chadhas's picture
Upload folder using huggingface_hub
0cd7d65 verified
|
Raw History Blame Contribute Delete
4.44 kB
metadata
license: cc-by-4.0
language:
  - en
pretty_name: Qdrant documentation with multilingual-e5-small embeddings
task_categories:
  - text-retrieval
  - feature-extraction
size_categories:
  - 1K<n<10K
tags:
  - qdrant
  - hybrid-search

qdrant-docs-multilingual-e5-small-embeddings

This dataset contains the Qdrant documentation, courses, and articles from qdrant.tech, as of October 25, 2026, split into 3,593 chunks by page section. A section that fits in 480 tokens becomes one chunk, and larger sections are split further. Code examples are kept in Python only.

Each row holds a chunk's ID, its payload, and a 384-dimensional embedding:

Column Type Description
id string Point ID in the Qdrant collection
url string Page URL, relative to qdrant.tech
title string Page title
section string Headings above the chunk, joined with " > ". Empty for page intros
content_type string documentation, course, or article
has_code bool Whether the chunk contains a code block
text string A heading line, then the chunk in Markdown
vector list of float32 multilingual-e5-small embedding, 384 dimensions, cosine

Generation process

The embeddings were generated with the multilingual-e5-small model from the text field with fenced code blocks, Markdown images, and link URLs removed, using the passage: prefix. Embed your search queries with the query: prefix instead:

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("intfloat/multilingual-e5-small")
text = "Final Project: Production-Ready Documentation Search Engine > Build Steps > Step 3: Parse and Chunk Documents"
embeddings = model.encode([f"passage: {text}"], normalize_embeddings=True)
query_embedding = model.encode(
    ["how to chunk documentation"], prompt="query: ", normalize_embeddings=True
)

Using the dataset with Qdrant

The quickest way is to restore the ready-made snapshot, which includes the payload indexes. It requires Qdrant 1.20 or later:

PUT /collections/qdrant-docs/snapshots/recover
{
  "location": "https://snapshots.qdrant.io/qdrant-docs-2026-10-25.snapshot"
}

To load the rows yourself, create the collection with a vector named dense, upload the rows, then create the payload indexes. The full-text index on text uses BM25 scoring, which also requires Qdrant 1.20 or later:

from datasets import load_dataset
from qdrant_client import QdrantClient, models

client = QdrantClient(url="http://localhost:6333")
client.create_collection(
    "qdrant-docs",
    vectors_config={
        "dense": models.VectorParams(size=384, distance=models.Distance.COSINE)
    },
    # The collection is small, so lower the threshold to build the HNSW graph anyway.
    optimizers_config=models.OptimizersConfigDiff(indexing_threshold=1000),
)

dataset = load_dataset(
    "Qdrant/qdrant-docs-multilingual-e5-small-embeddings", split="train"
)
for batch in dataset.iter(batch_size=256):
    payload_fields = [name for name in batch if name not in ("id", "vector")]
    client.upsert(
        "qdrant-docs",
        points=[
            models.PointStruct(
                id=batch["id"][i],
                vector={"dense": batch["vector"][i]},
                payload={name: batch[name][i] for name in payload_fields},
            )
            for i in range(len(batch["id"]))
        ],
    )

for field in ("url", "content_type"):
    client.create_payload_index("qdrant-docs", field, models.PayloadSchemaType.KEYWORD)
client.create_payload_index("qdrant-docs", "has_code", models.PayloadSchemaType.BOOL)
client.create_payload_index(
    "qdrant-docs",
    "text",
    models.TextIndexParams(
        type=models.TextIndexType.TEXT,
        tokenizer=models.TokenizerType.CODE,
        lowercase=True,
        stemmer=models.SnowballParams(
            type=models.Snowball.SNOWBALL, language=models.SnowballLanguage.ENGLISH
        ),
        stopwords=models.Language.ENGLISH,
        scoring=models.TextScoringParams(type=models.TextScoringType.BM25),
    ),
)

Pass using="dense" when you query the vector.

License

The content belongs to Qdrant and is released under CC BY 4.0.