uri, title, snippet, and metadata