zzstoatzz.io
semantic bufo search
find-bufo.com
bufo
add academic documentation and update landing page
- add comprehensive module documentation to search.rs explaining the
multimodal early fusion approach with references to voyage AI research
- update landing page to link "semantic search" directly to our search.rs
implementation on tangled
- remove outdated "multimodal hybrid search" link to turbopuffer docs
(no longer accurate since we removed BM25 hybrid search)
- simplify landing page header by removing the secondary info line
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
+23
src/search.rs
+23
src/search.rs
···
1
+
//! multimodal semantic search using early fusion embeddings
2
+
//!
3
+
//! this implementation uses voyage AI's multimodal-3 model which employs a
4
+
//! unified transformer encoder for early fusion of text and image modalities.
5
+
//!
6
+
//! ## approach
7
+
//!
8
+
//! - filename text (e.g., "bufo-jumping-on-bed" → "bufo jumping on bed") is combined
9
+
//! with image content in a single embedding request
10
+
//! - the unified encoder processes both modalities together, creating a single 1024-dim
11
+
//! vector that captures semantic meaning from both text and visual features
12
+
//! - vector search against turbopuffer using cosine distance similarity
13
+
//!
14
+
//! ## research backing
15
+
//!
16
+
//! voyage AI's multimodal-3 demonstrates 41.44% improvement on table/figure retrieval
17
+
//! tasks when combining text + images vs images alone, validating the early fusion approach.
18
+
//!
19
+
//! references:
20
+
//! - voyage multimodal embeddings: https://docs.voyageai.com/docs/multimodal-embeddings
21
+
//! - early fusion methodology: text and images are combined in the embedding generation
22
+
//! phase rather than fusing separate embeddings (late fusion)
23
+
1
24
use crate::config::Config;
2
25
use crate::embedding::EmbeddingClient;
3
26
use crate::turbopuffer::{QueryRequest, TurbopufferClient};
+8
-19
static/index.html
+8
-19
static/index.html
···
47
47
margin-bottom: 15px;
48
48
}
49
49
50
-
.info {
51
-
color: rgba(255, 255, 255, 0.85);
52
-
font-size: 0.9em;
53
-
text-align: center;
54
-
margin-top: 10px;
50
+
.subtitle a {
51
+
color: inherit;
52
+
text-decoration: underline;
53
+
text-decoration-color: rgba(255, 255, 255, 0.4);
54
+
transition: text-decoration-color 0.2s;
55
55
}
56
56
57
-
.info a {
58
-
color: rgba(255, 255, 255, 0.95);
59
-
text-decoration: none;
60
-
border-bottom: 1px solid rgba(255, 255, 255, 0.4);
61
-
transition: border-color 0.2s;
62
-
}
63
-
64
-
.info a:hover {
65
-
border-bottom-color: rgba(255, 255, 255, 0.9);
57
+
.subtitle a:hover {
58
+
text-decoration-color: rgba(255, 255, 255, 0.9);
66
59
}
67
60
68
61
.search-box {
···
210
203
<div class="container">
211
204
<div class="header">
212
205
<h1>find bufo</h1>
213
-
<p class="subtitle">semantic search for <a href="https://bufo.zone" target="_blank" style="color: inherit; text-decoration: underline;">bufo.zone</a></p>
214
-
<p class="info">
215
-
<a href="https://turbopuffer.com/docs/hybrid-search" target="_blank">multimodal hybrid search</a> ·
216
-
<a href="https://git.tangled.sh/zzstoatzz.io/find-bufo" target="_blank">source</a>
217
-
</p>
206
+
<p class="subtitle"><a href="https://git.tangled.sh/zzstoatzz.io/find-bufo/src/branch/main/src/search.rs" target="_blank">semantic search</a> for <a href="https://bufo.zone" target="_blank">bufo.zone</a></p>
218
207
</div>
219
208
220
209
<div class="search-box">