skills.rest
PALabs-v1
PALabs-v1OFFICIAL
Explore all skills by this author →
WebsiteGitHub

qdrant-memory-usage-optimization

Diagnoses and reduces Qdrant vector database memory usage through quantization and on-disk storage configuration.

3|1|Updated Nov 30, 2025
Search Tags:#vector database#qdrant#capacity planning#memory optimization#quantization#hnsw#on-disk storage
One-click install
npx skills add https://github.com/PALabs-v1/AI_friend --skill qdrant-memory-usage-optimization-palabs-v1
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: qdrant-memory-usage-optimization
Source: https://github.com/PALabs-v1/AI_friend/tree/main/.claude/skills/qdrant-performance-optimization/memory-usage-optimization
Command: npx skills add https://github.com/PALabs-v1/AI_friend --skill qdrant-memory-usage-optimization-palabs-v1

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Qdrant deployments can consume excessive RAM, causing out-of-memory crashes, growing resident memory, and confusion when actual usage doesn't match capacity calculations. This Skill helps diagnose whether memory pressure comes from resident memory or OS page cache and applies the right reduction techniques. ## Core Features & Use Cases - Memory Diagnosis: Distinguishes resident memory (RSSAnon) from OS page cache, interprets the /metrics endpoint, and flags resident memory above 80% of total RAM as a problem. - Footprint Reduction: Applies quantization, float16/int8/turbo4 datatypes, Matryoshka Representation Learning, and on-disk storage for HNSW indexes, payload indexes, and sparse vectors. - Version-Aware Configuration: Maps settings to the correct Qdrant version, using memory: pinned/cached/cold on 1.19+ and always_ram/on_disk booleans on 1.18 or older. - Use Case: A multi-tenant deployment keeps crashing during recovery. Use this Skill to move cold tenants' vectors and payload indexes to disk, enable quantization, and leave headroom for segment optimization. ## Quick Start Ask the assistant to diagnose why your Qdrant node's memory usage is so high and recommend configuration changes to reduce RAM consumption.

Frequently Asked Questions about qdrant-memory-usage-optimization

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I reduce Qdrant memory usage?▼

Reduce Qdrant memory by enabling quantization to keep only compressed vectors in RAM, using float16 or int8 datatypes for 2x or 4x savings, and moving rarely accessed components like payload indexes and sparse vectors to disk with on_disk or memory: cold settings.

Why is Qdrant using so much RAM?▼

Qdrant uses both resident memory for internal structures and OS page cache for disk reads. Page cache filling all RAM is normal, but resident memory above 80% of total RAM indicates a problem. Check the /metrics endpoint to distinguish the two.

When should I put the HNSW index on disk in Qdrant?▼

Put the HNSW index on disk when using fast local NVMe storage, in multi-tenant deployments where only some tenants are active, or with inline storage enabled. Frequent access to on-disk indexes causes significant latency degradation.

Does Qdrant memory configuration differ between versions?▼

Yes. Qdrant 1.19 and newer uses per-component memory tiers (pinned, cached, cold), while 1.18 and older uses always_ram and on_disk booleans on vectors, HNSW config, sparse indexes, and payload indexes.

How much memory headroom does Qdrant need for optimization?▼

During optimization, Qdrant fully loads optimized segments into RAM, so you must leave headroom beyond steady-state usage. The larger the max_segment_size setting, the more extra memory is required during optimization runs.

Related Skills

Explore More Software Engineering Skills→
githubgithub
38.5k

qdrant-monitoring-debugging

Diagnose Qdrant production issues using metrics, telemetry, and optimizer status endpoints.

Official
Intermediate
lucifertrjlucifertrj
17

qdrant-horizontal-scaling

Diagnose Qdrant capacity and guide horizontal scaling decisions.

Community
Advanced
lucifertrjlucifertrj
17

qdrant-memory-usage-optimization

Diagnose and optimize Qdrant deployment memory usage with quantization and on-disk storage.

Community
Advanced
summarybotngsummarybotng
2

exploratory-testing-advanced

Structure exploratory testing sessions with SBTM charters and heuristics.

Official
Advanced