Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add hamzabellouch/agent-skills --skill vector-databases-qdrant-milvus-pineconegit clone --depth 1 https://github.com/hamzabellouch/agent-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/hamzabellouch/agent-skills/vector-databases-qdrant-milvus-pinecone)<a href="https://agentmods.dev/skills/hamzabellouch/agent-skills/vector-databases-qdrant-milvus-pinecone"><img src="https://agentmods.dev/badge/skills/hamzabellouch/agent-skills/vector-databases-qdrant-milvus-pinecone/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/hamzabellouch/agent-skills/vector-databases-qdrant-milvus-pinecone"><img src="https://agentmods.dev/badge/skills/hamzabellouch/agent-skills/vector-databases-qdrant-milvus-pinecone.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00074 | $0.02210 |
| Opus 5 | $0.00037 | $0.01105 |
| Sonnet 5 | $0.00015 | $0.00442 |
| Haiku 4.5 | $0.00007 | $0.00221 |
Grade A, and why
vector-databases-qdrant-milvus-pinecone scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 202 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Vector Databases Architect Skill: Qdrant, Milvus, & Pinecone
1. Architectural Taxonomy & Selection Matrix
| Feature / Criteria | Qdrant | Milvus | Pinecone |
|---|---|---|---|
| Deployment Model | Self-hosted (Rust, Single/Distributed) or Qdrant Cloud | Self-hosted (Go/C++, Cloud-Native K8s) or Zilliz Cloud | Fully Managed Serverless / Pods (SaaS) |
| Primary Indexing | In-Memory HNSW, On-Disk HNSW, Memmap Vectors | HNSW, IVF_FLAT, IVF_PQ, SCaNN, DiskANN | Proprietary Graph / Serverless Blob-backed |
| Quantization Support | Scalar (SQ8), Product (PQ), Binary (BQ) | Scalar (SQ8), Product (PQ), Binary | Handled internally in Serverless |
| Filter Engine | Native Payload Indexing (B-Tree, Keyword, Geo) | Dynamic Schema & Expression Parsing | Metadata Filtering (JSON-like) |
| Hardware Efficiency | Extremely low memory footprint via Memmap + Quantization | High-throughput distributed scaling, GPU acceleration | Pay-per-read/write scaling |
| Best Used For | Low-latency, cost-efficient self-hosted or hybrid cloud RAG | Enterprise scale (>100M+ vectors), distributed K8s, GPU search | Zero-Ops managed scaling, quick time-to-market |
2. Index Selection, Memory Estimation & Quantization Math
Indexing Mechanisms
- HNSW (Hierarchical Navigable Small World)
- m (Max Edges per node): Default 16. Higher values (32-64) improve recall for high-dimensional vectors (>1024d) at the cost of memory and build time.
- ef_construction: Default 100-200. Controls index build precision.
- ef_search: Dynamic search depth. Higher = higher recall, lower QPS.
- IVF (Inverted File Index)
- nlist: Number of cluster centroids (e.g., $\sqrt{N}$ to $4\sqrt{N}$).
- nprobe: Number of centroids queried during search.
- DiskANN / Vamana
- Stores vectors on NVMe SSD with in-memory compressed graph edges. Crucial for massive scale (>1B vectors) with constrained RAM.
Quantization Techniques
- Scalar Quantization (SQ8): Maps 32-bit floats (
float32) to 8-bit integers (int8). Reduces RAM by ~75% with minimal recall drop (<1%). - Product Quantization (PQ): Splits high-dim vector into $m$ sub-vectors and quantizes each into centroid IDs (
uint8). Reduces RAM by up to 90-95%, with minor accuracy tradeoff. - Binary Quantization (BQ): Quantizes positive floats to
1and negative to0(1 bit per dimension). 32x reduction in size and ultra-fast Hamming distance, best combined with dense re-ranking.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 10d ago First seen · 202 lines · 74 tokens per session scan A 7be59dca43c9
vector-databases-qdrant-milvus-pinecone is a skill published in the GitHub repository hamzabellouch/agent-skills (4 stars, last pushed 1mo ago), licensed MIT. It adds 74 tokens to every session and 2,210 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
neo4j
Neo4j graph database with Cypher query language. Use for graph-based data.
mongodb-search-and-ai
Guides MongoDB users through implementing and optimizing Atlas Search (full-text), Vector Search (semantic), and Hybrid Search solutions. Use this skill when users need to build search functionality for text-based queries (autocomplete, fuzzy matching, faceted search), semantic similarity (embeddings, RAG…
sqlite-vec-skilld
ALWAYS use when writing code importing "sqlite-vec". Consult for debugging, best practices, or modifying sqlite-vec, sqlite vec.
qdrant-clients-sdk
Qdrant provides client SDKs for various programming languages, allowing easy integration with Qdrant deployments.
neo4j-document-import-skill
Ingests unstructured and semi-structured documents into Neo4j as a knowledge graph. Use when chunking PDFs, HTML, plain text, or Markdown; extracting entities and relationships from text with an LLM (SimpleKGPipeline, neo4j-graphrag); loading JSON via apoc.load.json; building Document→Chunk→Entity graph structures; or…
neo4j-vector-index-skill
Create and manage Neo4j vector indexes, run vector similarity search (ANN/kNN), store embeddings on nodes or relationships, use SEARCH clause (Neo4j 2026.01+, preferred) or db.index.vector.queryNodes() procedure (deprecated 2026.04, still works on 2025.x), configure HNSW and quantization options, pick similarity…