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 |
|---|---|---|---|---|---|---|---|
6ca33f3a-34d5-5c66-bf5d-2977a9cee07e | /documentation/embeddings/mixedbread/ | Mixedbread | documentation | false | Mixedbread
[Mixedbread](https://huggingface.co/mixedbread-ai) publishes open-weight embedding models on Hugging Face under the Apache 2.0 license. There are two ways to use them with Qdrant:
- Run the model yourself with [sentence-transformers](https://sbert.net/) and send the vectors to Qdrant. This works with any Q... | [
0.05984586104750633,
-0.045269519090652466,
-0.010216980241239071,
-0.06650666892528534,
0.09236548095941544,
-0.04867162927985191,
0.07421240955591202,
0.032529331743717194,
0.004803527146577835,
-0.05016227811574936,
0.029072729870676994,
0.008675356395542622,
0.047230567783117294,
-0.05... | |
6cbdd2d0-8eb3-5131-8ddf-1ee0a845656e | /documentation/tutorials-build-essentials/video-anomaly-edge-part-1/ | Video Anomaly Detection Part 1: Architecture, Twelve Labs, and NVIDIA VSS | What's Next | documentation | false | Video Anomaly Detection Part 1: Architecture, Twelve Labs, and NVIDIA VSS > What's Next
In **[Part 2 | Edge-to-Cloud Pipeline](/documentation/tutorials-build-essentials/video-anomaly-edge-part-2/)**, we'll implement the two-shard Qdrant Edge architecture, edge triage scoring, escalation flow with ensemble scoring, and... | [
0.06082647666335106,
-0.04244689643383026,
-0.0865761935710907,
-0.042751494795084,
0.08000641316175461,
0.0021712977904826403,
0.054511331021785736,
0.020166555419564247,
0.06768791377544403,
-0.029258975759148598,
0.0741720125079155,
0.01265901979058981,
0.06843157857656479,
-0.023698164... |
6cd611b6-fbab-536f-868b-33d04c4b9bad | /course/multi-vector-search/module-3/final-project/ | Final Project: Build Your Own Multi-Vector Search System | Choose Your Challenge | course | false | Final Project: Build Your Own Multi-Vector Search System > Choose Your Challenge
This is your project. Pick a problem that matters to you.
### Use Your Own Data
The most valuable learning comes from working with documents you actually care about:
- **Technical documentation** you reference frequently
- **Research p... | [
0.024810001254081726,
-0.05470674857497215,
-0.0706605613231659,
-0.0352325513958931,
0.06944657862186432,
-0.013161644339561462,
0.04842858016490936,
0.020646406337618828,
0.07238701730966568,
-0.0452939011156559,
0.013629426248371601,
0.017241159453988075,
0.06684081256389618,
-0.0357084... |
6d15e09a-c3a6-5f1c-a525-993f6e34a820 | /documentation/manage-data/quantization/ | Quantization | TurboQuant Quantization > Rescoring | documentation | false | Quantization > TurboQuant Quantization > Rescoring
Rescoring re-evaluates the top search results with the original vectors, which recovers accuracy lost to compression. It improves search quality, but it reads the original vectors, which can slow down search when they're served from disk.
Lower bit depths lose more p... | [
0.06075341999530792,
-0.06024529039859772,
-0.021045919507741928,
-0.03827032446861267,
0.0563792958855629,
-0.009527929127216339,
0.05608886852860451,
0.013061458244919777,
0.07625950127840042,
-0.018000422045588493,
0.04509493336081505,
0.025905491784214973,
0.052054788917303085,
-0.0131... |
6d4b41a9-896e-5d37-92ad-b65d428934c5 | /documentation/private-cloud/qdrant-cluster-management/ | Managing a Cluster | Exposing a Cluster | documentation | true | Managing a Cluster > Exposing a Cluster
By default, a QdrantCluster will be exposed through an internal `ClusterIP` service. To expose the cluster to the outside world, you can create a `NodePort` service, a `LoadBalancer` service or an `Ingress` resource.
This is an example on how to create a QdrantCluster with a `L... | [
0.028122853487730026,
-0.013724349439144135,
-0.07089487463235855,
-0.024296337738633156,
0.053049687296152115,
-0.06882971525192261,
0.06042115390300751,
0.08745655417442322,
0.054431792348623276,
-0.05317429080605507,
0.03330204635858536,
0.014322145842015743,
0.01178487204015255,
-0.029... |
6d5008bb-aa35-5538-82be-e0a14c5e338f | /documentation/examples/multitenancy/ | Multitenancy with Qdrant | Setting up the pipeline > Creating the collection | documentation | true | Multitenancy with Qdrant > Setting up the pipeline > Creating the collection
Since none of the search queries will be executed on the whole collection, we configure the collection for
multitenancy from the start. This is done for [performance reasons](/documentation/manage-data/multitenancy/#calibrate-performance):
-... | [
0.031019659712910652,
-0.02993198111653328,
-0.04484419524669647,
0.00884593091905117,
0.05571334809064865,
-0.03381115570664406,
0.038224950432777405,
0.032900530844926834,
0.06522059440612793,
-0.03130823373794556,
0.02813258208334446,
0.03847868740558624,
0.06758859008550644,
-0.0190094... |
6d55e14b-d3cb-5faf-a126-ec44f50cbb34 | /documentation/examples/rag-contract-management-stackit-cohere/ | Region-Specific Contract Management System | Implementation > Search application > Generate the answer | documentation | true | Region-Specific Contract Management System > Implementation > Search application > Generate the answer
Now we create the final step: the LLM of our choice generates the answer from the retrieved chunks. The prompt asks the
model to answer from the source only:
```python
prompt_template = """
Question: {question}
Answ... | [
0.025241224095225334,
-0.05038517713546753,
-0.057471394538879395,
-0.023874809965491295,
0.05456317588686943,
-0.0419798381626606,
0.039583880454301834,
0.03407752513885498,
0.07939433306455612,
-0.04225323721766472,
0.014356398023664951,
-0.00616089254617691,
0.05102920904755592,
-0.0394... |
6d92c331-f16f-5675-8170-39b03db3df07 | /documentation/manage-data/multitenancy/ | Multitenancy | documentation | false | Multitenancy
Many applications serve multiple users, customers, or organizations ("tenants") from a shared deployment. Each tenant's data must stay isolated: when a tenant searches, they should only see their own data. Qdrant offers several solutions for keeping tenants' data separated within a shared deployment.
Cre... | [
0.0175604410469532,
0.0008908980526030064,
-0.06145702674984932,
-0.038955263793468475,
0.06504611670970917,
-0.032815657556056976,
0.0689362958073616,
0.01824686862528324,
0.09391169995069504,
-0.022732174023985863,
0.028542332351207733,
0.0294465571641922,
0.05406897887587547,
-0.0245323... | |
6d9fe3b1-ea47-57f1-9a3b-b10dd7ec4491 | /articles/gridstore-key-value-storage/ | Introducing Gridstore: Qdrant's Custom Key-Value Store | Testing Gridstore Performance: Benchmarks > Final Result | article | false | Introducing Gridstore: Qdrant's Custom Key-Value Store > Testing Gridstore Performance: Benchmarks > Final Result
Data ingestion is **twice as fast and with a smoother throughput** — a massive win!
.
We use Qdrant Cloud and its [Free Embedding Inference](/documentation/cloud/inference/#qdrant-hosted-models).
Create a Free Tier [Qdrant Cloud cluster](https://cloud.qdrant.io/) and... | [
0.06549149751663208,
-0.01085606124252081,
-0.016332663595676422,
-0.05226024612784386,
0.06525053083896637,
-0.04754282906651497,
0.06599653512239456,
0.05365748703479767,
0.041931189596652985,
-0.02790170907974243,
0.0010189547901973128,
0.009369389154016972,
0.07171452045440674,
-0.0567... |
6dc6d05a-c366-542b-bcc8-17166d17cef5 | /documentation/embeddings/superlinked/ | Superlinked | Hybrid Search with Dense and Sparse Vectors | documentation | true | Superlinked > Hybrid Search with Dense and Sparse Vectors
`SIENamedVectorizer` asks for several output types in one SIE call and returns a dictionary keyed by output type. With `BAAI/bge-m3`, one call returns a dense vector and a learned sparse vector. Sparse vectors come back as `indices` and `values`, which is the f... | [
0.07471354305744171,
-0.04331090301275253,
-0.02153167687356472,
-0.07556749880313873,
0.042825665324926376,
-0.055103953927755356,
0.06047724559903145,
0.05896010622382164,
0.057223841547966,
-0.05479218810796738,
0.0249040387570858,
0.05742480605840683,
0.03801453113555908,
0.01336593274... |
6df69083-9e7a-588d-99f6-876dfe778716 | /documentation/hybrid-cloud/operator-configuration/ | Configure the Qdrant Operator | documentation | true | Configure the Qdrant Operator
The Qdrant Operator has several configuration options, which can be configured in the advanced section of your Hybrid Cloud Environment.
The following YAML shows all configuration options with their default values:
```yaml
# Additional pod annotations
podAnnotations: {}
# Configuration... | [
0.04438002035021782,
0.008058215491473675,
-0.042282503098249435,
-0.06018926203250885,
0.04844795912504196,
-0.030313070863485336,
0.061908602714538574,
0.03932620584964752,
0.056069422513246536,
0.00008555162639822811,
0.013822341337800026,
0.005192306824028492,
0.0663299560546875,
-0.05... | |
6e002aa9-4458-5712-8c5a-d4e34ace468c | /documentation/manage-data/indexing/ | Indexing | documentation | false | Indexing
A key feature of Qdrant is the effective combination of vector and traditional indexes. It is essential to have this because for vector search to work effectively with filters, having a vector index only is not enough. In simpler terms, a vector index speeds up vector search, and payload indexes speed up filt... | [
0.018732571974396706,
-0.06856091320514679,
-0.0502268485724926,
-0.013250472955405712,
0.04823705926537514,
-0.046947091817855835,
0.059674184769392014,
0.019477151334285736,
0.09653100371360779,
-0.011548490263521671,
0.006014975719153881,
0.05712010711431503,
0.02934410236775875,
-0.041... | |
6e0b6f99-5d59-55d0-ac15-1236f2a64330 | /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.expires-in` | No | Token expiry in seconds (when auth-type is `bearer`) |
## Gotchas
- **Named vectors:** The tool reads only the class's default vector. Classes that use named vecto... | [
0.02261515147984028,
-0.046018049120903015,
-0.006243695970624685,
-0.08361262083053589,
0.05456642061471939,
-0.07200440019369125,
0.01815503090620041,
0.024183103814721107,
0.016638420522212982,
-0.05405295640230179,
0.0020073249470442533,
0.03871408849954605,
0.026234304532408714,
-0.03... |
6e0f1618-72e0-58b3-8a23-872587e7c4e8 | /documentation/tutorials-search-engineering/turbo4-multivector-search/ | Compressed Multivector Search | Setup | documentation | true | Compressed Multivector Search > Setup
You'll use [Qdrant Cloud Inference](/documentation/cloud/inference/) to generate embeddings server-side, so `qdrant-client` is the only Qdrant dependency you need. `huggingface-hub` and `polars` download and process the dataset.
```bash
pip install qdrant-client huggingface-hub p... | [
0.055428363382816315,
-0.042537540197372437,
-0.02321881614625454,
-0.059191059321165085,
0.06834200769662857,
-0.018773438408970833,
0.12077406793832779,
0.0457775741815567,
0.047508593648672104,
-0.0009380843257531524,
0.027173688635230064,
0.009313447400927544,
0.04262511059641838,
-0.0... |
6e2f7fe1-3de7-5e6a-8afe-976bda0fdd1a | /documentation/migrate-to-qdrant/from-chroma/ | From Chroma | Concept Mapping | documentation | false | From Chroma > Concept Mapping
| Chroma | Qdrant | Notes |
| :--- | :--- | :--- |
| Collection | Collection | One-to-one mapping |
| Document | Point | Each document becomes a point |
| Embeddings | Vector | Mapped automatically |
| Metadata | Payload | Direct mapping |
| Documents (text) | Payload field | Stored via `... | [
0.05577492341399193,
-0.05859837681055069,
-0.0421670600771904,
-0.07789717614650726,
0.07638265192508698,
-0.008777806535363197,
0.06620826572179794,
0.07834292948246002,
0.057195957750082016,
-0.025017719715833664,
0.04403293877840042,
-0.026410982012748718,
0.062097031623125076,
-0.0235... |
6e57b031-381d-5572-b39e-400b7c28301e | /articles/minicoil-en-es/ | miniCOIL EN-ES: Sparse Neural Retrieval Across the Language Barrier | Takeaways > What's Next | article | false | miniCOIL EN-ES: Sparse Neural Retrieval Across the Language Barrier > Takeaways > What's Next
* **More concepts.** Train the remaining 10k heads of the pruned vocabulary.
* **But refine them.** Split noisy many-disconnected-words clusters, so wrong senses don't inject ranking noise into the top.
* **More training data... | [
0.03686709702014923,
-0.03853528946638107,
-0.02012523263692856,
-0.040812864899635315,
0.0747029110789299,
-0.056450873613357544,
0.051102414727211,
0.019069794565439224,
0.05342542380094528,
-0.044512614607810974,
0.03641028702259064,
0.035320550203323364,
0.07434733211994171,
-0.0286595... |
6e5e33fe-edfb-5522-b2c0-90bca56a269b | /documentation/interfaces/ | API & SDKs | Client Libraries | documentation | false | API & SDKs > Client Libraries
||Client Repository|Installation|Version|
|-|-|-|-|
|[](https://python-client.qdrant.tech/)|**[Python](https://github.com/qdrant/qdrant-client)**|`pip install qdrant-client[fastembed]`|[Latest Release](https://github.com/qdrant/qdrant-client/releases)|
|
- [Mastra Documentation](http://mastra.ai/docs... | [
0.038187697529792786,
-0.03554777055978775,
-0.019920330494642258,
-0.055603642016649246,
0.05449982360005379,
-0.009318399243056774,
0.05100630223751068,
0.030729170888662338,
0.08919177949428558,
-0.022486316040158272,
0.016329271718859673,
0.028991561383008957,
0.03863052651286125,
-0.0... |
6e7d070a-4dee-5841-9224-83275bc681f5 | /course/essentials/day-5/colbert-multivectors/ | Multivectors for Late Interaction Models | Generating Token-Level Embeddings | course | true | Multivectors for Late Interaction Models > Generating Token-Level Embeddings
You can generate the required token-level embeddings with FastEmbed:
```python
from fastembed import LateInteractionTextEmbedding
encoder = LateInteractionTextEmbedding("colbert-ir/colbertv2.0")
doc_multivectors = list(encoder.embed(["A lon... | [
0.02107052132487297,
-0.006342447828501463,
-0.011155818589031696,
-0.05945521220564842,
0.08402229845523834,
-0.060472358018159866,
0.06902293860912323,
0.05043922737240791,
0.028149748221039772,
-0.025931712239980698,
0.020393159240484238,
0.012130171991884708,
0.03342721611261368,
-0.05... |
6e815f30-da6c-51fc-a831-deb8a643b232 | /documentation/search/hybrid-queries/ | Hybrid Queries | Hybrid Search > Reciprocal Rank Fusion (RRF) > Setting RRF Constant k | documentation | true | Hybrid Queries > Hybrid Search > Reciprocal Rank Fusion (RRF) > Setting RRF Constant k
_Available as of v1.16.0_
To set the constant $k$ in the formula:
```python
from qdrant_client import QdrantClient, models
client = QdrantClient(url="http://localhost:6333")
client.query_points(
collection_name="{collection_... | [
0.01999577321112156,
-0.03021894581615925,
-0.026043443009257317,
-0.06616535037755966,
0.06641386449337006,
-0.03906484320759773,
-0.005539979785680771,
0.04505334421992302,
0.0878094956278801,
-0.028007231652736664,
0.004241574089974165,
-0.006311439443379641,
0.08413268625736237,
-0.009... |
6e81e36a-a2a6-57ff-8b79-34edc9cec3ae | /course/multi-vector-search/module-3/multi-stage-retrieval/ | Multi-Stage Retrieval with Universal Query API | Bringing It All Together | course | false | Multi-Stage Retrieval with Universal Query API > Bringing It All Together
Multi-stage retrieval is a powerful optimization technique, but it's just one tool in your arsenal. Real-world production systems often combine multiple techniques:
- **Multi-stage retrieval** for computational efficiency
- **Score boosting** t... | [
0.002989052562043071,
-0.042634639889001846,
-0.041095104068517685,
-0.029033668339252472,
0.04572883993387222,
-0.036328528076410294,
0.04041225463151932,
-0.007239551283419132,
0.08502271771430969,
-0.025016440078616142,
0.02568003535270691,
0.033228784799575806,
0.07389600574970245,
-0.... |
6e9702d7-2770-5778-8acf-7d63373e6f6d | /articles/relevance-feedback/ | Relevance Feedback in Qdrant | Experiments > Results | article | false | Relevance Feedback in Qdrant > Experiments > Results
Rescoring humongous datasets like MSMARCO on the user side for every query would not have been fun, especially while iterating on different formulas and hyperparameters. We faced the same problem that limits many relevance feedback researchers and practitioners, tes... | [
0.046875160187482834,
-0.020435472950339317,
-0.05148782953619957,
-0.0571097657084465,
0.06692253053188324,
-0.036299578845500946,
0.04852525517344475,
0.04568314179778099,
0.05756419897079468,
-0.030735839158296585,
0.014948147349059582,
0.023575717583298683,
0.08945491164922714,
-0.0317... |
6e97944f-fa33-5961-81f0-26adcd51090c | /documentation/private-cloud/changelog/ | Changelog | 1.5.1 (2025-03-04) | documentation | false | Changelog > 1.5.1 (2025-03-04)
| Component | Version |
|------------------------|---------|
| qdrant-kubernetes-api | v1.12.0 |
| operator | 2.1.26 |
| qdrant-cluster-manager | v0.3.2 |
* Fix scaling down clusters that have TLS with self-signed certificates configured
* Various performan... | [
0.05623917654156685,
-0.0631268322467804,
-0.0643928125500679,
-0.010816328227519989,
0.03455439954996109,
-0.046435002237558365,
0.06056276336312294,
0.03760199621319771,
0.05296165496110916,
-0.036231257021427155,
0.028250789269804955,
0.014493700116872787,
0.039577092975378036,
-0.04893... |
6ecdb2f0-72e7-5d92-b49e-cf3d0105b09e | /documentation/overview/ | Overview | Scaling Considerations > Filtering | documentation | false | Overview > Scaling Considerations > Filtering
Vector search alone can provide a decent search experience to your users; however, semantic similarity is rarely the only factor you have to consider. Embeddings won’t capture attributes such as price, and typically, a filter on a specific payload attribute has to be appli... | [
0.062296777963638306,
-0.05445607751607895,
-0.038990430533885956,
-0.01706671342253685,
0.0475168377161026,
-0.032732900232076645,
0.0690939724445343,
0.026031283661723137,
0.07763908803462982,
-0.023700173944234848,
0.05144811421632767,
0.03481575846672058,
0.07750537991523743,
-0.041782... |
6edb782c-86d5-580e-bb4c-c67a4d2556c1 | /course/multi-vector-search/module-0/qdrant-setup/ | Qdrant Setup | course | false | Qdrant Setup
Module 0
Before diving into multi-vector search, you need a running Qdrant instance. Whether you choose Qdrant Cloud for a managed solution or a local deployment, this lesson will get you up and running.
Multi-vector search requires specific collection configurations that differ from traditional single-... | [
0.006048677489161491,
-0.06284832954406738,
-0.04125276207923889,
-0.032731398940086365,
0.03879733383655548,
-0.019688785076141357,
0.049670957028865814,
0.02008948102593422,
0.0624401830136776,
-0.01851477473974228,
-0.0018991281976923347,
0.03584582731127739,
0.04256906732916832,
-0.041... | |
6ee4c81d-93cc-517c-8a0b-18ad96d30fd8 | /documentation/tutorials-search-engineering/collaborative-filtering/ | Collaborative Filtering | Implementation > Upload the data | documentation | true | Collaborative Filtering > Implementation > Upload the data
Here, you will initialize the Qdrant client and create a new collection to store the data.
Convert the user ratings to sparse vectors, one point per user.
```python
# Create a collection with one named sparse vector
qdrant_client.create_collection(
colle... | [
0.05200258642435074,
-0.03356574475765228,
0.006438813172280788,
-0.04698452726006508,
0.07303833216428757,
-0.01658429391682148,
0.07159008830785751,
0.05195818096399307,
0.07744452357292175,
-0.013835616409778595,
0.01177969854325056,
0.03103775903582573,
0.09156597405672073,
-0.01495355... |
6ee4e405-4565-5407-8c77-7a139b2756ba | /documentation/migrate-to-qdrant/from-chroma/ | From Chroma | Gotchas | documentation | false | From Chroma > Gotchas
- **Distance metric:** The tool does not read the collection's distance function and defaults to `euclid`, which matches Chroma's default `l2`. If your collection uses `cosine` or `ip`, set `--qdrant.distance-metric`.
## Next Steps
After migration, verify your data arrived correctly with the [M... | [
0.03273722901940346,
-0.025131020694971085,
-0.019318878650665283,
-0.001202940009534359,
0.08446725457906723,
-0.012761752121150494,
0.048252951353788376,
0.07363618165254593,
0.08584190160036087,
-0.0045947241596877575,
0.023965198546648026,
-0.021779732778668404,
0.053300946950912476,
-... |
6ee75fc6-9bcb-55ea-875f-09b13e88f231 | /documentation/tutorials-build-essentials/video-anomaly-edge-part-1/ | Video Anomaly Detection Part 1: Architecture, Twelve Labs, and NVIDIA VSS | Architecture Overview | documentation | false | Video Anomaly Detection Part 1: Architecture, Twelve Labs, and NVIDIA VSS > Architecture Overview
The system uses a three-tier architecture designed around a simple principle: **cheap, fast triage at the edge; accurate, rich analysis in the cloud.**
**Architecture:**
- **Edge Tier (NVIDIA Jetson)**
- *NVIDIA Metro... | [
0.03355138748884201,
-0.01956675760447979,
-0.09391400963068008,
-0.028630251064896584,
0.053578756749629974,
0.004047755617648363,
0.057484276592731476,
0.006575461942702532,
0.08209241926670074,
-0.04223090410232544,
0.035818830132484436,
-0.02366974763572216,
0.05237817391753197,
-0.018... |
6ef78843-d114-5f8f-8560-d8290415cfe0 | /documentation/migration-guidance/ | Migration Guidance | Who This Guide Is For | documentation | false | Migration Guidance > Who This Guide Is For
Engineers and teams migrating to Qdrant from another vector search system (Pinecone, Weaviate, Milvus/Zilliz, Elasticsearch, pgvector, or any custom FAISS/ScaNN deployment). The verification steps are system-agnostic on the source side. That is, you capture baselines from you... | [
0.062430474907159805,
-0.07892172783613205,
-0.042962536215782166,
-0.044539790600538254,
0.017298508435487747,
-0.0101304417476058,
0.07785072922706604,
0.011449803598225117,
0.042946528643369675,
-0.011785385198891163,
0.026767777279019356,
0.01158904004842043,
0.04085170850157738,
-0.01... |
6f1cdf9d-e014-50d4-b02f-3dbf53ae485b | /documentation/search-evaluation/retrieval-relevance/ | Measuring Retrieval Relevance | Next Steps | documentation | false | Measuring Retrieval Relevance > Next Steps
Once retrieval relevance is on target, the next layer is pipeline output quality: whether the full pipeline produces the right output when retrieval feeds into a consumer (LLM generator, ranker, or UI). See [Evaluating Pipeline Output Quality](/documentation/search-evaluation... | [
0.005234719719737768,
-0.04114241898059845,
-0.0433918796479702,
-0.06159542128443718,
0.04236571863293648,
-0.014928396791219711,
0.07358904182910919,
0.015217054635286331,
0.06680615991353989,
-0.028135724365711212,
0.0010610106401145458,
0.017038317397236824,
0.06588365882635117,
-0.011... |
6f218b04-d67b-558e-ab5a-4188c6d0f21a | /articles/relevance-feedback/ | Relevance Feedback in Qdrant | Our Idea | article | false | Relevance Feedback in Qdrant > Our Idea
Usually neither LLMs nor subject matter experts providing feedback know the dataset’s exact shape or the vector space it spans. For that they would need to process a large subset of stored vectors.
What they can do is help define a relevance direction from the examples they hav... | [
0.025205744430422783,
-0.019637782126665115,
-0.035156119614839554,
-0.06585250794887543,
0.04579075053334236,
-0.047411493957042694,
0.07317876070737839,
0.01777963526546955,
0.04188905656337738,
-0.018561450764536858,
-0.010553191415965557,
0.003878870978951454,
0.09842266887426376,
-0.0... |
6f451b8f-9666-56eb-82bc-5edbb128b8da | /documentation/examples/qdrant-dspy-medicalbot/ | Building a Chain-of-Thought Medical Chatbot with Qdrant and DSPy | Guardrails and Medical Question Detection | documentation | true | Building a Chain-of-Thought Medical Chatbot with Qdrant and DSPy > Guardrails and Medical Question Detection
Since this is a medical chatbot, we can introduce a simple guardrail to ensure it doesn’t respond to unrelated questions like the weather. This can be implemented using a DSPy module.
The chatbot checks if eve... | [
0.037801675498485565,
-0.004771401174366474,
-0.02885720692574978,
-0.017638135701417923,
0.05802037939429283,
-0.01958388462662697,
0.07582283020019531,
0.04036721959710121,
0.03383422642946243,
-0.01505354605615139,
0.043582428246736526,
0.04210012033581734,
0.05201805382966995,
-0.04004... |
6f45acf7-d180-5474-a6d5-223e177f7c13 | /documentation/tutorials-operations/incremental-embedding-updates/ | Incremental Embedding Updates | Characteristics of a Document Chunk | documentation | true | Incremental Embedding Updates > Characteristics of a Document Chunk
What usually happens to documentation?
Something completely new appears, information on pages gets fixed, pages get restructured and sections are moved as-is, pages get deleted.
It makes sense to monitor two independent characteristics of a document ... | [
0.030458731576800346,
-0.04062872752547264,
-0.07784419506788254,
-0.029952187091112137,
0.05392719432711601,
-0.02260676957666874,
0.04266316816210747,
0.03795721009373665,
0.09906072169542313,
-0.046472351998090744,
-0.005561469122767448,
0.0185923483222723,
0.017314715310931206,
-0.0267... |
6f6afdf3-46cf-5367-a199-6ab4a2919fc4 | /documentation/migration-guidance/search-quality/ | Search Quality | Tier 3: Gold-Standard Evaluation (When You Have Labeled Data) > Building Toward Tier 3 Incrementally | documentation | false | Search Quality > Tier 3: Gold-Standard Evaluation (When You Have Labeled Data) > Building Toward Tier 3 Incrementally
Most teams don't have labeled evaluation data on migration day. That's fine. Here's a practical path:
1. **Day 0 (migration):** Run Tier 1 and Tier 2. You now have a quantitative baseline.
2. **Week 1... | [
0.03584431856870651,
-0.05052175372838974,
-0.04419192671775818,
-0.03293028846383095,
0.06474749743938446,
-0.04591605067253113,
0.043877974152565,
0.0528704859316349,
0.06518812477588654,
-0.01834562048316002,
0.009606276638805866,
0.02615552768111229,
0.06266564130783081,
-0.04161032661... |
6f7d6e9e-5ad2-5d05-9d75-88fb9f0bdf7d | /documentation/production-operations/memory-tiers/ | Memory Tiers in Qdrant: What to Use and When | Combining Tiers and Quantization | documentation | false | Memory Tiers in Qdrant: What to Use and When > Combining Tiers and Quantization
Tiers and quantization combine into a handful of configurations that cover most collections, from a small one that fits entirely in RAM to one too large for RAM at any budget:
- **Full Cached, No Quantization.** Dense vectors and the HNSW... | [
0.05442211031913757,
-0.05053924769163132,
-0.06814167648553848,
-0.045060742646455765,
0.05858471617102623,
-0.03354748338460922,
0.05665338411927223,
0.02540011703968048,
0.11707800626754761,
0.005251977127045393,
0.015312843024730682,
0.017462586984038353,
0.03429931402206421,
-0.031584... |
6f80879f-3efa-520f-8948-1288eeb45285 | /documentation/tutorials-build-essentials/video-anomaly-edge-part-1/ | Video Anomaly Detection Part 1: Architecture, Twelve Labs, and NVIDIA VSS | Connecting with NVIDIA VSS | documentation | false | Video Anomaly Detection Part 1: Architecture, Twelve Labs, and NVIDIA VSS > Connecting with NVIDIA VSS
Before jumping into the code, let's cover what NVIDIA VSS is and the tools it provides.
VSS stands for **Video Search and Summarization**. It is an [NVIDIA AI Blueprint](https://blogs.nvidia.com/blog/ai-blueprint-vi... | [
0.02150825597345829,
-0.046804510056972504,
-0.08986280858516693,
-0.03817043825984001,
0.05777136608958244,
-0.009316425770521164,
0.03576255217194557,
-0.002990073524415493,
0.06177476793527603,
-0.039338115602731705,
0.03673027828335762,
-0.007672534324228764,
0.06170714274048805,
-0.00... |
6f837238-1030-5758-a190-7989dd70c51c | /documentation/upgrades/ | Upgrades | Upgrading to v1.20 | documentation | false | Upgrades > Upgrading to v1.20
As of v1.20.0, the internal gRPC API enforces authentication by default when an API key is configured. If your cluster sets only `read_only_api_key`, peers lose peer-to-peer communication after the upgrade. Set `api_key`, or set `enforce_internal_auth` to `false`. If your cluster sets onl... | [
0.04857442528009415,
-0.03635184094309807,
-0.05454297363758087,
-0.019048573449254036,
0.052937790751457214,
-0.03966650739312172,
0.023524653166532516,
0.09812101721763611,
0.026340201497077942,
-0.03322107717394829,
0.0374332033097744,
0.04089495539665222,
0.05913095921278,
-0.069505542... |
6fc58a53-d6a0-57e4-ae8e-5bfc4ca506bc | /documentation/tutorials-operations/blue-green-deployment/ | Blue-Green Cluster Deployment | When to Use It | documentation | false | Blue-Green Cluster Deployment > When to Use It
Use a blue-green cluster deployment when:
- You do a **monthly or regular full dataset replacement** and want to build and index the new data without touching the live cluster.
- You are doing a **version upgrade** and want to test the new Qdrant version on real data b... | [
0.0019858742598444223,
-0.042632728815078735,
-0.046637218445539474,
-0.04347744211554527,
0.10922291874885559,
-0.015879053622484207,
0.014577393420040607,
0.008142086677253246,
0.0494845025241375,
-0.034488361328840256,
0.023025354370474815,
-0.008921701461076736,
-0.006968172732740641,
... |
6fcd83ac-085d-519c-a205-efd1b262e93d | /articles/filtered-vector-search-acorn/ | Filtered Vector Search: What ACORN Fixes, and What Fixes ACORN | ACORN on a Normal Collection | article | false | Filtered Vector Search: What ACORN Fixes, and What Fixes ACORN > ACORN on a Normal Collection
The earlier tables pinned `full_scan_threshold` low to hold the three fixed strategies on the graph. Nobody runs a collection that way. This is the default configuration: extra edges, the default threshold, and the planner fr... | [
0.02758975140750408,
-0.0400652177631855,
-0.04577450454235077,
0.0007111231680028141,
0.0899457260966301,
0.008786722086369991,
0.06084638833999634,
0.0021948886569589376,
0.057062409818172455,
-0.01667601615190506,
0.053806740790605545,
0.021756697446107864,
0.05402382090687752,
-0.03049... |
6fd49551-2adb-5e4c-aa6c-34ca8951c545 | /documentation/tutorials-search-engineering/sparse-embeddings-ecommerce-part-3/ | Fine-Tuning Sparse Embeddings for E-Commerce Search | Part 3: Evaluation and Hard Negatives | What Fine-Tuning Actually Changes | documentation | false | Fine-Tuning Sparse Embeddings for E-Commerce Search | Part 3: Evaluation and Hard Negatives > What Fine-Tuning Actually Changes
Looking at the model's outputs before and after fine-tuning reveals what it learned:
**Query expansion improves:**
- "laptop" → adds "notebook", "computer", "macbook"
- "wireless earbuds" → ... | [
0.04842590540647507,
-0.0846785306930542,
-0.05866321921348572,
-0.043769437819719315,
0.04340110346674919,
-0.03540676832199097,
0.05940718203783035,
0.00013621115067508072,
0.07417335361242294,
-0.03858042135834694,
0.012863530777394772,
0.03375529870390892,
0.046107396483421326,
-0.0310... |
7021a9c0-5ebe-5d56-8354-11fd61331e9d | /documentation/manage-data/indexing/ | Indexing | Full-Text Index > Tokenizers > Word | documentation | false | Indexing > Full-Text Index > Tokenizers > Word
The `word` tokenizer is the default. It splits the string into words, separated by spaces, punctuation marks, and special characters. | [
0.012021142989397049,
-0.009162393398582935,
-0.03147037327289581,
-0.06444098055362701,
0.0609295591711998,
-0.009763608686625957,
0.008772707544267178,
0.023019447922706604,
0.060612987726926804,
-0.021722014993429184,
0.0369526743888855,
-0.0022084263619035482,
0.045193061232566833,
-0.... |
70486bd1-1fce-508d-beb6-f059bd8a5fbe | /documentation/cloud/encryption/ | Encryption at Rest | Prerequisites | documentation | false | Encryption at Rest > Prerequisites
- A [Premium Managed Cloud](/documentation/cloud-premium/) subscription
- An existing cluster in Qdrant Cloud (create one before requesting encryption)
- Whether your cluster is empty or contains data. We recommend starting with an empty cluster where possible. | [
0.01064855232834816,
-0.022167887538671494,
-0.046240001916885376,
-0.06589977443218231,
0.03664996102452278,
-0.059194035828113556,
0.04734515771269798,
0.04535529389977455,
0.0657263696193695,
-0.020963365212082863,
-0.0016223995480686426,
-0.0013042220380157232,
0.07835519313812256,
-0.... |
705ac20a-a4ad-5a5d-9ef8-abf8f485dd94 | /course/multi-vector-search/module-1/problems-multi-vector/ | Problems of Multi-Vector Search | What's Next | course | false | Problems of Multi-Vector Search > What's Next
Understanding these fundamental limitations prepares you for practical implementation. In the next lesson, you'll learn how to use **Qdrant multi-vector search**, including collection configuration and storage optimization strategies.
From theory to practice - let's see h... | [
0.04131362587213516,
-0.055124431848526,
-0.042463548481464386,
-0.02399618737399578,
0.06986227631568909,
-0.012373116798698902,
0.050603341311216354,
0.03316332772374153,
0.09212164580821991,
-0.002922661369666457,
0.004694454837590456,
0.01139821670949459,
0.05650630593299866,
-0.032577... |
706cfe10-35be-5fde-aaf4-e59defead5af | /documentation/private-cloud/api-reference/ | API Reference | qdrant.io/v1 > Resource Types > Storage | documentation | false | API Reference > qdrant.io/v1 > Resource Types > Storage
_Appears in:_
- [QdrantClusterSpec](#qdrantclusterspec)
| Field | Description | Default | Validation |
| --- | --- | --- | --- |
| `volumeAttributesClassName` _string_ | VolumeAttributesClassName specifies VolumeAttributeClass name to use for the storage PVCs | ... | [
0.026900628581643105,
-0.04324384406208992,
-0.00942928809672594,
-0.024011962115764618,
0.07096312940120697,
-0.03259638324379921,
0.030627010390162468,
0.07220219075679779,
0.04919738695025444,
-0.03522046282887459,
0.04432778060436249,
-0.009402332827448845,
0.013789727352559566,
-0.009... |
70abe354-76d2-510f-8c04-21a2486d38ce | /articles/relevance-feedback/ | Relevance Feedback in Qdrant | Relevance Feedback Query | article | false | Relevance Feedback in Qdrant > Relevance Feedback Query
The results were convincing enough to justify implementing Relevance Feedback Query. And [here it is](https://qdrant.tech/documentation/search/search-relevance/#relevance-feedback), ready for your retrieval pipelines! | [
0.025269487872719765,
-0.02029009349644184,
-0.024009481072425842,
-0.08396348357200623,
0.04296575486660004,
-0.040213122963905334,
0.06086175516247749,
0.032920025289058685,
0.04391436278820038,
-0.0009887950727716088,
0.008501315489411354,
0.04428660124540329,
0.10551604628562927,
-0.02... |
70bde88b-feb8-5c4a-b397-dcbd302b4172 | /course/essentials/day-3/sparse-vectors/ | Sparse Vectors and Inverted Indexes | Sparse Vector Representations > Organizing Sparse Vectors: Inverted Index | course | true | Sparse Vectors and Inverted Indexes > Sparse Vector Representations > Organizing Sparse Vectors: Inverted Index
Searching for similar sparse vectors boils down to finding the vectors that share non‑zero dimensions with the query and multiplying corresponding values.
Scanning every vector to check for matching non-zer... | [
0.0764855146408081,
-0.05344498157501221,
-0.04754182696342468,
-0.03911871835589409,
0.07762863487005234,
-0.014093874953687191,
0.04975396767258644,
-0.0008676385041326284,
0.07899078726768494,
-0.009674608707427979,
0.0009679364156909287,
0.034159813076257706,
0.04248696565628052,
-0.02... |
70c681f3-0200-52f6-bb02-086d9d440950 | /documentation/edge/edge-api/updating-data/ | Updating Data | Update Operations > Payload Indexes | documentation | false | Updating Data > Update Operations > Payload Indexes
The `update` operation also enables you to create and delete payload indexes.
| Operation | Parameters | Description |
| --- | --- | --- |
| `create_field_index` | `field_name`, `schema` | Index a payload field. |
| `delete_field_index` | `field_name` | Remove a pay... | [
0.03206265717744827,
-0.016748657450079918,
-0.03768740966916084,
-0.038933005183935165,
0.06737072020769119,
-0.014060917310416698,
-0.0009138863533735275,
0.028430672362446785,
0.045084625482559204,
-0.06425540894269943,
0.0534941591322422,
0.05826621875166893,
0.03310925140976906,
-0.01... |
70c7c958-fa1e-5696-a372-f9f66f15ee37 | /documentation/cloud-premium/ | Premium Tier | Multi-AZ Clusters | documentation | false | Premium Tier > Multi-AZ Clusters
A multi-AZ cluster spreads its nodes across several availability zones in one region, so the cluster stays available if a zone goes down. Qdrant places each shard's replicas in different zones and routes traffic between zones automatically.
Replication alone doesn't give you this prot... | [
-0.003892298322170973,
-0.03338010609149933,
-0.06436215341091156,
-0.007724430412054062,
0.0670953020453453,
-0.011785456910729408,
0.04761503264307976,
0.029477892443537712,
0.03565644100308418,
-0.01508721336722374,
0.04484939202666283,
0.026220589876174927,
0.026771368458867073,
-0.003... |
70d480c6-37aa-5e7b-ac05-fc34d5cd1156 | /documentation/overview/ | Overview | Scaling Considerations > Payload Indexes | documentation | false | Overview > Scaling Considerations > Payload Indexes
The payload index is a helper data structure that enables effective filtering on a particular payload attribute. It’s a concept familiar from relational databases, where we create an index on a column that we often filter by. Similarly, in Qdrant, you should also mak... | [
0.017524806782603264,
-0.042531147599220276,
-0.0661289244890213,
-0.031197212636470795,
0.06529612839221954,
-0.020596468821167946,
0.02162487432360649,
0.015118027105927467,
0.06847207993268967,
-0.02936711348593235,
0.04063504561781883,
0.024175124242901802,
0.0348125584423542,
-0.03505... |
70ef622f-8e9a-5fd7-bcb6-c04fd3182b15 | /articles/metric-learning-tips/ | Metric Learning Tips & Tricks | What is metric learning and why using it? | article | false | Metric Learning Tips & Tricks > What is metric learning and why using it?
The Comparator is usually a very simple function that could be calculated very quickly.
It might be cosine similarity or even a dot product.
Two-stage schema allows performing complex calculations only once per object.
Once transformed, the Comp... | [
0.057688698172569275,
-0.024919461458921432,
-0.049559902399778366,
-0.020118597894906998,
0.04571021348237991,
-0.021989649161696434,
0.04132900759577751,
0.03942454978823662,
0.07459264248609543,
0.005925304722040892,
0.028771227225661278,
-0.002772961976006627,
0.09119778126478195,
-0.0... |
70f5ef12-c65f-5714-b8de-9c067d2bd9ce | /documentation/tutorials-build-essentials/video-anomaly-edge-part-1/ | Video Anomaly Detection Part 1: Architecture, Twelve Labs, and NVIDIA VSS | Introduction | documentation | false | Video Anomaly Detection Part 1: Architecture, Twelve Labs, and NVIDIA VSS > Introduction
What if your surveillance cameras could autonomously detect any type of anomaly (fights, accidents, intrusions, equipment failures) without ever being trained on those specific events? What if instead of manually reviewing hours o... | [
0.055513832718133926,
-0.06755458563566208,
-0.09744429588317871,
-0.05418599769473076,
0.06571761518716812,
-0.02779373526573181,
0.03193650022149086,
0.014159469865262508,
0.07138712704181671,
-0.02733636274933815,
0.06134044751524925,
-0.0018427129834890366,
0.054385002702474594,
-0.019... |
71182f17-57c0-5ee8-acd5-2a5d8dffc11a | /course/multi-vector-search/module-1/late-interaction-basics/ | Late Interaction Basics | What's Next | course | false | Late Interaction Basics > What's Next
Understanding the conceptual foundation of late interaction is the first step. But how exactly do we compare sets of query vectors against sets of document vectors?
In the next lesson, you'll learn about **MaxSim** - the distance metric that powers late interaction search. MaxSim... | [
0.058311671018600464,
-0.038001470267772675,
-0.03230326622724533,
-0.015584046021103859,
0.05573546513915062,
-0.05696364492177963,
0.033193740993738174,
0.009906357154250145,
0.06592518836259842,
-0.034917011857032776,
0.03632121905684471,
-0.016930093988776207,
0.05818178132176399,
-0.0... |
711adf89-17d5-5509-946b-1dbe8e89a67e | /documentation/cloud/encryption/ | Encryption at Rest | Step 3: Reply to the Support Ticket | documentation | false | Encryption at Rest > Step 3: Reply to the Support Ticket
Send Qdrant Support your key identifier:
- **AWS**: Key ARN
- **GCP**: Key ID
- **Azure**: Key URL
Qdrant Support will configure encryption for your cluster. | [
0.038741886615753174,
-0.013105634599924088,
-0.04885873943567276,
-0.025077039375901222,
0.022491419687867165,
-0.029593005776405334,
0.043793685734272,
0.06339842826128006,
0.055450689047575,
0.008173327893018723,
0.007586178835481405,
-0.006392719689756632,
0.030924297869205475,
-0.0418... |
71221493-43ab-5527-a605-2247b018401a | /articles/minicoil/ | miniCOIL: on the Road to Usable Sparse Neural Retrieval | Eyes on the Prize: Usable Sparse Neural Retrieval > Standing on the Shoulders of BM25 | article | false | miniCOIL: on the Road to Usable Sparse Neural Retrieval > Eyes on the Prize: Usable Sparse Neural Retrieval > Standing on the Shoulders of BM25
BM25 has been a decent baseline across various domains for many years -- and for a good reason. So why discard a time-proven formula?
Instead of training our sparse neural re... | [
0.04836467653512955,
-0.025438429787755013,
-0.007853507995605469,
-0.03100760467350483,
0.058191411197185516,
-0.040981095284223557,
0.051544953137636185,
0.019908016547560692,
0.06554780900478363,
-0.004935070406645536,
0.04558515548706055,
0.01677568256855011,
0.07441852986812592,
-0.00... |
7133d736-464d-5e1b-b948-bb9759dcce12 | /articles/distance-based-exploration/ | Data Exploration with Qdrant's Distance Matrix API | Clustering | article | true | Data Exploration with Qdrant's Distance Matrix API > Clustering
Another approach to data structure understanding is clustering, which groups similar items.
*Note that there's no universally best clustering criterion or algorithm.*
.to(colpali_model.device)
# R... | [
0.04022487998008728,
-0.05495042726397514,
-0.02769029140472412,
-0.04589513689279556,
0.06514092534780502,
-0.01194986142218113,
0.025421930477023125,
0.035038672387599945,
0.0785682201385498,
0.007077168673276901,
0.017425544559955597,
0.022225501015782356,
0.05729467421770096,
-0.025942... |
7152b1f3-71c7-51e2-b930-115ab7ea46d1 | /documentation/capacity-planning/ | Capacity Planning | Cluster Topology | documentation | false | Capacity Planning > Cluster Topology
The formulas in the previous section tell you how much RAM and disk your data needs. They don't tell you how many nodes and shards to spread it across. That's driven by different considerations: fault tolerance and room for future growth. | [
-0.010985290631651878,
-0.03443656489253044,
-0.04839201644062996,
-0.08394377678632736,
0.06335660815238953,
-0.023560041561722755,
0.03577765077352524,
0.07335543632507324,
0.08851643651723862,
-0.005371052771806717,
0.012260980904102325,
0.00787947978824377,
0.08070960640907288,
0.00037... |
7157e50d-2ef5-55aa-890a-3f3987fcf48a | /documentation/data-synchronization/with-postgres/ | Keeping Postgres and Qdrant in Sync | Tier 1: Application-Level Dual-Write > When to Use This | documentation | false | Keeping Postgres and Qdrant in Sync > Tier 1: Application-Level Dual-Write > When to Use This
- Prototypes and MVPs
- Internal tools where occasional inconsistency is tolerable
- Low write throughput (< 10K products, < a few hundred writes/day)
- Teams that want to ship fast and revisit operational concerns later
---... | [
0.0028280233964323997,
-0.04919604957103729,
-0.05699381232261658,
-0.006002041511237621,
0.02535700984299183,
-0.056125689297914505,
0.02093842253088951,
0.014868935570120811,
0.08862615376710892,
0.012942856177687645,
0.033017732203006744,
0.0271060299128294,
0.050341881811618805,
-0.007... |
7159b97d-a025-5f6b-9075-0c403daeb5c6 | /articles/search-feedback-loop/ | Relevance Feedback in Information Retrieval | Query Refinement > Query As Text | article | false | Relevance Feedback in Information Retrieval > Query Refinement > Query As Text
The reason why you’re most probably not familiar with these models – their absence in the industry – is that their **training** itself is a **high upfront cost**, and even though it was “paid”, these models [struggle with generalization](ht... | [
0.02799292467534542,
-0.00522157596424222,
-0.024484975263476372,
-0.04116222262382507,
0.03270355984568596,
-0.030149979516863823,
0.03766174241900444,
0.0029811346903443336,
0.07413442432880402,
0.003321336582303047,
0.04021228849887848,
0.016511697322130203,
0.07502692192792892,
-0.0325... |
71809acb-1e73-5788-a6cb-049dd09835fa | /documentation/manage-data/multitenancy/ | Multitenancy | Tiered Multitenancy > Limitations | documentation | false | Multitenancy > Tiered Multitenancy > Limitations
- The fallback shard and the dedicated tenant shards must have a `shards_number` of 1. A new dedicated tenant shard also needs to have a `replication_factor` of 1 at creation time. This is because the shard transfer mechanism only works with single shards. If you wish, ... | [
-0.0028171762824058533,
0.0034885057248175144,
-0.03734426200389862,
-0.015768254175782204,
0.03660793974995613,
-0.04557918384671211,
0.03482498228549957,
0.06196548789739609,
0.07977639138698578,
-0.026121126487851143,
0.026988709345459938,
0.04345560446381569,
0.025689875707030296,
-0.0... |
7188f04e-d30d-59b1-84ca-73d628a6ee19 | /documentation/search/filtering/ | Filtering | Filtering Conditions > Slice | documentation | true | Filtering > Filtering Conditions > Slice
*Available as of v1.19.0*
The `slice` condition divides a collection into a specific number of deterministic, disjoint subsets and matches all points in one of those subsets.
Slicing is useful to [scroll](/documentation/manage-data/points/#scroll-points) through several subse... | [
0.029199806973338127,
-0.04958532750606537,
-0.08072301745414734,
-0.018104800954461098,
0.05370459333062172,
-0.009354488924145699,
0.07284718751907349,
0.010062486864626408,
0.0842260792851448,
-0.011536398902535439,
0.02205551043152809,
0.026132414117455482,
0.04419140890240669,
-0.0118... |
71aa661a-0910-537d-943e-49190a2d78fb | /articles/rag-is-dead/ | Is RAG Dead? Why Long Context Windows Don't Replace RAG | How to choose | article | false | Is RAG Dead? Why Long Context Windows Don't Replace RAG > How to choose
| Your situation | Start with |
|---|---|
| Corpus under about 200,000 tokens, stable, queried rarely | Long context |
| Same corpus queried often within minutes | Long context with caching. Compare its cost with RAG |
| Corpus larger than the win... | [
-0.024026190862059593,
-0.03988271206617355,
-0.03063795529305935,
-0.04449964687228203,
0.05925976112484932,
-0.03722549229860306,
0.0201090220361948,
0.05669548735022545,
0.08017107844352722,
-0.0516267754137516,
0.004056822508573532,
0.03469249606132507,
0.04186559095978737,
-0.05005929... |
71c1542f-7a97-5d93-a085-7bc426a211cd | /documentation/tutorials-operations/embedding-model-migration/ | Migrate to a New Embedding Model | Check Retrieval Quality Before Switching | documentation | false | Migrate to a New Embedding Model > Check Retrieval Quality Before Switching
A new embedding model is not automatically better on your data, so measure it before you move traffic. Both embeddings exist side by side at this point (two collections in a blue-green migration, two named vectors in the other option), which m... | [
0.04494703188538551,
-0.04633359983563423,
-0.03941132128238678,
-0.025698836892843246,
0.051291219890117645,
-0.043917179107666016,
0.07458377629518509,
0.05606774613261223,
0.039051685482263565,
-0.02998620830476284,
0.0035721887834370136,
0.03102741576731205,
0.02999711222946644,
-0.023... |
71c31e21-1b83-5cdd-94df-fb9cbfdbc0cc | /documentation/examples/graphrag-qdrant-neo4j/ | GraphRAG with Qdrant and Neo4j | Implementation > Retrieval & Generation > Querying Neo4j for Related Graph Data | documentation | true | GraphRAG with Qdrant and Neo4j > Implementation > Retrieval & Generation > Querying Neo4j for Related Graph Data
Starting from the chunks the retriever found, fetch the entities they mention and the relationships around them.
```python
def fetch_related_graph(neo4j_client, chunk_ids):
query = """
MATCH (c:Chu... | [
0.011522550135850906,
-0.029969291761517525,
-0.049074746668338776,
-0.04144291207194328,
0.054682131856679916,
-0.050844110548496246,
0.05049506947398186,
0.0707133561372757,
0.044248223304748535,
-0.02737337350845337,
0.040462080389261246,
0.045075517147779465,
0.07087069004774094,
0.012... |
71d3b29d-2f27-538f-87a8-177c970ca2ae | /documentation/edge/edge-api/updating-data/ | Updating Data | Update Operations | documentation | false | Updating Data > Update Operations
| Operation | Parameters | Description |
| --- | --- | --- |
| `upsert_points` | `points`, `condition`, `update_mode` | Insert or update points. |
| `delete_points` | `point_ids` | Delete points by ID. |
| `delete_points_by_filter` | `filter` | Delete every point matching a filter. |
... | [
0.053546760231256485,
-0.03887230530381203,
-0.04852178692817688,
-0.0013787782518193126,
0.06304353475570679,
-0.011798021383583546,
0.05341341719031334,
0.03390054777264595,
0.041031673550605774,
-0.0251043401658535,
0.037282176315784454,
0.04268459230661392,
0.02889403887093067,
-0.0139... |
71dffe6f-c5a3-5c51-8753-b21a30781c3e | /documentation/search-tuning/before-tuning-a-qdrant-collection/ | What to Check Before Tuning a Qdrant Collection | Silent Settings Can Break Quality | documentation | false | What to Check Before Tuning a Qdrant Collection > Silent Settings Can Break Quality
Check the stages you run before tuning anything else. Each prerequisite has an expected state for a given collection and can drift from it without an error. Fix them before you benchmark or compare settings, otherwise you are measuring... | [
0.03965190052986145,
-0.0573960617184639,
-0.023246830329298973,
-0.028756892308592796,
0.03658528998494148,
-0.04538705199956894,
0.019285622984170914,
0.021019481122493744,
0.0444379560649395,
-0.031221244484186172,
0.01578899845480919,
0.011180393397808075,
0.06110279634594917,
-0.03727... |
71e54551-0637-55e3-93c8-5b788b46512c | /documentation/tutorials-develop/async-api/ | Async API | When to use async API > Using Qdrant asynchronously | documentation | true | Async API > When to use async API > Using Qdrant asynchronously
To follow along, you need a Qdrant instance running at `localhost:6333`, as set up in the [Local Quickstart](/documentation/quickstart/), and the client library: `pip install qdrant-client`.
The simplest way to run asynchronous code is to define an `asyn... | [
0.018826445564627647,
-0.033894143998622894,
-0.01394213829189539,
-0.017604803666472435,
0.025646232068538666,
-0.04555455967783928,
0.04374203085899353,
0.016561513766646385,
0.08435104787349701,
-0.012100817635655403,
0.012608185410499573,
0.017646240070462227,
0.04509589076042175,
-0.0... |
71fc056d-b35e-5409-8c5b-af82dcf70b32 | /course/multi-vector-search/module-0/installing-dependencies/ | Installing Dependencies | Python Environment Setup | course | true | Installing Dependencies > Python Environment Setup
### Using uv (Recommended)
For this course, we recommend using [uv](https://docs.astral.sh/uv/), a modern Python package manager that's significantly faster and more reliable than traditional pip. It handles virtual environments and dependencies with better performan... | [
0.020383387804031372,
-0.04368293285369873,
-0.00277272192761302,
-0.021704722195863724,
0.05886528268456459,
-0.04556366801261902,
0.020905109122395515,
0.044059835374355316,
0.060427937656641006,
-0.020419955253601074,
-0.0006350507610477507,
0.03194336220622063,
0.03163909167051315,
-0.... |
72108d24-4307-5c55-bb3b-756bfe763325 | /documentation/security/ | Security | Authentication > Admin API Key | documentation | true | Security > Authentication > Admin API Key
*Available as of v1.2.0*
The admin API key is the primary key that grants full access to all operations and collections. It is configured with the `api_key` setting in the configuration file.
```yaml
service:
# Set an api-key.
# If set, all requests must include a header... | [
0.02337236888706684,
-0.03220684081315994,
-0.06630311161279678,
-0.018505806103348732,
0.041904762387275696,
-0.034315694123506546,
0.05802009999752045,
0.0830780416727066,
0.08057229965925217,
-0.02216355875134468,
0.015803074464201927,
0.008982510305941105,
0.044671040028333664,
-0.0413... |
721207be-9de3-555e-9932-07768d16082c | /articles/what-are-embeddings/ | Vector Embeddings Explained: How They Work in ML & Search | Why use vector embeddings? | article | false | Vector Embeddings Explained: How They Work in ML & Search > Why use vector embeddings?
The **nuances** of natural language or the hidden **meaning** in large datasets of images, sounds, or user interactions are hard to fit into a table. Traditional relational databases can't efficiently query most types of data being ... | [
0.0245245099067688,
-0.04942859336733818,
-0.07567380368709564,
-0.05006073787808418,
0.06753245741128922,
-0.03172777220606804,
0.06533078849315643,
0.04256938770413399,
0.06309840828180313,
-0.01897364854812622,
0.018449563533067703,
0.012962579727172852,
0.06218830496072769,
-0.03408360... |
72139c8b-7819-5acd-9859-26570fa3ac54 | /documentation/tutorials-search-engineering/sparse-embeddings-ecommerce-part-4/ | Fine-Tuning Sparse Embeddings for E-Commerce Search | Part 4: Specialization vs Generalization | The Case for Fine-Tuning | documentation | false | Fine-Tuning Sparse Embeddings for E-Commerce Search | Part 4: Specialization vs Generalization > The Case for Fine-Tuning
Why fine-tune when off-the-shelf models already beat BM25?
**Domain knowledge matters.** Generic models don't know that "AirPods Max" is a specific product, that "prime" means fast shipping, or th... | [
0.019832851365208626,
-0.04274652153253555,
-0.05313188210129738,
-0.046629901975393295,
0.04381116107106209,
-0.026134774088859558,
0.07067091763019562,
0.008317261934280396,
0.05681242048740387,
-0.0076021733693778515,
0.017636161297559738,
0.010489257052540779,
0.06280016154050827,
-0.0... |
72214376-1ba7-5b3f-b493-2887c89e8a37 | /documentation/embeddings/nvidia/ | NVIDIA | Reduce Vector Size | documentation | false | NVIDIA > Reduce Vector Size
`nvidia/nemotron-3-embed-1b` has no `dimensions` parameter and always returns 2048 values. To store smaller vectors, keep the first 1024 or 512 values of each embedding, for example `item.embedding[:1024]`, and set the collection's `size` to match. Shortened vectors must be normalized again... | [
0.03881165012717247,
-0.018170565366744995,
-0.023775875568389893,
-0.06921561807394028,
0.04353172332048416,
-0.03785467892885208,
0.046253036707639694,
0.06848330795764923,
0.08816321939229965,
-0.03579975292086601,
0.03993794694542885,
0.02191382832825184,
0.04996485263109207,
-0.014048... |
72223541-ccb7-5617-91d9-c81da476672d | /documentation/hybrid-cloud/hybrid-cloud-setup/ | Setup Hybrid Cloud | Installation > Generate Installation Command | documentation | false | Setup Hybrid Cloud > Installation > Generate Installation Command
Clicking **Save and Generate Installation Command** generates a script that you can run in your Kubernetes cluster to perform the initial installation of the Kubernetes operator and agent.
It will:
- Create the Kubernetes namespace, if not present.
- ... | [
0.027825819328427315,
-0.02828715369105339,
-0.03914318233728409,
-0.047360219061374664,
0.04706461355090141,
-0.031388554722070694,
0.07210646569728851,
0.0270912554115057,
0.04267441853880882,
-0.0400858111679554,
0.06167197600007057,
0.008117706514894962,
0.06755466759204865,
-0.0539216... |
7229a6be-9c22-51b3-bb03-290472d58336 | /documentation/search-precision/automate-filtering-with-llms/ | LLM-Powered Filter Automation | Search with the generated filter | documentation | true | LLM-Powered Filter Automation > Search with the generated filter
The generated filter is a regular `models.Filter`, so you can pass it straight to `query_points` and combine it with
a semantic query:
```python
results = client.query_points(
collection_name="test_filter",
query=models.Document(text="white T-sh... | [
0.020321032032370567,
-0.02895483933389187,
-0.07824327796697617,
-0.0412648469209671,
0.042209893465042114,
0.010935157537460327,
0.030693773180246353,
0.05014553293585777,
0.06766831874847412,
-0.01160458941012621,
0.05759485065937042,
0.03900766000151634,
0.057764142751693726,
-0.024166... |
722f4d97-9a49-5246-bf26-84607ea7b57c | /course/essentials/day-5/universal-query-demo/ | Demo: Universal Query for Hybrid Retrieval | Step 2: The Universal Query in Action | course | false | Demo: Universal Query for Hybrid Retrieval > Step 2: The Universal Query in Action
Let's build a sophisticated research discovery query step by step. We'll orchestrate dense search, sparse search, RRF fusion, and ColBERT reranking - all in a single API call. | [
0.03364305570721626,
-0.025645114481449127,
-0.06134887784719467,
-0.02775385230779648,
0.07874780148267746,
-0.03237275406718254,
0.041022397577762604,
0.06937793642282486,
0.06270669400691986,
-0.024985408410429955,
0.05765586346387863,
0.047439392656087875,
0.05132044106721878,
0.005338... |
723657b8-f7ce-5cc9-970e-cb3a1f4139cf | /course/beginners/module-4/module-overview/ | Module Overview | course | false | Module Overview
Module 4
So far, rebuilding a collection has taken only a few seconds. Once generating embeddings takes hours, you need to get the design right before ingesting your data.
**Follow-along code**: [Module 4 notebook](https://github.com/qdrant/examples/blob/master/course/beginners/Module4.ipynb)
#### O... | [
0.03880302235484123,
-0.04516971856355667,
-0.03630306199193001,
-0.04071640968322754,
0.09884121268987656,
-0.05642753466963768,
0.025648998096585274,
0.06369037181138992,
0.07401854544878006,
-0.04290161654353142,
0.024385547265410423,
0.04240068420767784,
0.039974041283130646,
-0.039782... | |
724ad05e-a2ba-5e2a-9533-9099438cf88a | /course/multi-vector-search/module-0/installing-dependencies/ | Installing Dependencies | Verification Steps | course | true | Installing Dependencies > Verification Steps
Let's verify that everything is installed correctly.
**Test your imports:**
```python
from qdrant_client import QdrantClient
from fastembed import TextEmbedding
print("All dependencies installed successfully!")
```
**Quick connection test:**
If you set up Qdrant in the... | [
0.02698933705687523,
-0.03876366838812828,
-0.022761918604373932,
-0.03343599662184715,
0.013228814117610455,
-0.04099671170115471,
0.03644973039627075,
0.039040058851242065,
0.07043574005365372,
-0.03476082161068916,
0.023548942059278488,
0.044044189155101776,
0.05521747097373009,
-0.0432... |
724fc87c-7d8a-5d4e-baea-6e884564805f | /documentation/tutorials-build-essentials/video-anomaly-edge-part-2/ | Video Anomaly Detection Part 2: Edge-to-Cloud Pipeline | Mutable Shard Retention > Option B: Time-Windowed Eviction (time_window) | documentation | true | Video Anomaly Detection Part 2: Edge-to-Cloud Pipeline > Mutable Shard Retention > Option B: Time-Windowed Eviction (time_window)
```env
RETENTION_STRATEGY=time_window
RETENTION_SECONDS=1800 # 30 minutes
```
A background worker runs every `RETENTION_CHECK_INTERVAL_S` seconds and deletes points older than `RETENTION... | [
0.01874507963657379,
-0.028559226542711258,
-0.061858128756284714,
-0.04432648792862892,
0.047591131180524826,
-0.02034197747707367,
0.04824545606970787,
-0.021421782672405243,
0.08919910341501236,
-0.041132114827632904,
0.051111962646245956,
0.027235576882958412,
0.05166689306497574,
-0.0... |
725f5768-cafa-56e5-a06f-36595e490dbc | /documentation/tutorials-operations/embedding-model-migration/ | Migrate to a New Embedding Model | Migrate Using Named Vectors > Step 3: Re-Embed Existing Points | documentation | true | Migrate to a New Embedding Model > Migrate Using Named Vectors > Step 3: Re-Embed Existing Points
Searches and new inserts can continue during backfill. Pause updates and deletes to existing points, and wait for those in-flight operations to finish before starting. A concurrent update could change the payload after it... | [
0.031145289540290833,
-0.05520705506205559,
-0.013309831731021404,
-0.015592309646308422,
0.04897172376513481,
-0.03200286254286766,
0.043132681399583817,
0.025358883664011955,
0.03674915060400963,
-0.019742131233215332,
0.03600911423563957,
0.029637617990374565,
0.007913542911410332,
-0.0... |
7268b1cf-ab6e-5894-a216-848f8c14fb96 | /articles/universal-io/ | Universal I/O in Qdrant | Neither Async nor Blocking Fits | article | false | Universal I/O in Qdrant > Neither Async nor Blocking Fits
Because of this gap in cost, the design of the interface seemed to be limited to two approaches, and each with a clear downside:
- **Make everything async.** This is a great fit for network storage, but a poor fit for `mmap` and for data that is already cached... | [
0.03191065043210983,
-0.019861755892634392,
-0.07699228078126907,
-0.01029675081372261,
0.0691424012184143,
-0.020472459495067596,
0.05674408748745918,
0.016780197620391846,
0.058327510952949524,
0.014355954714119434,
0.03806670755147934,
0.0015066358027979732,
0.06713353842496872,
0.00324... |
7276b017-c336-5bc3-a12f-abbd99f8dda2 | /documentation/search-tuning/hybrid-search/ | Hybrid Search in Qdrant | What to Test Next | documentation | false | Hybrid Search in Qdrant > What to Test Next
- **Add more stages.** [Multi-stage queries](/documentation/search/hybrid-queries/#multi-stage-queries) retrieve with a cheap representation and rescore with an expensive one. Cross-encoder reranking puts the query and chunk into a model together. [When Is a Reranker Worth I... | [
0.01307108998298645,
-0.04668588191270828,
-0.032657042145729065,
-0.05052769184112549,
0.05563554912805557,
-0.019693100824952126,
0.026913540437817574,
0.01427762396633625,
0.06971851736307144,
-0.020042037591338158,
0.027920089662075043,
-0.0015326774446293712,
0.06448110938072205,
-0.0... |
728596e8-dd4c-5c8d-97f8-6fb756de1945 | /documentation/examples/graphrag-qdrant-neo4j/ | GraphRAG with Qdrant and Neo4j | RAG & Its Challenges | documentation | false | GraphRAG with Qdrant and Neo4j > RAG & Its Challenges
[RAG](https://qdrant.tech/rag/) combines retrieval-based and generative AI to enhance LLMs with relevant, up-to-date information from a knowledge base, like a vector search engine. However, RAG faces several challenges:
1. **Understanding Context:** Models may mis... | [
0.01102779246866703,
-0.054816581308841705,
-0.0707508772611618,
-0.0606742724776268,
0.07333636283874512,
-0.047197431325912476,
0.0566241592168808,
0.03324291110038757,
0.03492517024278641,
-0.038608111441135406,
0.02824259176850319,
0.02279972843825817,
0.07892847061157227,
-0.022605946... |
728a06c1-585b-565d-b34f-c082a340e722 | /articles/io_uring/ | Faster Disk I/O for Vector Search Using io_uring | And what about Qdrant? | article | false | Faster Disk I/O for Vector Search Using io_uring > And what about Qdrant?
Qdrant can store everything in memory, but not all data sets may fit, which can
require storing on disk. Before io\_uring, Qdrant used mmap to do its IO. This
led to some modest overhead in case of disk latency. The kernel may
stop a user threa... | [
0.041196081787347794,
-0.04367554932832718,
-0.03511018678545952,
-0.012677338905632496,
0.08259836584329605,
-0.03537992388010025,
0.04594393074512482,
-0.01594443805515766,
0.0946706011891365,
-0.0049435230903327465,
0.002838752930983901,
-0.018594233319163322,
0.08242031186819077,
-0.00... |
729385a5-6b72-535c-9251-e40e789f7436 | /documentation/tutorials-build-essentials/agentic-rag-crewai-zoom/ | Agentic RAG with CrewAI | What You'll Build > Architecture | documentation | false | Agentic RAG with CrewAI > What You'll Build > Architecture
The system is built on three main components:
- **Qdrant Vector Database**: Stores meeting transcripts and summaries as vector embeddings, enabling semantic search
- **CrewAI Framework**: Coordinates AI agents that handle different aspects of meeting analysis
... | [
0.015103559009730816,
-0.030407404527068138,
-0.049660202115774155,
-0.03362704813480377,
0.09624655544757843,
-0.04853373020887375,
0.05626607686281204,
0.033302344381809235,
0.053294334560632706,
-0.03719016909599304,
0.034588370472192764,
-0.03476238623261452,
0.07100605964660645,
-0.03... |
72a06def-a9f5-5afe-9c64-837d83d0b887 | /articles/relevance-feedback/ | Relevance Feedback in Qdrant | Vector Search Optimization > What is Relevance Feedback | article | false | Relevance Feedback in Qdrant > Vector Search Optimization > What is Relevance Feedback
**Relevance feedback** distills signals from current search results into the next retrieval iteration to surface more relevant documents.
It can be provided by a human or a model, be binary (relevant–irrelevant) or granular, rescor... | [
0.01166237611323595,
-0.03959919139742851,
-0.0456092394888401,
-0.06169818341732025,
0.04991106316447258,
-0.03774387389421463,
0.04267681762576103,
0.029627816751599312,
0.08457773923873901,
-0.0030306940898299217,
0.03491421788930893,
0.022297583520412445,
0.07099329680204391,
-0.023730... |
72af0322-a83e-5b02-8c96-ebd73de39c19 | /documentation/ops-optimization/read-write-contention/ | Read-Write Contention | Step 3: Lower the Optimizer's CPU Budget | documentation | true | Read-Write Contention > Step 3: Lower the Optimizer's CPU Budget
By default, Qdrant auto-selects the number of CPU threads the optimizer can use, typically leaving one or two cores free and giving the rest to the optimizer. Under write pressure, this can leave little headroom for queries. Use the `optimizer_cpu_budget... | [
0.00942794419825077,
-0.049484409391880035,
-0.04960716515779495,
-0.020485395565629005,
0.05586723983287811,
-0.028088629245758057,
0.006113118026405573,
0.03376102074980736,
0.09282052516937256,
-0.041480980813503265,
0.036565374583005905,
-0.015126209706068039,
0.02483363449573517,
-0.0... |
72c51100-bb37-5403-be05-7132194a6647 | /documentation/scaling/resilience/ | Resilience | Temporary Node Failure | documentation | false | Resilience > Temporary Node Failure
Running Qdrant in distributed mode makes your cluster resistant to outages when one node fails temporarily. Here's how differently-configured Qdrant clusters respond:
- 1-node clusters: All operations time out or fail for up to a few minutes, depending on how long it takes to resta... | [
0.045317087322473526,
-0.025326533243060112,
-0.062224406749010086,
-0.04032211750745773,
0.0886770635843277,
-0.020988401025533676,
0.047218337655067444,
0.0026403774973005056,
0.08779233694076538,
-0.00902337022125721,
0.04332592710852623,
-0.012756816111505032,
0.04539002850651741,
-0.0... |
72d24a37-3cd9-54df-a7d7-56be4a92546c | /documentation/tutorials-operations/embedding-model-migration/ | Migrate to a New Embedding Model | Two Options > Option 2: Named Vectors | documentation | false | Migrate to a New Embedding Model > Two Options > Option 2: Named Vectors
The [named vectors approach](#migrate-using-named-vectors) keeps everything in a single collection. Start by [adding the new model as an additional named vector](/documentation/manage-data/collections/#update-vector-schema): a schema-only operati... | [
0.0569649264216423,
-0.032299771904945374,
-0.03483213484287262,
-0.006822798401117325,
0.038218021392822266,
-0.00976184569299221,
0.03606776148080826,
0.02592642791569233,
0.054354727268218994,
-0.012233392335474491,
0.015931488946080208,
0.03562133386731148,
0.013268924318253994,
-0.012... |
72e9efa3-5595-53e3-968c-919dde3ae293 | /documentation/security/ | Security | Authentication > Table of Access | documentation | false | Security > Authentication > Table of Access
| Action | manage | read-only | collection read-write | collection read-only |
|--------|--------|-----------|----------------------|-----------------------|
| delete collection snapshot | ✅ | ❌ | ✅ | ❌ |
| download collection snapshot | ✅ | ✅ | ✅ | ✅ |
| upload collection s... | [
0.01825053058564663,
-0.023719072341918945,
-0.04748394712805748,
-0.019994661211967468,
0.03154102712869644,
-0.01561217661947012,
0.03236302360892296,
0.06278283149003983,
0.06109466403722763,
-0.03729432821273804,
0.06532814353704453,
0.03317924588918686,
0.04023054242134094,
-0.0339996... |
72ebfca3-574c-520a-a911-7753b5e905f4 | /documentation/ops-configuration/administration/ | Administration | Strict Mode > Maximum Vectors per Multivector | documentation | true | Administration > Strict Mode > Maximum Vectors per Multivector
A multivector with many vectors per point is expensive to store, index and query.
Setting `multivector_config` caps the maximum number of vectors per multivector for each named vector.
```python
from qdrant_client import QdrantClient, models
client = Qd... | [
-0.007501936983317137,
-0.04148275405168533,
-0.042139820754528046,
-0.012299194000661373,
0.01579361781477928,
-0.0288022980093956,
0.04768069460988045,
0.03779090940952301,
0.08468935638666153,
-0.029058611020445824,
-0.0035092178732156754,
0.021319428458809853,
0.06109276041388512,
-0.0... |
731ea5ce-9190-53f4-8074-482bd92ba2ea | /documentation/scaling/horizontal-scaling/ | Horizontal Scaling | How Many Qdrant Nodes Should I Run? > One Node | documentation | false | Horizontal Scaling > How Many Qdrant Nodes Should I Run? > One Node
One node gives you the lowest cost and is easy to set up, but it offers no high availability. This is not recommended for production environments. Drawbacks:
- Resilience: Users will experience downtime during node restarts. If the data on the node ... | [
0.009540441446006298,
-0.046792738139629364,
-0.033936578780412674,
-0.03117215447127819,
0.03061077743768692,
-0.0331486240029335,
0.08137048035860062,
0.04404636472463608,
0.07989255338907242,
-0.02096257545053959,
0.021329687908291817,
-0.001637425972148776,
0.050633225589990616,
-0.028... |
73218a8f-ac5c-5169-a595-de1167660f73 | /documentation/search/filtering/ | Filtering | Filtering Conditions > Geo | documentation | true | Filtering > Filtering Conditions > Geo
#### Geo Bounding Box
```python
models.FieldCondition(
key="location",
geo_bounding_box=models.GeoBoundingBox(
bottom_right=models.GeoPoint(
lon=13.455868,
lat=52.495862,
),
top_left=models.GeoPoint(
lon=13.4036... | [
-0.002936153905466199,
-0.03443983569741249,
-0.028467752039432526,
-0.0254523903131485,
0.0554860383272171,
-0.018065476790070534,
0.07091369479894638,
0.06587089598178864,
0.1096605509519577,
-0.024798011407256126,
-0.004355677403509617,
0.03734875097870827,
0.05067139118909836,
0.011869... |
73368148-99a3-5d38-a435-08ea854d7c51 | /course/beginners/module-6/module-overview/ | Module Overview | course | false | Module Overview
Module 6
Modules 1 through 5 showed you how to design and build a complete retrieval pipeline. This bonus module covers the next layer: measuring and improving its results.
#### Overview
> You'll match common search problems to Qdrant features: score boosting and reranking for order, Maximal Margina... | [
0.025281997397542,
-0.04895934835076332,
-0.05031343922019005,
-0.01332035381346941,
0.0753515288233757,
-0.06080739572644234,
0.02972484566271305,
0.05876154080033302,
0.06738340854644775,
-0.024397052824497223,
-0.003710284596309066,
0.03384539857506752,
0.05882349982857704,
-0.033999569... | |
73420a99-3bf3-51d3-aa94-9075f9273c07 | /articles/immutable-data-structures/ | Immutable Data Structures in Qdrant | How Does Immutability Help? | article | false | Immutable Data Structures in Qdrant > How Does Immutability Help?
For more info on payload-aware connections in HNSW, read our [previous article](/articles/filterable-hnsw/).
This time around, we will focus on the latest additions to Qdrant:
- **the immutable hash map with perfect hashing**
- **defragmented vector ... | [
0.056758664548397064,
-0.03803182393312454,
-0.06446296721696854,
-0.055394772440195084,
0.03406728059053421,
-0.0644201785326004,
0.07938015460968018,
0.03817244991660118,
0.0688670203089714,
-0.009134933352470398,
-0.011283270083367825,
0.011699937283992767,
0.035085540264844894,
-0.0182... |
734a26aa-80a9-5ff8-a7c1-425c14ee7142 | /articles/what-is-a-vector-database/ | What is a Vector Database? | Data Security and Access Control | article | true | What is a Vector Database? > Data Security and Access Control
A common security risk in vector databases is the possibility of **embedding inversion attacks**, where attackers could reconstruct the original data from embeddings. There are many layers of protection you can use to secure your instance that are very impo... | [
0.039919864386320114,
-0.04840393364429474,
-0.047421883791685104,
-0.06176600977778435,
0.039454810321331024,
-0.02732478268444538,
0.06734003871679306,
0.021903574466705322,
0.0854230672121048,
-0.029846003279089928,
0.012785174883902073,
-0.001987810479477048,
0.059757016599178314,
-0.0... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.