skeptrune

Complete configuration: ``` replicaCount: 4 image: repository: docker.io/qdrant/qdrant pullPolicy: IfNotPresent tag: "" useUnprivilegedImage: false imagePullSecrets: [] nameOverride: "" fullnameOverride: "" args: ["./config/initialize.sh"] env: {} # - name: QDRANT_ALLOW_RECOVERY_MODE # value:...

Collection config: ``` { "params": { "vectors": { "1536_vectors": { "size": 1536, "distance": "Cosine" } }, "shard_number": 4, "replication_factor": 2, "write_consistency_factor": 1, "on_disk_payload": true, "sparse_vectors": { "sparse_vectors": { "index": {...

payload schema: ``` { "tag_set": { "data_type": "text", "points": 0 }, "content": { "data_type": "text", "params": { "type": "text", "tokenizer": "prefix", "min_token_len": 2, "max_token_len": 10, "lowercase": true }, "points": 0...

> That is expected behavior. With that configuration, and indexing turned off, everything is kept in memory. Quantization is only applied at time of indexing. > > Instead you probably...

> > We actually want to use memory. We don't want to put things on disk. > > Note that with `on_disk: true` you'd still be using memory. It does...

> > Ideally there would be a way to leave the points on disk until the `optimizer` gets to it during ingest. > > For that you can use this...