Crate klieo_memory_qdrant 

Qdrant-backed implementation of klieo_core::memory::LongTermMemory.

MemoryQdrant::connect(url) is the capability-shaped default — connects to a Qdrant gRPC URL, ensures the per-scope collection exists with the embedder’s vector dimension, and returns an Arc<dyn LongTermMemory> ready to drop into klieo_core::AgentContext. Defaults to DummyEmbedder and a standard QdrantConfig.

For a custom QdrantConfig (API key, collection prefix, plaintext-remote opt-in) or a real Embedder, reach for MemoryQdrant::new directly.

Quickstart

use klieo_memory_qdrant::MemoryQdrant;

async fn example() {
    let mem = MemoryQdrant::connect("http://127.0.0.1:6334").await.unwrap();
    let _ = mem.long_term;
}

Advanced wiring — custom config + embedder

use klieo_memory_qdrant::{MemoryQdrant, QdrantConfig, DummyEmbedder};
use std::sync::Arc;

async fn example() {
    let cfg = QdrantConfig::new("http://qdrant.internal:6334")
        .with_api_key("supersecret")
        .with_collection_prefix("triage");
    let mem = MemoryQdrant::new(cfg, Arc::new(DummyEmbedder)).await.unwrap();
    let _ = mem.long_term;
}

What’s implemented

Only LongTermMemory. ShortTermMemory and EpisodicMemory are not Qdrant-shaped — see klieo-memory-sqlite and klieo-memory-neo4j for those.

Re-exports

pub use client::QdrantConfig;

pub use long_term::QdrantLongTerm;

pub use factory::MemoryQdrant;

Modules

client

Qdrant connection configuration + bootstrap helpers.

embedder

Re-exports of the shared Embedder trait + dummy/fake impls.

error

Error mapping from qdrant_client errors to klieo_core::error::MemoryError.

factory

MemoryQdrant — convenience factory wrapping the LongTermMemory handle in an Arc<dyn …> ready for AgentContext.

long_term

QdrantLongTerm — LongTermMemory over a single Qdrant collection with scope filtering on the payload.

Structs

DummyEmbedder

Default embedder that returns zero vectors.

FakeEmbedder

Test-only embedder that hashes each input text into a deterministic vector. Identical texts produce identical embeddings, so cosine recall behaves predictably under test.

Traits

Embedder

Compute embeddings for one or more texts.