Adding GraphParams to be able to save graph parameters of index to SavedParams

[backurs]

This PR addresses the following issue:

We want to save alpha, l_build, backedge_ratio and vector_dtype somewhere and the best place to do it (in my opinion) is SavedParams.
For that we need to save GraphParams in BfTreeProviderParameters and in BfTreeProvider. This is what this PR does.

[Copilot]

Pull request overview

This PR adds GraphParams struct to persist graph configuration parameters (l_build, alpha, backedge_ratio, vector_dtype) alongside the BfTreeProvider index. The changes enable saving and loading these parameters as part of the index's SavedParams, allowing the DiskANNIndex configuration to be reconstructed on load.

Changes:

• Introduced GraphParams struct with fields for l_build, alpha, backedge_ratio, and vector_dtype
• Added optional graph_params field to BfTreeProvider, BfTreeProviderParameters, and SavedParams
• Updated save/load implementations to persist and restore graph_params
• Updated all tests and documentation examples to set graph_params: None

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
diskann-providers/src/model/graph/provider/async_/bf_tree/provider.rs	Added GraphParams struct, integrated graph_params field into BfTreeProvider/BfTreeProviderParameters/SavedParams, updated save/load logic, updated all doc examples and tests
diskann-providers/src/model/graph/provider/async_/bf_tree/mod.rs	Exported GraphParams in public API

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.00%. Comparing base (e5c83f5) to head (7dc182e).
⚠️ Report is 4 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #786      +/-   ##
==========================================
- Coverage   89.00%   89.00%   -0.01%     
==========================================
  Files         428      431       +3     
  Lines       78417    78455      +38     
==========================================
+ Hits        69795    69828      +33     
- Misses       8622     8627       +5     

Flag	Coverage Δ
miri	89.00% <ø> (-0.01%)	⬇️
unittests	89.00% <ø> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.
see 18 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[hildebrandmw]

Is this the data type as in f32/f16 etc.? It's probably better for it to be an enum instead of a raw string.

[backurs]

Good point. I changed the line to pub vector_dtype: VectorDtype, and added code

/// The element type of the full-precision vectors stored in the index.
#[derive(Serialize, Deserialize, Clone, Debug, PartialEq, Eq)]
#[serde(rename_all = "lowercase")]
pub enum VectorDtype {
    F32,
    F16,
    U8,
    I8,
}

impl VectorDtype {
    /// Derive the `VectorDtype` from a concrete [`VectorRepr`] type parameter.
    #[allow(clippy::panic)]
    pub fn from_type<T: VectorRepr>() -> Self {
        let name = std::any::type_name::<T>();
        match name {
            "f32" => Self::F32,
            "half::f16" | "f16" => Self::F16,
            "i8" => Self::I8,
            "u8" => Self::U8,
            _ => panic!("unsupported VectorRepr type: {}", name),
        }
    }
}

[hildebrandmw]

mmmm, this isn't the best approach. First std::any::type_name is not a stable representation and can change at any time. But more-so, almost any use of std::any::type_name or std::any::TypeId in patterns like this is almost always a bad idea. Nothing is preventing passing something like MinMax to this type and you only discover that it doesn't work at runtime.

Instead, I'd suggest something like

trait AsVectorDtype {
    const DATA_TYPE: VectorDtype;
}

and use that trait. Misuses will be caught at compile time, it's guaranteed not to break, and is much friendlier on the compiler to optimize.

[backurs]

Thank you for the suggestion and for describing the alternative! I added the following code:

pub trait AsVectorDtype {
    const DATA_TYPE: VectorDtype;
}

impl AsVectorDtype for f32 {
    const DATA_TYPE: VectorDtype = VectorDtype::F32;
}

impl AsVectorDtype for half::f16 {
    const DATA_TYPE: VectorDtype = VectorDtype::F16;
}

impl AsVectorDtype for i8 {
    const DATA_TYPE: VectorDtype = VectorDtype::I8;
}

impl AsVectorDtype for u8 {
    const DATA_TYPE: VectorDtype = VectorDtype::U8;
}

Is this what you meant by your comment?