Merge pull request #1 from KShivendu/feat/refactor

refactor: Use static dir for common images
KShivendu · May 11, 2024 · cef0015 · cef0015
2 parents ca93aea + 7855071
commit cef0015
Show file tree

Hide file tree

Showing 17 changed files with 963 additions and 0 deletions.
diff --git a/qdrant-future-of-search/imgs/dragon-search.png b/qdrant-future-of-search/imgs/dragon-search.png
diff --git a/qdrant-future-of-search/imgs/intro-bg.png b/qdrant-future-of-search/imgs/intro-bg.png
diff --git a/qdrant-future-of-search/imgs/mystical-creature.jpg b/qdrant-future-of-search/imgs/mystical-creature.jpg
diff --git a/qdrant-future-of-search/imgs/search-space.png b/qdrant-future-of-search/imgs/search-space.png
diff --git a/qdrant-future-of-search/imgs/similarity-bubble-1.png b/qdrant-future-of-search/imgs/similarity-bubble-1.png
diff --git a/qdrant-future-of-search/imgs/similarity-bubble-2.png b/qdrant-future-of-search/imgs/similarity-bubble-2.png
diff --git a/qdrant-future-of-search/index.html b/qdrant-future-of-search/index.html
diff --git a/qdrant-future-of-search/index.md b/qdrant-future-of-search/index.md
@@ -0,0 +1,278 @@
+---
+theme: gaia
+_class: lead
+style: |
+  :root {
+    --color-background: #fff !important;
+    --color-foreground: #182b3a !important;
+    --color-highlight: #bc1439 !important;
+    --color-dimmed: #888 !important;
+    border-top: 4px solid var(--color-highlight);
+  }
+
+marp: true
+inlineSVG: true
+# paginate: true
+---
+
+<!-- ![bg](./imgs/cover-foss-united-jan-2024.png) -->
+![bg](./imgs/intro-bg.png)
+
+
+---
+
+## $ whoami
+
+![bg right:40% 80%](/static/shivendu.jpg)
+
+* Kumar Shivendu
+
+* Software Engineer @ Qdrant
+
+* I ❤️ information retrieval, performance, and data mining.
+
+* First talk!
+
+* Qdrant: Future of search and beyond
+
+<!-- ---
+
+## Topics to cover
+
+* Evolution of search
+  * Traditional approaches and drawbacks
+  * Vectors search and how it harnesses ML models
+
+* Qdrant
+  * Building HNSW index and vector search
+  * Beyond similarity search:
+    * Recommendations
+    * Discovery
+  * Sparse vectors -->
+
+---
+
+## Traditional ways to build search
+
+* Approaches:
+  * Keyword match, Regex, Boolean operators
+  * Extracting metadata using NLP and CV
+  * Knowledge graphs, Vocabs
+
+* Challenges:
+  * Effort to maintain the knowledge
+  * Growth of unstructured data
+  * Multimodal search remains hard: Text, Image, Audio, Video
+
+---
+
+## Vectors
+
+* Points in an N-dimensional space
+* Anything -> Vector
+* Generated from:
+  * ML models
+  * Metric learning
+    * CLIP
+
+<!-- Replace search space image -->
+![bg vertical right:50% 90%](./imgs/search-space.png)
+![bg right:50% 90%](/static/clip-model.png)
+
+
+---
+
+## Vector search
+
+
+![bg right:50% 50%](/static/lens-reverse-image.png)
+
+* Nearest points
+* Example: Google Lens
+* But this is expensive and not easy to scale
+* Solution: Indexing and approximation
+
+<!-- Image showing vector search -->
+
+---
+
+## What is **Qdrant**
+
+![bg right:40% 50%](/static/qdrant.png)
+
+* Vector Search Engine (aka Vector DB)
+
+* 15k+ stars on Github
+
+* Written in Rust 🦀
+
+* SDKs for Python, JS, Go, Java, etc
+
+* Twitter, Canva, Meesho, Flipkart
+
+---
+
+## The HNSW Index
+
+![bg right:50% 100%](/static/hnsw-layers.png)
+
+* Skip Lists + Graphs
+* Approximate and Tunable
+* Filter during search
+* Quantization
+
+---
+
+## Running search:
+
+```js
+POST /collections/rentals/points/search
+{
+  "query": [0.2, 0.3, 0.4, 0.5], // vector generated from image/text/video
+  "filter": { "must": [{"key": "locality", "match": {"value": "Indiranagar"}}] },
+  "limit": 10
+}
+```
+
+* ```js
+  [
+    {"id": 4, "score": 0.56, "payload": {...}},
+    {"id": 2, "score": 0.40, "payload": {...}},
+    {"id": 5, "score": 0.23, "payload": {...}},
+  ]
+  ```
+
+---
+
+## Beyond search: Recommendations
+
+* Realtime addition of points is possible.
+* `average_vector` and `best_score`
+* ```js
+  POST /collections/rentals/points/recommend
+  {
+    "positive": [100, 231], // vector ids
+    "negative": [718, [0.2, 0.3, 0.4, 0.5]], // vector id and vector
+    "filter": { "must": [{"key": "locality", "match": {"value": "Indiranagar"}}] },
+    "strategy": "best_score",
+  }
+  ```
+* DailyMotion (Qdrant), Spotify (Annoy)
+
+---
+## How to find this thing on the internet?
+
+![bg right:35%](./imgs/mystical-creature.jpg)
+
+* No reverse image search
+* No known name
+
+---
+## Strategy One
+
+* Describe the thing
+  * "Combination of human, dragon and chicken"
+  * "Mythology creature of human and dragon"
+
+![bg right:50%](./imgs/dragon-search.png)
+
+---
+## Strategy Two
+
+* Search for similar images
+  * Similarity bubble
+
+![bg right:60%](./imgs/similarity-bubble-1.png)
+![bg](./imgs/similarity-bubble-2.png)
+
+---
+
+## Beyond search: Discovery API
+
+* Unique iterative search by Qdrant
+* Combine multi-modal vectors in single query
+
+* ```js
+  POST collections/my-collection/points/discover
+  {
+    "target": [0.63, 0.10, 0.91, 0.55],
+    "context": [
+      {
+        "positive": 7125, // <-- ID of the example
+        "negative": 122   // <-- This can also be a vector
+      }
+    ],
+  }
+  ```
+
+---
+<!-- ---
+
+## How discovery uses that?
+
+##### Remember Metric Learning?
+
+
+![bg 80%](./imgs/triplet-loss.png)
+
+---
+-->
+
+
+<!-- ![bg 95%](./imgs/discovery-context.png)
+
+![bg 90%](./imgs/context-pairs.png)
+
+![bg 90%](./imgs/context-with-target.png)
+
+--- -->
+
+<!--
+## How multi-modal embeddings look like?
+
+
+![bg 90%](./imgs/cross-modal-space.png)
+
+
+
+---
+
+![bg](./imgs/clip-discovery.png)
+
+
+---
+
+
+![bg](./imgs/complex-context-search.png)
+
+--- -->
+
+<!-- ## Sparse vectors
+
+* VS text search
+* BM25 & TF-IDF
+* Transformer's attention weights
+* SPLADE
+
+![bg right:60% 90%](./imgs/sparse-vectors.png)
+
+--- -->
+
+# Summary
+
+* Anything => vector
+
+* Vectors >> similarity search
+
+* Thousands of use-cases with Qdrant
+
+<!-- * Navigating search (read vector) space is powerful! -->
+
+* Find me at
+  * kshivendu.dev/bio
+  * kshivendu.dev/twitter
+
+![bg right:50% 50%](/static/linkedin-qr.png)
+
+<!-- * Thank you! -->
+
diff --git a/reranking/imgs/rrf.jpg b/reranking/imgs/rrf.jpg
diff --git a/reranking/presentation.html → reranking/index.html b/reranking/presentation.html → reranking/index.html
diff --git a/reranking/presentation.md → reranking/index.md b/reranking/presentation.md → reranking/index.md
diff --git a/static/clip-model.png b/static/clip-model.png
diff --git a/static/hnsw-layers.png b/static/hnsw-layers.png
diff --git a/static/lens-reverse-image.png b/static/lens-reverse-image.png
diff --git a/static/linkedin-qr.png b/static/linkedin-qr.png
diff --git a/static/qdrant.png b/static/qdrant.png
diff --git a/static/shivendu.jpg b/static/shivendu.jpg