basemind 0.2.1

Full AI context layer over MCP — tree-sitter code-map, document RAG (PDF/Office/HTML/email + OCR + reranker), shared agent memory, on-demand web crawl, git history + blame + per-symbol diff. 300+ languages, 8 coding-agent harnesses, content-addressed Fjall + LanceDB.
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

//! Shared embedding engine for the memory + documents MCP tools.

use anyhow::{Context, Result, anyhow};
use kreuzberg::embeddings::EMBEDDING_PRESETS;
use kreuzberg::{EmbeddingConfig, EmbeddingModelType};

/// Loaded, ready-to-query embedding engine. `Clone` is cheap (config is stack-only).
#[derive(Clone)]
pub struct SharedEmbedder {
    config: EmbeddingConfig,
    dim: u16,
    model_name: String,
}

impl SharedEmbedder {
    /// Build a `SharedEmbedder` from a named kreuzberg preset.
    pub fn load(preset: &str) -> Result<Self> {
        let meta = EMBEDDING_PRESETS
            .iter()
            .find(|p| p.name == preset)
            .ok_or_else(|| {
                anyhow!(
                    "unknown embedding preset '{preset}'; \
                     available: fast, balanced, quality, multilingual"
                )
            })?;
        let dim = u16::try_from(meta.dimensions).with_context(|| {
            format!(
                "preset '{preset}' dimension {} exceeds u16",
                meta.dimensions
            )
        })?;
        let config = EmbeddingConfig {
            model: EmbeddingModelType::Preset {
                name: preset.to_string(),
            },
            normalize: true,
            batch_size: 32,
            show_download_progress: false,
            cache_dir: None,
            acceleration: None,
            max_embed_duration_secs: Some(60),
        };
        Ok(Self {
            config,
            dim,
            model_name: preset.to_string(),
        })
    }

    /// Vector dimension produced by this embedder.
    pub fn dim(&self) -> u16 {
        self.dim
    }

    /// The preset name (e.g. `"balanced"`).
    pub fn model(&self) -> &str {
        &self.model_name
    }

    /// Embed a single text string. Returns a `Vec<f32>` of length `self.dim()`.
    pub fn embed(&self, text: &str) -> Result<Vec<f32>> {
        if text.is_empty() {
            return Err(anyhow!("embed: input text must not be empty"));
        }
        let mut results = kreuzberg::embeddings::embed_texts(&[text], &self.config)
            .with_context(|| format!("embed_texts(preset={})", self.model_name))?;
        results
            .pop()
            .ok_or_else(|| anyhow!("embed_texts returned empty result"))
    }
}