id stringlengths 36 36 | url stringclasses 504
values | title stringclasses 475
values | section stringlengths 0 148 | content_type stringclasses 3
values | has_code bool 2
classes | text stringlengths 147 22.5k | vector listlengths 384 384 |
|---|---|---|---|---|---|---|---|
3449a332-143e-59e7-a5b0-9fc5ee3334ba | /documentation/tutorials-search-engineering/sparse-embeddings-ecommerce-part-2/ | Fine-Tuning Sparse Embeddings for E-Commerce Search | Part 2: Training SPLADE on Modal | Running Training | documentation | true | Fine-Tuning Sparse Embeddings for E-Commerce Search | Part 2: Training SPLADE on Modal > Running Training
With everything in place, launch training:
```bash
# Quick test run (100K samples)
uv run modal run modal_app.py \
--config-path configs/splade_standard.yaml \
--mode train
# Full dataset (set max_sample... | [
0.06048942729830742,
-0.045256104320287704,
-0.018198158591985703,
-0.06517543643712997,
0.03218454867601395,
-0.05122962221503258,
0.06533382833003998,
0.03620770201086998,
0.0541207417845726,
-0.023028893396258354,
0.036450594663619995,
0.0425407849252224,
0.03535911440849304,
-0.0414240... |
3452a721-c999-5516-baf8-e454134884cb | /documentation/private-cloud/api-reference/ | API Reference | qdrant.io/v1 > Resource Types > HelmRelease | documentation | false | API Reference > qdrant.io/v1 > Resource Types > HelmRelease
_Appears in:_
- [QdrantCloudRegionSpec](#qdrantcloudregionspec)
| Field | Description | Default | Validation |
| --- | --- | --- | --- |
| `markedForDeletionAt` _string_ | MarkedForDeletionAt specifies the time when the helm release was marked for deletion |... | [
0.04907340183854103,
-0.04995526745915413,
-0.0015432956861332059,
-0.06126890704035759,
0.07061435282230377,
-0.03375430777668953,
0.00787606555968523,
0.06129695475101471,
0.02260950766503811,
-0.062083374708890915,
0.05964155122637749,
0.031109659001231194,
0.0724785178899765,
-0.019837... |
3457489e-6e0b-5344-8095-5feb40bc4624 | /course/essentials/day-7/tensorlake/ | Integrating with Tensorlake | TensorLake Knowledge Graph Integration > Key Benefits | course | false | Integrating with Tensorlake > TensorLake Knowledge Graph Integration > Key Benefits
TensorLake's integration with Qdrant provides several advantages:
- **Enhanced Accuracy**: More accurate embeddings through structured data extraction
- **Complete Document Understanding**: Preserves document hierarchy and relationshi... | [
0.027946164831519127,
-0.039233237504959106,
-0.07670921832323074,
0.013156813569366932,
0.07211724668741226,
-0.04452945291996002,
0.07847422361373901,
0.042826540768146515,
0.05539686605334282,
-0.009266054257750511,
0.010349276475608349,
0.014251040294766426,
0.060378506779670715,
0.024... |
346b61bd-9c7f-5c44-95a6-a1dcd487c78b | /documentation/ops-configuration/administration/ | Administration | documentation | false | Administration
Qdrant exposes administration tools which enable to modify at runtime the behavior of a qdrant instance without changing its configuration manually. | [
0.03902088478207588,
-0.06443793326616287,
-0.0745931938290596,
-0.03997012600302696,
0.03320927917957306,
-0.016108835116028786,
0.059487856924533844,
0.054269324988126755,
0.08935187757015228,
0.009182889014482498,
0.03296913206577301,
0.03187764436006546,
0.09704191982746124,
-0.0499684... | |
348bd9bd-db81-52d3-bbe1-9a3ebc6f9315 | /documentation/examples/ | Build Prototypes | End-to-End Code Samples | documentation | false | Build Prototypes > End-to-End Code Samples
The following guided samples help you get started with real-world projects using Qdrant and other ecosystem tools.
| Guided Sample | Description | Stack ... | [
0.026983967050909996,
-0.029656605795025826,
-0.0644446462392807,
0.0050452402792871,
0.08260475844144821,
-0.04705863073468208,
0.07344180345535278,
0.07608567923307419,
0.05657950043678284,
-0.06720952689647675,
0.037787530571222305,
0.022815171629190445,
0.034593366086483,
-0.0472992137... |
348dd8ee-0c65-5ef9-a874-1d6fc1972ca7 | /documentation/examples/rag-contract-management-stackit-cohere/ | Region-Specific Contract Management System | Prerequisites | documentation | true | Region-Specific Contract Management System > Prerequisites
### Cohere account
Since you will be using Cohere's models, sign up on the [Cohere dashboard](https://dashboard.cohere.com/api-keys) and create an API key. Once you have it ready, store it as an environment variable:
```shell
export COHERE_API_KEY="<your-tok... | [
0.027923699468374252,
-0.023246068507432938,
-0.020982423797249794,
-0.0438896045088768,
0.0356421023607254,
-0.06797997653484344,
0.06058134883642197,
0.04119139164686203,
0.03585422411561012,
-0.027778536081314087,
0.005333538167178631,
0.017597176134586334,
0.0315965935587883,
-0.041583... |
34a906ae-a154-5521-9fed-1fc6d5a1d5b5 | /articles/vector-similarity-beyond-search/ | Vector Similarity: Going Beyond Full-Text Search | Dissimilarity Search | article | false | Vector Similarity: Going Beyond Full-Text Search > Dissimilarity Search
Dissimilarity search, or farthest search, is the simplest step beyond nearest-neighbor search, and full-text search can't reproduce it.
It aims to find the most dissimilar or distant documents across the collection.
 to help increase the relevance of vector search results. **It is built for scale, cheap, customizable, and universal.**
**It is cheap to use**, as... | [
0.040458451956510544,
-0.052316758781671524,
-0.05079888552427292,
-0.06993979215621948,
0.05163147673010826,
-0.031303368508815765,
0.043535273522138596,
0.03643462434411049,
0.07291065901517868,
-0.019877886399626732,
0.014040535315871239,
0.011901547200977802,
0.07964786887168884,
-0.02... |
34ea457b-ac64-58b3-b2ed-e3715cd7e9bb | /course/essentials/day-1/distance-metrics/ | Distance Metrics | course | false | Distance Metrics
Day 1
After vectors are stored, we can use their spatial properties to perform [nearest neighbor searches](/documentation/search/search/) that retrieve semantically similar items based on how close they are in this space.
The position of a vector in embedding space only reflects meaning as far as th... | [
0.07023819535970688,
-0.028806425631046295,
-0.014180322177708149,
-0.029423987492918968,
0.06320903450250626,
-0.04592097923159599,
0.024633284658193588,
0.03090992197394371,
0.07788579165935516,
-0.005074876360595226,
0.02412799745798111,
0.01628543995320797,
0.06663520634174347,
0.01417... | |
350ed2e5-71f3-51dc-8e90-69ce1ba91f88 | /documentation/private-cloud/api-reference/ | API Reference | qdrant.io/v1 > Resource Types > QdrantConfigurationService | documentation | false | API Reference > qdrant.io/v1 > Resource Types > QdrantConfigurationService
_Appears in:_
- [QdrantConfiguration](#qdrantconfiguration)
| Field | Description | Default | Validation |
| --- | --- | --- | --- |
| `api_key` _[QdrantSecretKeyRef](#qdrantsecretkeyref)_ | ApiKey for the qdrant instance | | Optional: \{\} ... | [
0.04828440025448799,
-0.018946388736367226,
-0.03379066661000252,
-0.03218597173690796,
0.03498481214046478,
-0.034397777169942856,
0.04921872168779373,
0.07258002460002899,
0.05330439656972885,
-0.03758286312222481,
0.02586994878947735,
0.02799336053431034,
0.06481529027223587,
-0.0325566... |
350fbca8-cd57-5c3b-851a-29b4387b7fad | /documentation/manage-data/vectors/ | Vectors | Quantization | documentation | false | Vectors > Quantization
Apart from changing the datatype of the original vectors, Qdrant can create quantized representations of vectors alongside the original ones.
This quantized representation can be used to quickly select candidates for rescoring with the original vectors or even used directly for search.
Quantiza... | [
0.05069160833954811,
-0.07140970230102539,
-0.01920999214053154,
-0.048010095953941345,
0.07610592991113663,
-0.02078814059495926,
0.07151739299297333,
0.031162213534116745,
0.07686767727136612,
-0.03701001778244972,
0.044640976935625076,
0.03177449479699135,
0.06736785918474197,
-0.037841... |
3523feea-c05f-582c-bd61-21aad55d1f61 | /articles/discovery-search/ | Discovery Search in Qdrant | Real-world discovery search applications | article | false | Discovery Search in Qdrant > Real-world discovery search applications
Let's talk about the first case: context with a target.
To understand why this is useful, let's take a look at a real-world example: using a multimodal encoder like [CLIP](https://openai.com/blog/clip/) to search for images, from text __and__ image... | [
0.043074652552604675,
-0.030649419873952866,
-0.060144126415252686,
0.004131758119910955,
0.08260339498519897,
-0.014683504588901997,
0.05579022690653801,
0.027512049302458763,
0.029491132125258446,
-0.0020022359676659107,
0.0021872580982744694,
0.031288791447877884,
0.06583640724420547,
-... |
352f92bf-c5bd-59b4-8964-fc5f29c9a1d2 | /course/beginners/module-3/the-two-families-of-search/ | The Two Families of Search | Dense Search | course | true | The Two Families of Search > Dense Search
Module 3
Every retrieval system is built from one or both of these.

A dense vector has a small, fixed number of dimensions, 384 for the model used here, a... | [
0.010768857784569263,
-0.012301736511290073,
-0.08774552494287491,
-0.023076653480529785,
0.06662077456712723,
-0.027256052941083908,
0.043333087116479874,
0.09074433892965317,
0.06596964597702026,
-0.0132217388600111,
0.008288993500173092,
0.040292784571647644,
0.062234457582235336,
0.016... |
35331e8f-780e-5f4e-b565-2cde02e2734e | /course/multi-vector-search/ | Multi-Vector Search Course | How the course works | course | false | Multi-Vector Search Course > How the course works
- icon: /icons/outline/training-purple.svg
title: Video-first lessons
content: Clear, concise modules by the Qdrant team
- icon: /icons/outline/hacker-purple.svg
title: Final project
content: Build a production-ready multi-modal search system
- icon: /icons/out... | [
0.033582545816898346,
-0.05749461054801941,
-0.0724143534898758,
-0.01790192537009716,
0.07938668131828308,
0.01445381436496973,
0.06534672528505325,
0.04956147447228432,
0.04463700205087662,
-0.007145623676478863,
0.01833093725144863,
0.05195390805602074,
0.048925258219242096,
-0.04675632... |
35365d36-db7a-50af-8193-c0607c1bb3ed | /course/essentials/day-6/congratulations/ | Course Completion and Next Steps | What's Next? | course | false | Course Completion and Next Steps > What's Next?
**Explore Advanced Integrations**: Check out [Day 7 Partner Integrations](/course/essentials/day-7/) to see how Qdrant works with leading AI frameworks and data platforms.
**Join the Community**: Share your final project results and connect with other practitioners buil... | [
0.03138747811317444,
-0.05555831640958786,
-0.032026506960392,
-0.008198915049433708,
0.08070651441812515,
-0.04081038385629654,
0.06089169904589653,
0.020663553848862648,
0.06327783316373825,
-0.047287434339523315,
0.0211963951587677,
0.004226834513247013,
0.028664667159318924,
-0.0589707... |
3553d5d6-6026-595e-aa50-c895b6264572 | /articles/rag-is-dead/ | Is RAG Dead? Why Long Context Windows Don't Replace RAG | Sources | article | false | Is RAG Dead? Why Long Context Windows Don't Replace RAG > Sources
- Liu et al., ["Lost in the Middle: How Language Models Use Long Contexts"](https://arxiv.org/abs/2307.03172), Transactions of the Association for Computational Linguistics, 2024.
- Hsieh et al., ["RULER: What's the Real Context Size of Your Long-Contex... | [
0.033386923372745514,
-0.010474243201315403,
-0.04893924668431282,
-0.056092895567417145,
0.07159475237131119,
-0.045409489423036575,
0.03838428482413292,
0.0592518113553524,
0.062149059027433395,
-0.061521224677562714,
0.032444458454847336,
-0.03170127794146538,
0.047871094197034836,
-0.0... |
357d0d02-c959-51d9-afc4-4dfa4243d82e | /documentation/fastembed/fastembed-splade/ | Working with SPLADE | Design choices | documentation | false | Working with SPLADE > Design choices
- The weights are not normalized. This means that the sum of the weights is not 1 or 100. This is a common practice in sparse embeddings, as it allows the model to capture the importance of each token in the document.
- Tokens are included in the sparse vector only if they are pres... | [
0.029229845851659775,
-0.05296582728624344,
-0.05139828473329544,
-0.06607440859079361,
0.06563499569892883,
-0.02281571924686432,
0.04938555136322975,
0.042075254023075104,
0.04426039382815361,
-0.02724967896938324,
0.034928202629089355,
0.03291161730885506,
0.0567176416516304,
-0.0064247... |
35a2f33b-2078-5712-9816-bcc860866fc5 | /documentation/tutorials-operations/data-streaming-kafka-qdrant/ | Kafka Streaming into Qdrant | Installation of MongoDB | documentation | true | Kafka Streaming into Qdrant > Installation of MongoDB
For Kafka to use MongoDB as a source, your MongoDB instance must run as a replica set, because the source connector reads changes from the oplog through change streams. The [Atlas CLI](https://www.mongodb.com/docs/atlas/cli/current/) creates a local single-node rep... | [
0.023781590163707733,
-0.0711197704076767,
-0.07143780589103699,
-0.027392398566007614,
0.04679583013057709,
-0.037037525326013565,
0.0728750228881836,
0.05892251431941986,
-0.018728051334619522,
-0.011503780260682106,
0.025203730911016464,
0.008153538219630718,
-0.0027837620582431555,
-0.... |
35aa56a7-c614-5c75-a4da-ea89aadaf103 | /documentation/search-tuning/hybrid-search/ | Hybrid Search in Qdrant | Dense and Sparse Retrieval Miss Different Things | documentation | false | Hybrid Search in Qdrant > Dense and Sparse Retrieval Miss Different Things
Dense retrieval embeds the query and each document, then ranks the documents by vector similarity. The model can place paraphrases near each other, but exact strings may lose influence among documents with similar meanings.
Sparse retrieval re... | [
0.03684361279010773,
-0.04590902477502823,
-0.02378157339990139,
-0.06073871999979019,
0.0778408795595169,
-0.057458583265542984,
0.06513067334890366,
0.029102889820933342,
0.07058244198560715,
-0.03417691960930824,
0.0026525058783590794,
0.03645392507314682,
0.035954322665929794,
-0.01443... |
35ade9d8-4dd6-5f90-85d8-aa78613c4795 | /articles/predicting-weak-retrieval/ | Predicting Weak Retrieval Without an LLM | Adapt This to Your Corpus | article | false | Predicting Weak Retrieval Without an LLM > Adapt This to Your Corpus
The winners here are not portable; the recipe is:
1. **Define the window** and the evidence each query needs.
2. **Label** weak retrievals on a calibration split.
3. **Benchmark** the cheap signals; keep what separates, drop the redundant.
4. **Thre... | [
0.021834686398506165,
-0.0673745796084404,
-0.05767566338181496,
-0.04926513507962227,
0.07658744603395462,
-0.007694059982895851,
0.03169892728328705,
0.013208959251642227,
0.059554778039455414,
-0.01100151427090168,
0.005395124666392803,
-0.015477912500500679,
0.061970632523298264,
-0.00... |
35b019a3-08ba-5e92-b05e-d5e47e7f91a3 | /documentation/private-cloud/api-reference/ | API Reference | qdrant.io/v1 > Resource Types > RegionCapabilities | documentation | false | API Reference > qdrant.io/v1 > Resource Types > RegionCapabilities
_Appears in:_
- [QdrantCloudRegionStatus](#qdrantcloudregionstatus)
| Field | Description | Default | Validation |
| --- | --- | --- | --- |
| `volumeSnapshot` _boolean_ | VolumeSnapshot specifies whether the Kubernetes cluster supports volume snapsho... | [
0.04815103858709335,
-0.03382926434278488,
0.00017097368254326284,
-0.08717386424541473,
0.06375013291835785,
-0.030794091522693634,
0.07885629683732986,
0.056091804057359695,
0.05756225064396858,
-0.04090512916445732,
0.05199746787548065,
0.015462816692888737,
0.05300494655966759,
-0.0047... |
35ce38a3-17ea-552d-bb33-cd8721ccf0b6 | /course/essentials/day-4/rescoring-oversampling-indexing/ | Accuracy Recovery with Rescoring | Implementation | course | true | Accuracy Recovery with Rescoring > Implementation
```python
from qdrant_client import QdrantClient, models
import os
client = QdrantClient(url=os.getenv("QDRANT_URL"), api_key=os.getenv("QDRANT_API_KEY"))
# For Colab:
# from google.colab import userdata
# client = QdrantClient(url=userdata.get("QDRANT_URL"), api_key... | [
0.03098168410360813,
-0.036306530237197876,
-0.035388682037591934,
-0.05357950180768967,
0.0714249461889267,
-0.005975449923425913,
0.027423366904258728,
-0.046306997537612915,
0.04765484109520912,
-0.0033562490716576576,
0.0523676872253418,
0.008384407497942448,
0.08563557267189026,
-0.03... |
35dbe23d-9762-528c-851e-3aec4787b047 | /course/essentials/day-1/embedding-models/ | Points, Vectors and Payloads | Common Embedding Sources > 3. On-Premise, Customizable: Open Source Models | course | false | Points, Vectors and Payloads > Common Embedding Sources > 3. On-Premise, Customizable: Open Source Models
Libraries like [Sentence Transformers](https://www.sbert.net/) give you access to thousands of open-source models on the [Hugging Face Hub](https://huggingface.co/models?pipeline_tag=sentence-similarity). This app... | [
0.01868516206741333,
-0.04151149094104767,
-0.03960433974862099,
-0.026053020730614662,
0.05422027036547661,
-0.011750034987926483,
0.05595167353749275,
0.005585212726145983,
0.06603960692882538,
-0.04151403158903122,
0.030762886628508568,
0.0019136791815981269,
0.04580114781856537,
-0.031... |
35f7c1cc-8f5b-5e23-80de-8e70805ac934 | /documentation/migrate-to-qdrant/from-weaviate/ | From Weaviate | Run the Migration > All Weaviate-Specific Flags | documentation | false | From Weaviate > Run the Migration > All Weaviate-Specific Flags
| Flag | Required | Description |
| :--- | :--- | :--- |
| `--weaviate.host` | Yes | Weaviate host address |
| `--weaviate.scheme` | No | `http` or `https` (default: `http`) |
| `--weaviate.class-name` | Yes | Weaviate class to migrate |
| `--weaviate.aut... | [
0.025691905990242958,
-0.0377531498670578,
-0.047555211931467056,
-0.03461761027574539,
0.049982983618974686,
-0.09297849237918854,
0.04378915950655937,
0.01871759444475174,
0.02808487042784691,
-0.05931692197918892,
0.01987314783036709,
0.036137793213129044,
0.039271533489227295,
-0.00755... |
36138e41-dcc2-5a05-b20e-1a361c0d7733 | /documentation/manage-data/vectors/ | Vectors | Datatypes > Float32 | documentation | false | Vectors > Datatypes > Float32
This is the default datatype for vectors in Qdrant. It is a 32-bit (4 bytes) floating-point number.
The standard OpenAI embedding of 1536 dimensionality will require 6KB of memory to store in `float32`.
You don't need to explicitly specify this datatype for vectors in Qdrant, as it is s... | [
0.05876004323363304,
-0.03256746008992195,
-0.0241059772670269,
-0.04409565031528473,
0.036403656005859375,
-0.0518675334751606,
0.011399022303521633,
0.09023996442556381,
0.09650660306215286,
-0.030956456437706947,
0.040668830275535583,
-0.02572673372924328,
0.0451318621635437,
-0.0235729... |
3618f5e5-851b-51d0-b47f-f12b9e867e1c | /documentation/tutorials-build-essentials/agentic-rag-crewai-zoom/ | Agentic RAG with CrewAI | Usage > 2. Launch the Interface | documentation | true | Agentic RAG with CrewAI > Usage > 2. Launch the Interface
The [`streamlit_app.py`](https://github.com/qdrant/examples/blob/master/agentic_rag_zoom_crewai/vector/streamlit_app.py) is located in the `vector` folder. To launch it, run:
```bash
streamlit run vector/streamlit_app.py
```
When you run this script, you will... | [
0.022073617205023766,
-0.05481792613863945,
-0.034173525869846344,
-0.007362631149590015,
0.09108331054449081,
-0.042095594108104706,
0.05036012828350067,
0.05550619587302208,
0.016278093680739403,
-0.040482454001903534,
0.04167422652244568,
-0.02911117486655712,
0.05690476670861244,
0.017... |
362c7ff3-7b68-5fe5-ab36-14deb093d896 | /documentation/tutorials-search-engineering/food-discovery-recommendations/ | Multimodal Food Search Demo | Different search modes | documentation | false | Multimodal Food Search Demo > Different search modes
The FastAPI backend [exposes just a single endpoint](https://github.com/qdrant/demo-food-discovery/blob/6b49e11cfbd6412637d527cdd62fe9b9f74ac699/backend/main.py#L37),
however it handles multiple scenarios. Let's dive into them one by one and understand why they are... | [
0.001782282255589962,
-0.04436502233147621,
-0.04532995447516441,
-0.028540313243865967,
0.06066253036260605,
-0.010094289667904377,
0.06563872843980789,
0.03164779394865036,
0.05176536738872528,
0.010126973502337933,
0.00594124523922801,
0.062417224049568176,
0.06141732633113861,
-0.02964... |
3635b1d9-9588-5ba0-aaa5-75ece7b1a61f | /articles/sparse-vectors/ | Understanding SPLADE and Sparse Vectors | What are sparse and dense vectors? | article | true | Understanding SPLADE and Sparse Vectors > What are sparse and dense vectors?
Sparse vectors are like the Marie Kondo of data: they keep only what sparks joy (or relevance, in this case).
Consider a simplified example of 2 documents, each with 200 words. A dense vector would have several hundred non-zero values, where... | [
0.024898728355765343,
-0.015310089103877544,
-0.0424075573682785,
-0.06706762313842773,
0.06779949367046356,
-0.04892852157354355,
0.06541121751070023,
0.039365846663713455,
0.07850822806358337,
-0.011160730384290218,
0.016152948141098022,
0.034225307404994965,
0.03697400540113449,
-0.0062... |
3666ecd0-2371-54ef-9f3c-16e53c15830a | /course/essentials/day-7/jina/ | Integrating with Jina AI | Jina AI Multimodal Embeddings > Real-World Applications | course | false | Integrating with Jina AI > Jina AI Multimodal Embeddings > Real-World Applications
This architecture enables various sophisticated use cases:
- **Content Discovery**: Multimodal search across text and image libraries
- **E-commerce**: Product search using both text descriptions and visual features
- **Research Platfo... | [
0.014937425963580608,
-0.03906122222542763,
-0.07317997515201569,
-0.00674058124423027,
0.08674640953540802,
-0.005235580727458,
0.032119277864694595,
0.03407230228185654,
0.03343847021460533,
-0.016841014847159386,
0.042905449867248535,
-0.031298212707042694,
0.07611872255802155,
-0.01468... |
367280ac-513d-52df-99a6-4a668b5600d8 | /documentation/inference/external-inference-providers/ | External Providers | OpenAI | documentation | true | External Providers > OpenAI
When you prepend a model name with `openai/`, the embedding request is automatically routed to the [OpenAI Embeddings API](https://platform.openai.com/docs/guides/embeddings).
For example, to use OpenAI's `text-embedding-3-large` model when ingesting data, prepend the model name with `open... | [
0.03833693265914917,
-0.03356674686074257,
0.011812429875135422,
-0.06752429902553558,
0.060716819018125534,
-0.034730859100818634,
0.04570852592587471,
0.09700415283441544,
0.03995930030941963,
-0.030139664188027382,
0.03202028200030327,
0.025709118694067,
0.038550861179828644,
-0.0372071... |
36862d9e-f42f-57bd-af33-497b647dd929 | /documentation/interfaces/ | API & SDKs | API Reference | documentation | true | API & SDKs > API Reference
All interaction with Qdrant takes place via the REST API. We recommend using REST API if you are using Qdrant for the first time or if you are working on a prototype.
| API | Documentation |
| -------- | -----------... | [
0.0006047922652214766,
-0.05317191407084465,
-0.023505980148911476,
-0.008315184153616428,
0.05848054587841034,
-0.043551769107580185,
0.08452766388654709,
0.08401474356651306,
0.06835976988077164,
-0.014978214167058468,
0.035955678671598434,
0.024620765820145607,
0.012976703234016895,
-0.... |
3690eab4-0f0d-56b4-aaa7-32d8cd9ce679 | /articles/relevance-feedback/ | Relevance Feedback in Qdrant | Our Idea > Using Entire Vector Space | article | false | Relevance Feedback in Qdrant > Our Idea > Using Entire Vector Space
So the idea is to use feedback to define a more relevant direction (**closer** to this, **further** from that) within the vector space spanned by the dataset.
That implies warping the notion of "closer" and "further", the distance (or similarity) **s... | [
0.05728965997695923,
-0.030700579285621643,
-0.05227501690387726,
-0.03978786617517471,
0.07785912603139877,
0.013401004485785961,
0.04842602089047432,
0.019793052226305008,
0.057439401745796204,
-0.005697976797819138,
0.033174995332956314,
0.006750867702066898,
0.10826048254966736,
0.0053... |
36bf716a-bd13-5d69-891a-a55f2fe5bcde | /articles/turboquant-quantization/ | TurboQuant in Qdrant | What Qdrant Adds | article | false | TurboQuant in Qdrant > What Qdrant Adds
TurboQuant is not the only rotation-based vector quantization algorithm in this design space. **[RaBitQ (Gao & Long, SIGMOD 2024)](https://arxiv.org/abs/2405.12497)** also develops the same rotate-then-quantize foundation, with different implementation details. Qdrant borrows th... | [
0.029749805107712746,
-0.03875705227255821,
-0.05198447033762932,
-0.052113618701696396,
0.01057319063693285,
-0.004909529350697994,
0.08624529093503952,
0.022563498467206955,
0.05944420397281647,
-0.03819238767027855,
0.07501059025526047,
0.008692318573594093,
0.04362351447343826,
-0.0251... |
36c0c09d-8032-59da-83a4-0b1fc5eac292 | /articles/rag-is-dead/ | Is RAG Dead? Why Long Context Windows Don't Replace RAG | article | false | Is RAG Dead? Why Long Context Windows Don't Replace RAG
Every time a model ships a larger context window, someone declares that [retrieval augmented generation (RAG)](/articles/what-is-rag-in-ai/) is obsolete. If a model can read a million tokens, why build a retrieval pipeline?
A larger window changes the trade-off,... | [
-0.019552679732441902,
-0.026653295382857323,
-0.04807949811220169,
-0.05169641226530075,
0.07440531253814697,
-0.046588197350502014,
0.02108496055006981,
0.013257544487714767,
0.06260058283805847,
-0.06055799499154091,
0.008070546202361584,
0.02567184716463089,
0.057613689452409744,
-0.03... | |
36ed7508-43fa-50a1-a6b9-eddc38db67db | /documentation/faq/qdrant-fundamentals/ | Qdrant Fundamentals | Compatibility > Should I create one Qdrant collection per user? | documentation | false | Qdrant Fundamentals > Compatibility > Should I create one Qdrant collection per user?
No. Creating one collection per user is more resource intensive.
Instead of creating separate collections for each user, we recommend creating a [single collection](/documentation/manage-data/multitenancy/) and separate access using... | [
-0.0059290798380970955,
-0.00785007793456316,
-0.0282141026109457,
-0.020237425342202187,
0.0435066819190979,
-0.05441640689969063,
0.06373074650764465,
0.062365565448999405,
0.08389688283205032,
-0.026434384286403656,
0.033653486520051956,
0.021731417626142502,
0.07516106963157654,
-0.016... |
3707c73a-ac3a-540f-9563-e97810a3c861 | /documentation/tutorials-operations/gpu-accelerated-hnsw-indexing/ | GPU-Accelerated HNSW Indexing | Monitoring Indexing and Query Latency > Query the Collection | documentation | true | GPU-Accelerated HNSW Indexing > Monitoring Indexing and Query Latency > Query the Collection
`query` runs on its own coroutine, independent of the polling loop above. It repeatedly cycles through the `queries` sample built earlier, issuing one `query_points` request after another as fast as the client and server allow... | [
0.022985156625509262,
-0.056281983852386475,
-0.024067416787147522,
-0.03586259484291077,
0.06020234152674675,
-0.04832105338573456,
0.07394564896821976,
0.039462823420763016,
0.07413958013057709,
-0.020317833870649338,
0.06204737722873688,
0.011381188407540321,
0.056714050471782684,
-0.00... |
371990bd-e217-5c18-8bf8-3e7de516d280 | /documentation/private-cloud/qdrant-cluster-management/ | Managing a Cluster | Ephemeral Snapshot Volumes | documentation | true | Managing a Cluster > Ephemeral Snapshot Volumes
If you do not [create snapshots](https://api.qdrant.tech/api-reference/snapshots/create-snapshot), or there is no need
to keep them available after cluster restart, the snapshot storage classname can be set to `emptyDir`:
```yaml
apiVersion: qdrant.io/v1
kind: QdrantClu... | [
0.023450549691915512,
-0.009992429055273533,
0.008829209953546524,
-0.08095529675483704,
0.03734273836016655,
-0.059138793498277664,
0.039135128259658813,
0.035690564662218094,
0.05301770567893982,
-0.048162106424570084,
0.03451386094093323,
0.020365431904792786,
0.043241698294878006,
-0.0... |
375b4e18-164e-5f08-a289-1a786d726a0c | /course/essentials/day-3/sparse-retrieval-demo/ | Demo: Keyword Search with Sparse Vectors | Sparse Neural Retrieval > The Idea Behind Sparse Neural Retrieval | course | false | Demo: Keyword Search with Sparse Vectors > Sparse Neural Retrieval > The Idea Behind Sparse Neural Retrieval
Instead of assigning word weights solely based on the corpus statistics, we could use for that **machine learning models** that have shown the ability to capture a word’s **meaning in context**.
 - List all collections in the Qdrant instance
- [Create Collection](https://api.qdrant.tech/v-1-14-x/api-reference/collections/create-collection... | [
0.030023831874132156,
-0.061133190989494324,
-0.05331325903534889,
-0.00023508595768362284,
0.07733841985464096,
-0.01982555352151394,
0.044662028551101685,
0.01887160912156105,
0.07986830919981003,
-0.03938049450516701,
0.07797499746084213,
0.041363928467035294,
-0.012710102833807468,
-0.... |
383f510a-08ef-5fd6-b733-5d743e8169d5 | /course/beginners/module-2/core-data-model/ | Core Data Model | Vector | course | false | Core Data Model > Vector
A vector is a list of numbers. An embedding is a vector created by a model to represent the meaning of content. In semantic search, a dense vector is usually an embedding generated from text, images, or other data.
Each number represents one dimension of the vector. Similar content produces s... | [
0.04032376781105995,
-0.035479143261909485,
-0.06685981899499893,
-0.060621604323387146,
0.031034423038363457,
-0.016229115426540375,
0.08374876528978348,
0.10933072865009308,
0.06571577489376068,
-0.011752999387681484,
0.04546530917286873,
0.04925861209630966,
0.05553491786122322,
-0.0251... |
3846c3d1-5a80-5b6e-9063-64bbbbdb831a | /documentation/ops-monitoring/ | Monitoring & Telemetry | documentation | false | Monitoring & Telemetry
These pages cover how to observe and measure a running Qdrant deployment using its built-in metrics endpoints and external monitoring tools.
## Monitoring & Telemetry
[Monitoring & Telemetry](/documentation/ops-monitoring/monitoring/) describes the Prometheus/OpenMetrics-compatible `/metrics` ... | [
0.01536557450890541,
-0.02735116146504879,
-0.037866897881031036,
-0.007351357024163008,
0.05891113728284836,
-0.02260761149227619,
0.07689829915761948,
0.05385034903883934,
0.06392300128936768,
-0.037013087421655655,
0.01434310618788004,
0.004485671408474445,
0.0350600928068161,
-0.033486... | |
38501f3e-6411-5855-ab2d-e1875d988551 | /articles/bm42/ | BM42: Attention-Based Sparse Embeddings for Hybrid Search | Why has BM25 stayed relevant for so long? | article | false | BM42: Attention-Based Sparse Embeddings for Hybrid Search > Why has BM25 stayed relevant for so long?
To understand why, we need to analyze its components.
The famous BM25 formula is defined as:
$$
\text{score}(D,Q) = \sum_{i=1}^{N} \text{IDF}(q_i) \times \frac{f(q_i, D) \cdot (k_1 + 1)}{f(q_i, D) + k_1 \cdot \left(... | [
0.05679579824209213,
0.012050017714500427,
-0.027687078341841698,
-0.05306757614016533,
0.0529949888586998,
-0.032689034938812256,
0.04489847272634506,
0.0593024417757988,
0.0964791476726532,
-0.013697300106287003,
0.03918503224849701,
0.005079133436083794,
0.048981133848428726,
-0.0417291... |
3892a6f2-4263-51c9-9437-813ae3d1616c | /documentation/cloud-account-setup/ | Account Setup | Managing Accounts | documentation | false | Account Setup > Managing Accounts
Open the **Accounts** page from the user menu (**Accounts**) or from the account switcher (**Manage accounts**) to see every account you own or have access to. You can filter by **All**, **Owned**, or **Invited**, and the page shows your current **account limit** (for example, *Accoun... | [
0.014482894912362099,
-0.033781640231609344,
-0.0521388053894043,
-0.039976198226213455,
0.037341028451919556,
-0.019443389028310776,
-0.0020858380012214184,
0.05808379501104355,
0.05961665138602257,
-0.04161624237895012,
0.022381050512194633,
0.044361308217048645,
0.0703660249710083,
-0.0... |
389b8c9f-92b9-5d0e-aeca-e76b37f00065 | /documentation/interfaces/ | API & SDKs | documentation | false | API & SDKs
Qdrant supports these "official" clients.
> **Note:** If you are using a language that is not listed here, you can use the REST API directly or generate a client for your language
using [OpenAPI](https://github.com/qdrant/qdrant/blob/master/docs/redoc/master/openapi.json)
or [protobuf](https://github.com/q... | [
0.007170513737946749,
-0.029108885675668716,
-0.021429765969514847,
-0.02990568056702614,
0.0643414855003357,
-0.05802301689982414,
0.0567658506333828,
0.018182137981057167,
0.05731198936700821,
-0.04536814242601395,
0.047248002141714096,
0.005338758230209351,
0.0625334084033966,
-0.057937... | |
38c085f3-bfb1-5ffe-acb9-e01ff6fea50a | /documentation/migration-guidance/search-quality/ | Search Quality | Tier 1: Spot-Check (Every Migration) | documentation | true | Search Quality > Tier 1: Spot-Check (Every Migration)
Run your [baseline queries](/documentation/migration-guidance/pre-migration-baseline/) against Qdrant and eyeball the results. This catches configuration-level errors that would affect every query: wrong distance metric, missing index, broken filter logic.
```py
f... | [
0.06518466025590897,
-0.062481604516506195,
-0.05178770050406456,
-0.01867719367146492,
0.04918786510825157,
-0.018541114404797554,
0.07472550123929977,
0.022605789825320244,
0.09130557626485825,
-0.0457068495452404,
0.013971692882478237,
0.006770184263586998,
0.0358581580221653,
-0.025848... |
38d4c12e-f563-5d7b-997c-8839a2ee7ce0 | /articles/what-is-vector-quantization/ | What is Vector Quantization? | Wrapping Up > Learn More | article | false | What is Vector Quantization? > Wrapping Up > Learn More
If you want to learn more about improving accuracy, memory efficiency, and speed when using quantization in Qdrant, we have a dedicated [Quantization tips](https://qdrant.tech/documentation/manage-data/quantization/#quantization-tips) section in our docs that exp... | [
0.0541110523045063,
-0.06772755831480026,
-0.030991487205028534,
-0.060261279344558716,
0.04123551398515701,
-0.02724415995180607,
0.04636034741997719,
0.03894633799791336,
0.09741393476724625,
-0.03395679593086243,
0.02889663353562355,
0.00932741817086935,
0.0560748465359211,
-0.025727920... |
38daf08c-a18b-5d5a-a74b-20c644cee95a | /course/essentials/day-4/what-is-quantization/ | Vector Quantization Methods | The Memory Economics | course | false | Vector Quantization Methods > The Memory Economics
Consider the mathematics of scale. OpenAI's `text-embedding-3-small` produces 1536-dimensional vectors requiring 6 KB each (1536 × 4 bytes per float32). This scales predictably: 1 million vectors consume 6 GB, 10 million require 60 GB, and 100 million demand 600 GB of... | [
0.034044310450553894,
-0.034360405057668686,
-0.020500296726822853,
-0.07362101972103119,
0.05357726663351059,
-0.013071762397885323,
0.03472619503736496,
0.0797240287065506,
0.10470102727413177,
-0.04373214393854141,
0.006379075348377228,
0.0032251537777483463,
0.045921992510557175,
-0.04... |
38f0627f-df95-5327-8b20-d65a4c3289ea | /course/essentials/day-4/ | Day 4: Optimization and Scale | course | false | Day 4: Optimization and Scale
Day 4
Compression, advanced tuning, and high‑throughput ingestion.
---
## Today’s path
1. Vector Quantization Methods
2. Accuracy Recovery with Rescoring
3. High-Throughput Data Ingestion
4. Project: Quantization Performance Optimization
You’ll compare memory footprint, recall, and t... | [
0.06851071119308472,
-0.024624092504382133,
-0.029936838895082474,
-0.04962297901511192,
0.06882686167955399,
-0.007445866242051125,
0.030321618542075157,
0.02007298171520233,
0.1040755957365036,
-0.028168193995952606,
0.012168784625828266,
0.017297616228461266,
0.043220724910497665,
-0.06... | |
39212f14-87da-5186-acab-dae0c812ef71 | /documentation/tutorials-search-engineering/collaborative-filtering/ | Collaborative Filtering | Recommendations | documentation | true | Collaborative Filtering > Recommendations
Running the code prints the top 5 recommendations:
```text
Star Wars: Episode VI - Return of the Jedi (1983), Score: 98.03
Star Wars: Episode V - The Empire Strikes Back (1980), Score: 97.58
Shawshank Redemption, The (1994), Score: 78.51
Dark Knight, The (2008), Score: 73.24
... | [
0.03550621494650841,
-0.051536086946725845,
-0.03436971828341484,
-0.07888653129339218,
0.060560885816812515,
-0.04297788068652153,
0.07105148583650589,
0.058543696999549866,
0.06681712716817856,
0.0021081510931253433,
0.04536961391568184,
0.028301840648055077,
0.10636469721794128,
-0.0355... |
39214280-3f48-53a0-a51b-94ea2652510a | /documentation/production-operations/access-control/ | Data Privacy with Qdrant's Role-Based Access Control (RBAC) | How Qdrant Handles Data Privacy and Security > Database API Keys | documentation | false | Data Privacy with Qdrant's Role-Based Access Control (RBAC) > How Qdrant Handles Data Privacy and Security > Database API Keys
Database API Keys enable [Database Access Control](https://qdrant.tech/documentation/cloud/authentication/). They are used to read and write data inside your Qdrant collections.
Qdrant support... | [
0.02018432319164276,
-0.031166912987828255,
-0.06551159918308258,
-0.0380525216460228,
0.039587005972862244,
-0.04441577568650246,
0.029595550149679184,
0.029367923736572266,
0.07962903380393982,
-0.023671554401516914,
0.015094274654984474,
0.010718731209635735,
0.06420238316059113,
-0.025... |
392bbb8d-837f-5d32-8041-76b4209aa974 | /documentation/edge/edge-api/configuration/ | Configuration | Dense Vector Parameters | documentation | true | Configuration > Dense Vector Parameters
`EdgeVectorParams` configures one named dense vector. `size` and `distance` are required and cannot be changed after the shard is created.
```python
EdgeVectorParams(
size: int,
distance: Distance,
on_disk: Optional[bool] = None,
multivector_config: Optional[Mul... | [
0.020134346559643745,
-0.045416053384542465,
-0.03852348402142525,
-0.0370560884475708,
0.04144899547100067,
-0.04651947319507599,
0.06822603940963745,
0.10197179019451141,
0.07516051828861237,
-0.032042134553194046,
0.02756674215197563,
0.038677915930747986,
0.031271886080503464,
-0.01304... |
3936f008-5b71-5f4c-83db-8cddc55fc853 | /documentation/faq/qdrant-fundamentals/ | Qdrant Fundamentals | Search > If limit is higher than hnsw_ef, does Qdrant automatically adjust hnsw_ef? | documentation | false | Qdrant Fundamentals > Search > If limit is higher than hnsw_ef, does Qdrant automatically adjust hnsw_ef?
Yes. Qdrant internally sets `ef = max(ef, limit)` so that the candidate list is always at least as large as the requested result count. | [
0.010987428948283195,
-0.043245043605566025,
-0.013133948668837547,
-0.052020203322172165,
0.04745262488722801,
-0.04593454301357269,
0.05746873468160629,
0.03124651312828064,
0.0859057754278183,
-0.025738339871168137,
0.041733771562576294,
0.0360863022506237,
0.08278040587902069,
-0.03276... |
393a9ef1-6d9c-5606-a544-8478220cb2c8 | /documentation/manage-data/indexing/ | Indexing | Payload Index | documentation | false | Indexing > Payload Index
Payload index in Qdrant is similar to the index in conventional document-oriented databases.
This index is built for a specific field and type, and is used for quick point requests by the corresponding filtering condition. The index is also used to accurately estimate the filter cardinality, w... | [
0.004969407804310322,
-0.03687362000346184,
-0.05982225760817528,
-0.011306075379252434,
0.05640261247754097,
-0.03724157437682152,
0.06184236332774162,
0.03320123627781868,
0.11170782893896103,
-0.04569423943758011,
0.029756611213088036,
0.011723547242581844,
0.0404827818274498,
-0.015705... |
39408375-9a40-5639-9309-83804069bc2d | /documentation/tutorials-operations/large-scale-search/ | Large-Scale Search | Memory Usage > Size of vectors | documentation | true | Large-Scale Search > Memory Usage > Size of vectors
In our scenario, we store only quantized vectors in RAM, so it is relatively easy to calculate the required size:
```text
400_000_000 * 512d / 8 bits / 1024 (Kb) / 1024 (Mb) / 1024 (Gb) = 23.84Gb
``` | [
0.0020404106471687555,
-0.05202370509505272,
-0.025370731949806213,
-0.06038705259561539,
0.043714649975299835,
-0.029577001929283142,
0.02729235589504242,
0.011204471811652184,
0.11167481541633606,
-0.03229183331131935,
0.007826177403330803,
0.03690875694155693,
0.08756791055202484,
-0.02... |
3955d263-4d63-53ac-ba70-3064774fb17c | /documentation/capacity-planning/ | Capacity Planning | Choosing Disk over RAM > Scaling Disk Space in Qdrant Cloud | documentation | false | Capacity Planning > Choosing Disk over RAM > Scaling Disk Space in Qdrant Cloud
Clusters supporting vector search require substantial disk space compared to other search systems. If you're running low on disk space, you can use the UI at [cloud.qdrant.io](https://cloud.qdrant.io/) to scale your cluster.
When running ... | [
0.011889751069247723,
-0.05527207627892494,
-0.04168048873543739,
-0.035750195384025574,
0.07370122522115707,
-0.03412595018744469,
0.047903768718242645,
0.028497176244854927,
0.1002616137266159,
-0.030147293582558632,
-0.009955774061381817,
0.006999239791184664,
0.03491286560893059,
-0.02... |
395d2e4a-52c4-56fb-988d-e1436ba6b6d5 | /documentation/installation/ | Installation | Production > Kubernetes | documentation | true | Installation > Production > Kubernetes
| Feature | Qdrant Helm Chart | Qdrant Cloud |
|--------------------------------------------------------|:-----------------:|:-------------:|
| Open-source | ✅ | ... | [
0.024626431986689568,
-0.048868484795093536,
-0.036278173327445984,
-0.02196928858757019,
0.06624782830476761,
-0.036809805780649185,
0.06679782271385193,
0.061914753168821335,
0.0526917427778244,
-0.03812316060066223,
0.040296584367752075,
0.018701627850532532,
0.03981460630893707,
-0.065... |
3975a438-03a2-5fc4-a1b4-4b2d0feebb93 | /documentation/tutorials-build-essentials/video-anomaly-edge-part-2/ | Video Anomaly Detection Part 2: Edge-to-Cloud Pipeline | Edge-to-Cloud Escalation > Temporal Boosting | documentation | true | Video Anomaly Detection Part 2: Edge-to-Cloud Pipeline > Edge-to-Cloud Escalation > Temporal Boosting
Consecutive escalations from the same device within a 5-minute window receive a score boost:
```python
boost = min(0.3, (consecutive_count - 1) * TEMPORAL_BOOST_FACTOR) # Factor = 0.1
final_score = ensemble_score + ... | [
0.018080994486808777,
-0.028570029884576797,
-0.020852265879511833,
-0.049715183675289154,
0.02220046892762184,
-0.0024683920200914145,
0.038657840341329575,
0.024242443963885307,
0.07530682533979416,
-0.023196756839752197,
0.02700718492269516,
0.011018438264727592,
0.04974467307329178,
0.... |
397d720a-8b12-5061-940e-12236b6dd8bb | /documentation/faq/qdrant-fundamentals/ | Qdrant Fundamentals | Search > Why do I get duplicate results when paginating through search results? | documentation | false | Qdrant Fundamentals > Search > Why do I get duplicate results when paginating through search results?
Because HNSW is an approximate algorithm, the ranking of results can shift slightly between requests. As a result, paginating with `offset` can return the same point on multiple pages or skip points entirely. This is ... | [
0.020509440451860428,
-0.051784541457891464,
-0.06300241500139236,
-0.021856321021914482,
0.06375950574874878,
-0.02066182531416416,
0.07013552635908127,
0.002469742903485894,
0.05364912003278732,
-0.004049899987876415,
0.01758183352649212,
0.061470743268728256,
0.03709441423416138,
-0.016... |
3980d5d8-adbc-51e8-8ccb-2995187eb9ac | /documentation/ops-configuration/configuration/ | Configuration | Configuration Validation | documentation | true | Configuration > Configuration Validation
Qdrant validates the configuration during startup. If any issues are found, the server will terminate immediately, providing information about the error. For example:
```console
Error: invalid type: 64-bit integer `-1`, expected an unsigned 64-bit or smaller integer for key `s... | [
0.02922561764717102,
-0.03457434102892876,
-0.021628201007843018,
-0.05090764909982681,
0.031053712591528893,
-0.05296948179602623,
0.06188476085662842,
0.05329127609729767,
0.07807356864213943,
0.01190994493663311,
0.004778757691383362,
-0.005880742333829403,
0.09493888914585114,
-0.03569... |
398bcca7-312c-5084-8c5d-8440f3e6a0c9 | /documentation/faq/qdrant-fundamentals/ | Qdrant Fundamentals | Search > When should I use Reciprocal Rank Fusion (RRF) vs. Distribution-Based Score Fusion (DBSF) for hybrid search? | documentation | false | Qdrant Fundamentals > Search > When should I use Reciprocal Rank Fusion (RRF) vs. Distribution-Based Score Fusion (DBSF) for hybrid search?
Both methods combine scores from multiple retrieval legs (for example, dense and BM25), but they work differently:
* **RRF (Reciprocal Rank Fusion)** combines ranked lists based ... | [
0.019685551524162292,
-0.012065274640917778,
-0.05523649975657463,
-0.06900852173566818,
0.054115504026412964,
-0.04282040148973465,
0.04117225855588913,
0.03219855949282646,
0.0791415274143219,
-0.01806645654141903,
0.04026739299297333,
-0.016820436343550682,
0.04647141322493553,
0.010224... |
398d4123-9ec2-5e35-9cea-89112741de0c | /documentation/examples/graphrag-qdrant-neo4j/ | GraphRAG with Qdrant and Neo4j | Implementation > Ingestion > Creating Qdrant Collection | documentation | true | GraphRAG with Qdrant and Neo4j > Implementation > Ingestion > Creating Qdrant Collection
You can create a collection once you have set up your Qdrant instance. A collection in Qdrant holds vectors for search and retrieval.
```python
def create_collection(client, collection_name, vector_dimension):
if client.colle... | [
0.02084488980472088,
-0.06096487492322922,
-0.011469732969999313,
-0.036241304129362106,
0.06846430897712708,
-0.05928359553217888,
0.047447945922613144,
0.06654758751392365,
0.056314609944820404,
-0.05291442573070526,
0.02351503074169159,
0.04772379994392395,
0.085609070956707,
-0.0292873... |
39969252-6633-5534-8a0c-bb68896263c3 | /documentation/production-operations/vector-search-in-production/ | Vector Search in Production: Scaling, HA & Tuning Guide | 3. What's the Best Way to Scale the Database and Optimize Resources? > What to Monitor? | documentation | false | Vector Search in Production: Scaling, HA & Tuning Guide > 3. What's the Best Way to Scale the Database and Optimize Resources? > What to Monitor?
For retrieval, focus on P99 latency metrics (the response time for the slowest 1% of requests) rather than just average latency. This gives you a better understanding of wor... | [
-0.0041978564113378525,
-0.04462987929582596,
-0.07245989143848419,
-0.016692183911800385,
0.05552411824464798,
-0.03215671330690384,
0.021444572135806084,
0.01454202551394701,
0.0801163986325264,
-0.025308987125754356,
0.01948496513068676,
0.00486924359574914,
0.039466068148612976,
-0.047... |
399b1be1-6cf2-5c96-8af7-c397db21d779 | /course/essentials/day-5/colbert-multivectors/ | Multivectors for Late Interaction Models | Next | course | false | Multivectors for Late Interaction Models > Next
With multivectors in your toolkit, you can achieve high-precision retrieval for both text and visual documents. In the next lesson, we'll explore the Universal Query API, where you'll learn how to combine multiple retrieval strategies and use ColBERT for reranking, a mor... | [
0.02313467115163803,
-0.023152291774749756,
-0.04475603252649307,
-0.036758191883563995,
0.07222364097833633,
-0.0326048918068409,
0.0663590207695961,
0.03216147795319557,
0.07191267609596252,
-0.00587840611115098,
0.01940983533859253,
0.04467865824699402,
0.06589098274707794,
-0.021649857... |
39a5b417-1aca-51cf-b705-2646cdae5104 | /documentation/search-evaluation/pipeline-output-quality/ | Evaluating Pipeline Output Quality | Pitfalls to Watch For | documentation | false | Evaluating Pipeline Output Quality > Pitfalls to Watch For
**Judge bias.** LLM judges reward verbose, confident, or well-formatted answers even when the underlying claim is weaker. Calibrate by running a sample of outputs through human raters and comparing; if judge and human scores disagree often, adjust the rubric o... | [
0.015696031972765923,
-0.0371282622218132,
-0.05485805124044418,
-0.05850895866751671,
0.03547636792063713,
0.0069773937575519085,
0.061259545385837555,
0.02688778005540371,
0.08146721869707108,
-0.048504702746868134,
0.013870593160390854,
-0.0038212186191231012,
0.06567469984292984,
-0.01... |
39bc05d7-c470-599a-9464-b3e98c2b9432 | /course/beginners/module-5/clustering-risk-signals/ | Clustering Risk Signals | Reading a Cluster Back | course | true | Clustering Risk Signals > Reading a Cluster Back
This is what indexing `cluster_id` bought you. One integer on each point turns a night of clustering into a view an analyst can page through:
```python
def signals_in_cluster(supplier_id: str, cluster_id: int):
"""Every signal the daily job put in one cluster."""
... | [
0.021151339635252953,
-0.00927409715950489,
-0.07121482491493225,
-0.05768073722720146,
0.06853830069303513,
-0.025240452960133553,
0.003971482627093792,
0.04915650561451912,
0.06241685897111893,
0.00914346519857645,
0.025911927223205566,
0.0033068598713725805,
0.05118843540549278,
0.00268... |
39df578e-5f6b-5d85-9ed5-20e074b5b076 | /articles/vector-similarity-beyond-search/ | Vector Similarity: Going Beyond Full-Text Search | Discovery and Context Search | article | true | Vector Similarity: Going Beyond Full-Text Search > Discovery and Context Search
Context search takes only the pairs and returns points from the zones where the loss is lowest, which gives a constrained but diverse result.
These two pairs prefer velvet over metal and wicker over brass:
```python
result = client.query_... | [
0.05126334726810455,
-0.07748259603977203,
-0.023427121341228485,
-0.008682210929691792,
0.040498048067092896,
-0.023893482983112335,
0.06582915782928467,
0.001648074947297573,
0.04677826166152954,
-0.0033112666569650173,
0.026271162554621696,
0.030722683295607567,
0.026230821385979652,
0.... |
39e1010e-7aad-5d70-95d5-8652203b8a3b | /documentation/tutorials-search-engineering/slicing-filter/ | Deterministic Collection Slicing | Stratified Sampling with Payload Filters | documentation | true | Deterministic Collection Slicing > Stratified Sampling with Payload Filters
`slice` is a normal filter condition, so it combines with any other condition. This gives a reproducible sample restricted to one category, useful for a canary rollout or for evaluating a change against a single segment of the data.
```python... | [
0.00581428874284029,
-0.014246977865695953,
-0.08825419843196869,
-0.056533332914114,
0.03644254058599472,
-0.00018371253099758178,
0.10834041237831116,
0.0325356088578701,
0.07175955921411514,
-0.052339937537908554,
0.03170990198850632,
0.015920650213956833,
0.06076677143573761,
0.0176412... |
39ffb57f-a796-5c52-8d7a-15c976f6eaaa | /course/essentials/day-5/universal-query-api/ | The Universal Query API | The Universal Query API Structure | course | false | The Universal Query API > The Universal Query API Structure
The Universal Query API enables complex search patterns through a simple declarative interface:
- **Prefetch Stage**: Execute multiple searches in parallel against different vector fields. Each prefetch can have its own filters, limits, and vector types used... | [
0.009858045727014542,
-0.04566342756152153,
-0.028784560039639473,
-0.009610685519874096,
0.06196817383170128,
-0.021355150267481804,
0.03975161164999008,
0.04427064582705498,
0.07823271304368973,
-0.030443688854575157,
0.035467762500047684,
0.01655924692749977,
0.06813710182905197,
-0.018... |
3a161c21-bb92-5858-9257-994f945fccb0 | /documentation/ops-optimization/optimize/ | Optimize Performance | 1. High-Speed Search with Low Memory Usage | documentation | true | Optimize Performance > 1. High-Speed Search with Low Memory Usage
To achieve high search speed with minimal memory usage, you can store vectors on disk while minimizing the number of disk reads. Vector quantization is a technique that compresses vectors, allowing more of them to be stored in memory, thus reducing the ... | [
0.04303361102938652,
-0.05432792752981186,
-0.048893097788095474,
-0.0272000003606081,
0.06619659811258316,
-0.0034539077896624804,
0.039659690111875534,
0.0032214196398854256,
0.07942739129066467,
-0.0052900114096701145,
0.010897516272962093,
0.025911524891853333,
0.045631956309080124,
-0... |
3a21b9fd-11b7-5c8e-8aaa-d246cfe28012 | /documentation/search/filtering/ | Filtering | Filtering Conditions > Nested Key > Nested Object Filter | documentation | true | Filtering > Filtering Conditions > Nested Key > Nested Object Filter
*Available as of v1.2.0*
By default, each condition in a filter is checked against the whole payload of a point. If a filter has several conditions on fields inside the same array, each condition can match a different element of the array.
For inst... | [
0.04510286822915077,
-0.01457696408033371,
-0.07147154211997986,
-0.029255622997879982,
0.06042873486876488,
0.010992469266057014,
0.04025501012802124,
0.02467852458357811,
0.07996032387018204,
-0.014751601964235306,
0.004807205405086279,
0.05038207396864891,
0.0673571527004242,
-0.0159436... |
3a70c043-998d-5c37-8ac3-b3c9bbe509c9 | /documentation/tutorials-develop/serverless-semantic-search/ | Serverless Semantic Search | Costs and Cleanup | documentation | false | Serverless Semantic Search > Costs and Cleanup
* **AWS.** New accounts on the AWS Free plan get 100 dollars in credits, and up to 100 dollars more for exploring services, for six months. Lambda also has an always-free monthly allowance of one million requests and 400,000 GB-seconds of compute, which this function stay... | [
0.040584370493888855,
-0.016711100935935974,
-0.06532919406890869,
-0.026523571461439133,
0.06005144491791725,
-0.023580286651849747,
0.022399088367819786,
0.000052931427489966154,
0.07492665946483612,
-0.04267091304063797,
0.04624488204717636,
-0.002356842393055558,
0.03661903366446495,
-... |
3a741d55-8139-59ce-b008-37b5a72e34a6 | /course/essentials/day-6/final-project/ | Final Project: Production-Ready Documentation Search Engine | Share Your Discovery | course | true | Final Project: Production-Ready Documentation Search Engine > Share Your Discovery
### Step 1: Reflect on Your Findings
As you test your search engine, consider:
* **Chunking granularity:** Are sections large enough to answer questions but small enough for precision?
* **Payload design:** Do fields help attribution,... | [
0.013053156435489655,
-0.03730423375964165,
-0.054816748946905136,
-0.030897442251443863,
0.040677156299352646,
-0.051807716488838196,
0.031367819756269455,
0.030229391530156136,
0.0898190513253212,
-0.05153936892747879,
0.041197218000888824,
0.00619403226301074,
0.05764055624604225,
-0.03... |
3a75415e-c6d4-5469-a8a7-2793ace9ae26 | /documentation/fastembed/fastembed-splade/ | Working with SPLADE | Dictionary output | documentation | true | Working with SPLADE > Dictionary output
The dictionary is then sorted by weights in descending order.
```python
{
"chandra": 1.7163975238800049,
"third": 1.5655333995819092,
"##ya": 1.535199522972107,
"india": 1.5310232639312744,
"3": 1.385086178779602,
"mission": 1.3676567077636719,
"luna... | [
0.06278123706579208,
-0.04271874576807022,
-0.01447040680795908,
-0.04244251176714897,
0.10498262196779251,
-0.019716117531061172,
0.0003943727642763406,
0.04932316392660141,
0.09342118352651596,
-0.01622653752565384,
0.04702369496226311,
0.030375508591532707,
0.03248758241534233,
0.025343... |
3a78c911-c06e-5762-9194-9284c60af063 | /documentation/ops-configuration/memory-tiers/ | Memory Tiers | Migrating > Legacy Settings > Quantized Vectors | documentation | false | Memory Tiers > Migrating > Legacy Settings > Quantized Vectors
The legacy parameter is `always_ram`. `always_ram: true` always resolves to `pinned`. Otherwise, quantized vectors inherit the original vectors' placement: `pinned` if the vectors are in RAM, `cold` if they're on disk.
| `memory` | Legacy value |
|---|---... | [
0.03380974382162094,
-0.06961269676685333,
-0.009500810876488686,
-0.04927565157413483,
0.037067711353302,
-0.03136419132351875,
0.03857438266277313,
0.03462224826216698,
0.09015607833862305,
-0.05257438123226166,
0.0283189844340086,
0.052362892776727676,
0.048782188445329666,
-0.008918569... |
3a9ab493-5bc4-5b64-922c-782ff98d283e | /documentation/cloud-quickstart/ | Cloud Quickstart | 5. Populate the collection | documentation | true | Cloud Quickstart > 5. Populate the collection
Next, we will populate the collection with menu items. Each item will be represented as a point in the collection with its associated metadata. Instead of generating embeddings locally, we pass a `Document` object with the text and model name — Qdrant Cloud Inference handl... | [
0.04590783268213272,
-0.01887562684714794,
-0.013360417447984219,
-0.07730084657669067,
0.10358230024576187,
-0.04919266328215599,
0.08706265687942505,
0.06479280441999435,
0.06383254379034042,
0.01055921707302332,
0.03217001631855965,
0.03458471968770027,
0.03733789548277855,
-0.034644007... |
3aa083bc-0607-5501-b0d6-36d890a22385 | /articles/search-feedback-loop/ | Relevance Feedback in Information Retrieval | Similarity Scoring | article | false | Relevance Feedback in Information Retrieval > Similarity Scoring
Another family of approaches is built around the idea of incorporating relevance feedback directly into the similarity scoring function.
It might be desirable in cases where we want to preserve the original query intent, but still adjust the similarity ... | [
0.062258873134851456,
-0.0415433906018734,
-0.08085007220506668,
-0.08009187132120132,
0.052376750856637955,
-0.0262902844697237,
0.04220805689692497,
0.020523887127637863,
0.07228484004735947,
-0.004941475111991167,
0.03724166750907898,
0.021438226103782654,
0.09870125353336334,
-0.008855... |
3aaa69ce-f542-59fb-affb-af4a4f819075 | /documentation/faq/qdrant-fundamentals/ | Qdrant Fundamentals | Vectors > What does the version field on a point represent? | documentation | false | Qdrant Fundamentals > Vectors > What does the version field on a point represent?
The version field in the Query API response represents the internal shard-level operation number of the last modification to that point. It is incremented by internal processes, so it is not a reliable proxy for application-level write c... | [
0.0012120158644393086,
-0.06142236292362213,
-0.007909072563052177,
-0.03487313166260719,
0.057611048221588135,
-0.014357424341142178,
0.055519551038742065,
0.04494967311620712,
0.05585286021232605,
-0.0623958557844162,
0.03294156864285469,
0.027512770146131516,
0.0344250313937664,
-0.0260... |
3abba72a-25af-54ee-8776-27fa4be483f1 | /articles/neural-search-tutorial/ | Neural Search 101: A Complete Guide and Step-by-Step Tutorial | Step-by-step neural search tutorial using Qdrant > Step 4: Make a search API | article | true | Neural Search 101: A Complete Guide and Step-by-Step Tutorial > Step-by-step neural search tutorial using Qdrant > Step 4: Make a search API
Now that all the preparations are complete, let's start building a neural search class.
First, install all the requirements:
```bash
pip install sentence-transformers numpy
```... | [
0.0211187731474638,
-0.08978787809610367,
-0.03709583729505539,
-0.017149336636066437,
0.050788361579179764,
-0.03599963337182999,
0.07542463392019272,
0.07032445818185806,
0.08882354944944382,
-0.0051082889549434185,
0.043738581240177155,
0.046197935938835144,
0.0664873868227005,
-0.05527... |
3abf4cb6-d771-5517-bc97-71955e27ce16 | /documentation/search/explore/ | Explore | Discovery API > Discovery Search | documentation | true | Explore > Discovery API > Discovery Search
This type of search works specially well for combining multimodal, vector-constrained searches. Qdrant already has extensive support for filters, which constrain the search based on its payload, but using discovery search, you can also constrain the vector space in which the ... | [
0.08663296699523926,
-0.05656540021300316,
-0.06574878841638565,
0.010742326267063618,
0.07349070906639099,
0.013445528224110603,
0.08811978995800018,
0.021176138892769814,
0.08333738893270493,
-0.02337501011788845,
0.03631480410695076,
0.0315164215862751,
0.06045737490057945,
-0.011669297... |
3ad60964-b176-5da2-948a-fd0a6db7719d | /documentation/production-checklist/ | Production Checklist | documentation | false | Production Checklist
A practical checklist to ensure Qdrant is optimized, stable, and ready to handle real-world load.
An agent can run the checklist with you. The Review my deployment prompt asks it for a prioritized list of what would break first under real load.
--- | [
0.020814241841435432,
-0.03908110037446022,
-0.07219208776950836,
-0.07962507009506226,
0.023128308355808258,
-0.03230024129152298,
0.054751038551330566,
0.030924344435334206,
0.07418778538703918,
-0.022640055045485497,
0.03863275796175003,
-0.008922947570681572,
0.04249952360987663,
-0.03... | |
3adb66bc-75b3-5d7c-beb7-a92be0b3a045 | /course/essentials/day-4/rescoring-oversampling-indexing/ | Accuracy Recovery with Rescoring | course | false | Accuracy Recovery with Rescoring
Day 4
When we use quantization methods like Scalar, Binary, or Product Quantization, we're compressing our vectors to save memory and improve performance. However, this compression can slightly reduce the accuracy of our similarity searches because the quantized vectors are approximat... | [
0.07653291523456573,
-0.0672411322593689,
-0.015193037688732147,
-0.04682152718305588,
0.07242190092802048,
-0.01503845676779747,
0.06460622698068619,
0.023759955540299416,
0.07252244651317596,
-0.016020504757761955,
0.05084146186709404,
0.035639915615320206,
0.07091278582811356,
-0.031053... | |
3aea53de-8b27-5334-9e46-ff2487646954 | /course/multi-vector-search/module-2/colpali-family/ | ColPali Family Overview | The Impact of Bidirectional Attention > Unidirectional Attention: Designed for Generation | course | true | ColPali Family Overview > The Impact of Bidirectional Attention > Unidirectional Attention: Designed for Generation
Most large language models (LLMs) and vision-language models (VLMs) like GPT, Llama, and the base models of [ColPali](https://huggingface.co/vidore/colpali) use **unidirectional (causal) attention**. In ... | [
0.02542809583246708,
-0.004715936724096537,
-0.010576379485428333,
-0.04028306528925896,
0.07819490134716034,
0.00013571638555731624,
0.06380819529294968,
0.042959339916706085,
0.04769265279173851,
-0.020818645134568214,
0.07090801000595093,
-0.01440840121358633,
0.02641349472105503,
-0.02... |
3af3bbb2-0d6b-5ed7-878a-0905fd9a64ed | /documentation/embeddings/bedrock/ | Amazon Bedrock | Models | documentation | false | Amazon Bedrock > Models
| Model | Model ID | Dimensions | Max Input | Input |
|---|---|---|---|---|
| Amazon Titan Text Embeddings V2 | `amazon.titan-embed-text-v2:0` | 1024 (default), 512, 256 | 8,192 tokens | Text |
| Amazon Nova Multimodal Embeddings | `amazon.nova-2-multimodal-embeddings-v1:0` | 3072 (default), 10... | [
0.03857799246907234,
0.005760659463703632,
-0.0386202335357666,
-0.011662567965686321,
0.06346319615840912,
-0.06725399196147919,
0.05663040280342102,
0.05528838559985161,
0.032987166196107864,
-0.032851625233888626,
0.05128740891814232,
0.02582721970975399,
0.04422531649470329,
-0.0512433... |
3b0adfc3-6cdc-5dd1-85c6-eec31ea0d9e9 | /documentation/search/filtering/ | Filtering | documentation | false | Filtering
With Qdrant, you can set conditions when searching or retrieving points.
For example, you can impose conditions on both the [payload](/documentation/manage-data/payload/) and the `id` of the point.
Setting additional conditions is important when it is impossible to express all the features of the object in ... | [
0.028765922412276268,
-0.056415095925331116,
-0.03837677463889122,
-0.03656407818198204,
0.05309698358178139,
-0.05048341676592827,
0.057953353971242905,
0.06085870414972305,
0.06575794517993927,
0.0034754551015794277,
0.03551333397626877,
0.04828861355781555,
0.07684172689914703,
-0.03570... | |
3b2f9439-21ec-54a7-a618-0268073ff2ca | /articles/what-is-a-vector-database/ | What is a Vector Database? | The Core Functionalities of Vector Databases > 3. Updating Points: Real-Time and Bulk Adjustments | article | true | What is a Vector Database? > The Core Functionalities of Vector Databases > 3. Updating Points: Real-Time and Bulk Adjustments
Data isn't static, and neither are vectors. Keeping your points up to date is crucial for maintaining relevance in your searches.
Point updates don’t always need to happen instantly, but when... | [
0.013149389065802097,
-0.04205605015158653,
-0.01848234049975872,
-0.050458699464797974,
0.04261377081274986,
-0.048953570425510406,
0.06729269027709961,
0.010046986863017082,
0.07965497672557831,
-0.031548768281936646,
0.035385530441999435,
0.05412565916776657,
0.04666570946574211,
-0.037... |
3b624bb7-b231-5b35-8b1b-149440306fa2 | /documentation/tutorials-develop/serverless-semantic-search/ | Serverless Semantic Search | What You Need | documentation | false | Serverless Semantic Search > What You Need
* A [Rust](https://rust-lang.org) toolchain.
* [Cargo Lambda](https://www.cargo-lambda.info/guide/installation.html), which builds and deploys Rust functions to Lambda. Install it with a package manager, from a [downloaded binary](https://github.com/cargo-lambda/cargo-lambda/... | [
0.0312984474003315,
-0.04002106562256813,
-0.06049361824989319,
-0.03673362359404564,
0.05275435373187065,
-0.06081320717930794,
0.04847584664821625,
0.005941824987530708,
0.025803064927458763,
-0.03352922946214676,
0.041514746844768524,
-0.003310198662802577,
0.022719569504261017,
-0.0361... |
3b6bdb05-2e74-5e5f-ad8b-71ebe935b72f | /course/multi-vector-search/ | Multi-Vector Search Course | course | false | Multi-Vector Search Course
**Build production-ready multi-vector search pipelines**
Go beyond single-vector embeddings with late interaction models like ColBERT and ColPali. Learn the MaxSim distance metric, optimize for billion-scale search, and evaluate your retrieval pipelines with industry-standard metrics.
- ic... | [
0.004935181699693203,
-0.01847046986222267,
-0.035883110016584396,
-0.01813274435698986,
0.06677395105361938,
-0.026038018986582756,
0.04925430938601494,
0.06605623662471771,
0.036193545907735825,
-0.016643889248371124,
0.008018835447728634,
0.023999152705073357,
0.03749627619981766,
-0.05... | |
3b6d8f54-c53d-5287-9b37-16b46b35084f | /course/beginners/module-4/ | Module 4: Designing a Vector Search System | course | false | Module 4: Designing a Vector Search System
Module 4
Start with the [Module Overview](/course/beginners/module-4/module-overview/), then work through the path.
## Today's Path
1. [Where Design Decisions Live](/course/beginners/module-4/where-design-decisions-live/)
2. [Decide Before You Ingest](/course/beginners/mod... | [
0.036084357649087906,
-0.07603979110717773,
-0.05277812480926514,
0.011194968596100807,
0.08031903952360153,
-0.024447474628686905,
0.013196034356951714,
0.04817406088113785,
0.058283451944589615,
-0.01477584708482027,
0.02564975619316101,
0.031571123749017715,
0.04144984483718872,
-0.0266... | |
3b70fe51-1c13-5968-a7eb-cf1e39f92e3d | /documentation/migration-recovery-options/ | Migration and Recovery | Comparison > Resource Requirements | documentation | false | Migration and Recovery > Comparison > Resource Requirements
Disk and RAM headroom needed on the target cluster, and how each option handles an interrupted operation.
| | Migration Tool | Snapshots | Cloud Backups |
|---|---|---|---|
| **Target disk headroom** | 2× source | 2× source | ≥ source size |
| **Target RAM h... | [
0.02110646665096283,
-0.04616560414433479,
-0.041509922593832016,
-0.040140457451343536,
0.05143341049551964,
0.01807740516960621,
0.053872544318437576,
0.0008990692440420389,
0.07821688801050186,
-0.031491510570049286,
-0.006026928313076496,
-0.012266062200069427,
0.029921559616923332,
-0... |
3b935c54-d462-51a8-a9e7-ee4feae7f91f | /documentation/examples/multitenancy/ | Multitenancy with Qdrant | Indexing documents | documentation | true | Multitenancy with Qdrant > Indexing documents
No matter how our documents are generated, the flow is the same: split them into chunks, encode each chunk
with the embedding model, and store it in the collection. Let's define some documents manually. Each one has
a single metadata attribute: the name of the library it b... | [
0.015945814549922943,
-0.011805643327534199,
-0.06519697606563568,
-0.009005434811115265,
0.03319910913705826,
-0.06100489944219589,
0.05136576294898987,
0.05433833599090576,
0.04602418094873428,
-0.009375722147524357,
0.024877449497580528,
0.01616133749485016,
0.06647109240293503,
-0.0185... |
3b94224b-7611-5788-9c03-c37af103cc74 | /documentation/tutorials-basics/huggingface-datasets/ | Hugging Face Dataset Ingestion | documentation | false | Hugging Face Dataset Ingestion
| Time: 15 min | Level: Beginner | Stack: Python, Hugging Face |
| --- | ----------- | ----------- |
[Hugging Face](https://huggingface.co/) provides a platform for sharing and using ML models and
datasets. [Qdrant](https://huggingface.co/Qdrant) also publishes datasets along with the
e... | [
0.036144908517599106,
-0.03627714887261391,
-0.05021507292985916,
-0.06689480692148209,
0.07231562584638596,
-0.05590060353279114,
0.041868411004543304,
0.030980423092842102,
0.060382161289453506,
-0.03136163577437401,
0.04987557604908943,
0.023103851824998856,
0.052472472190856934,
-0.043... | |
3bb01c2e-a9a2-5f99-b544-3b93c3d6e9e0 | /documentation/private-cloud/changelog/ | Changelog | 1.12.0 (2026-10-21) | documentation | false | Changelog > 1.12.0 (2026-10-21)
| Component | Version |
|-------------------------|---------|
| qdrant-kubernetes-api | v1.39.0 |
| operator | 2.36.1 |
| qdrant-cluster-manager | v0.3.25 |
| qdrant-cluster-exporter | 1.8.3 |
Latest validated Qdrant version: 1.19.2
* Updated qdrant-... | [
0.03840905800461769,
-0.024493562057614326,
-0.060348957777023315,
-0.03626243397593498,
0.04685197025537491,
-0.05332641303539276,
0.04965800419449806,
0.036400001496076584,
0.0520622618496418,
-0.034436289221048355,
0.037816938012838364,
0.03399357572197914,
0.06245587021112442,
-0.06385... |
3bbb9a36-8c9b-5467-a418-0d4ec5f62660 | /documentation/tutorials-search-engineering/collaborative-filtering/ | Collaborative Filtering | Conclusion | documentation | false | Collaborative Filtering > Conclusion
As demonstrated, it is possible to build an interesting movie recommendation system without intensive model training using Qdrant and Sparse Vectors. This approach not only simplifies the recommendation process but also makes it scalable and interpretable. In future tutorials, we c... | [
0.04175824299454689,
-0.05341798812150955,
-0.04259027913212776,
-0.05335383489727974,
0.07338615506887436,
-0.028490161523222923,
0.10669451951980591,
0.05883798748254776,
0.04927932098507881,
-0.006198294460773468,
0.026719102635979652,
0.03404850885272026,
0.0693269744515419,
-0.0424484... |
3bc58b05-5823-594f-ba46-92d142227824 | /documentation/search-tuning/candidate-depth/ | Candidate Depth: How Much Retrieval Is Enough? | What to Tune Next | documentation | false | Candidate Depth: How Much Retrieval Is Enough? > What to Tune Next
The gap between the best possible score and the current score tells you whether the next experiment should focus on ranking or retrieval. A large gap means relevant candidates are present but not ranked highly enough. In hybrid search, test fusion sett... | [
-0.011337739415466785,
-0.03635408356785774,
-0.051846038550138474,
-0.06065545603632927,
0.03963249921798706,
-0.033982083201408386,
0.05040480196475983,
0.04675854369997978,
0.06449975073337555,
-0.038514867424964905,
0.018237581476569176,
0.033327843993902206,
0.07369045913219452,
-0.02... |
3bd04268-0635-5cea-ae1b-9d7b61362beb | /articles/what-is-a-vector-database/ | What is a Vector Database? | The Core Functionalities of Vector Databases > 2. Searching: Approximate Nearest Neighbors (ANN) Search | article | false | What is a Vector Database? > The Core Functionalities of Vector Databases > 2. Searching: Approximate Nearest Neighbors (ANN) Search
Similarity search allows you to search by **meaning**. This way you can do searches such as similar songs that evoke the same mood, finding images that match your artistic vision, or eve... | [
0.0346064418554306,
-0.07033984363079071,
-0.07712415605783463,
-0.027295881882309914,
0.04952993243932724,
-0.01073494739830494,
0.052405573427677155,
0.015832185745239258,
0.06626927107572556,
-0.026113688945770264,
0.06564751267433167,
0.021132860332727432,
0.06416726112365723,
0.005670... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.