Merge branch 'chore/customer_quivr' of github.com:supabase/supabase into chore/customer_quivr

apps/www/_blog/2023-02-03-openai-embeddings-postgres-vector.mdx

@@ -378,11 +378,13 @@ The OpenAI API supports [completion streaming](https://platform.openai.com/docs/
 
 Storing embeddings in Postgres opens a world of possibilities. You can combine your search function with telemetry functions, add an user-provided feedback (thumbs up/down), and make your search feel more integrated with your products.
 
-The [pgvector extension](https://supabase.com/docs/guides/ai/vector-columns) is available on all new Supabase projects today. If you want to try it out, launch a new Postgres database today: [database.new](https://database.new)
+The [pgvector extension](https://supabase.com/docs/guides/ai/vector-columns) is available on all new Supabase projects today. To try it out, launch a new Postgres database: [database.new](https://database.new)
 
-## More pgvector and ChatGPT resources
+## More pgvector and AI resources
 
 - [Supabase Clippy: ChatGPT for Supabase Docs](https://supabase.com/blog/chatgpt-supabase-docs)
-- [A ChatGPT Plugins Template built with Supabase Edge Runtime](https://supabase.com/blog/building-chatgpt-plugins-template)
-- [Template for building your own custom ChatGPT style doc search](https://github.com/supabase-community/nextjs-openai-doc-search)
-- [Supabase + LangChain Starter Template](https://blog.langchain.dev/langchain-x-supabase/)
+- [Hugging Face is now supported in Supabase](https://supabase.com/blog/hugging-face-supabase)
+- [How to build ChatGPT Plugin from scratch with Supabase Edge Runtime](https://supabase.com/blog/building-chatgpt-plugins-template)
+- [Docs pgvector: Embeddings and vector similarity](https://supabase.com/docs/guides/database/extensions/pgvector)
+- [Choosing Compute Add-on for AI workloads](https://supabase.com/docs/guides/ai/choosing-compute-addon)
+- [pgvector v0.5.0: Faster semantic search with HNSW indexes](https://supabase.com/blog/pgvector-v0-5-0-hnsw)

apps/www/_blog/2023-05-15-building-chatgpt-plugins-template.mdx

@@ -274,8 +274,7 @@ In a next step you can add authentication to your plugin, let us know on [Twitte
 
 ## More AI resources
 
+- [Hugging Face is now supported in Supabase](https://supabase.com/blog/hugging-face-supabase)
+- [pgvector v0.5.0: Faster semantic search with HNSW indexes](https://supabase.com/blog/pgvector-v0-5-0-hnsw)
 - [OpenAI ChatGPT Plugin docs](https://platform.openai.com/docs/plugins/introduction)
-- [Advanced plugin template](https://github.com/openai/chatgpt-retrieval-plugin)
-- Learn to build your own ChatGPT-style docs search with [Deno Fresh](https://deno.com/blog/build-chatgpt-doc-search-with-supabase-fresh) or [Next.js](https://vercel.com/templates/next.js/nextjs-openai-doc-search-starter)
-- [How we builtChatGPT for the Supabase Docs](https://supabase.com/blog/chatgpt-supabase-docs).
 - [Docs pgvector: Embeddings and vector similarity](https://supabase.com/docs/guides/database/extensions/pgvector)

apps/www/_blog/2023-07-13-pgvector-performance.mdx

@@ -15,6 +15,14 @@ image: 2023-07-13-pgvector-performance/vector-benchmarks-og.jpeg
 thumb: 2023-07-13-pgvector-performance/vector-benchmarks-thumb.jpeg
 ---
 
+<div className="bg-gray-300 rounded-lg px-6 py-2 bold">
+
+🚀 The incorporation of the HNSW index in pgvector v0.5.0 ensures lightning-fast vector searches. We tested it and benchmarked and share everything.
+[Read the new post](https://supabase.com/blog/pgvector-v0-5-0-hnsw)
+
+</div>
+pgvector v0.5.0: Faster semantic search with HNSW indexes
+
 There are a few pgvector benchmarks floating around the internet, most recently a [pgvector vs Qdrant](https://nirantk.com/writing/pgvector-vs-qdrant/) comparison by NirantK. We wanted to reproduce (or improve!) the results.
 
 There is an obvious bias here: we're a Postgres company. It's not our goal to prove that pgvector is better than Qdrant for running vector workloads. From everything we hear about Qdrant, it's fantastic.
@@ -87,7 +95,7 @@ Let's start with NirantK's results as a baseline:
   />
 </div>
 
-They aren't very flattering! Repeating our statements above, these benchmarks are using the defaults for both engines. Our goal now is to replicate the results, and then see what improvements need to be made as developers scale up their workload.
+They aren't very flattering! Repeating our statements above, these benchmarks use the defaults for both engines. Our goal now is to replicate the results, and then see what improvements need to be made as developers scale up their workload.
 
 ## Results
 
@@ -301,7 +309,8 @@ pgvector is still early in development. As with any open source tool, it needs t
 
 What's next on the roadmap? Andrew has an impressive list of features [planned for v0.5.0](https://github.com/pgvector/pgvector/issues/27):
 
-- Adding HNSW: an index with better speed & precision than IVFFlat (at a higher memory cost)
+✅ Adding HNSW: an index with better speed & precision than IVFFlat (at a higher memory cost)
+
 - Product quantization: better storage for IVFFLAT, improving speed and recall
 - Parallel index builds: building your IVFFLAT indexes will be much faster
 
@@ -311,3 +320,4 @@ What's next on the roadmap? Andrew has an impressive list of features [planned f
 - [Docs pgvector: Embeddings and vector similarity](https://supabase.com/docs/guides/database/extensions/pgvector)
 - [pgvector vs Qdrant](https://nirantk.com/writing/pgvector-vs-qdrant)
 - [Choosing Compute Add-on for AI workloads](https://supabase.com/docs/guides/ai/choosing-compute-addon)
+- [pgvector v0.5.0: Faster semantic search with HNSW indexes](https://supabase.com/blog/pgvector-v0-5-0-hnsw)

apps/www/_blog/2023-08-03-fewer-dimensions-are-better-pgvector.mdx

@@ -148,7 +148,7 @@ Ultimately the choice of embedding model will depend on your specific use case a
 - **Dimension size:** What is the operational cost of storing and comparing embeddings?
 - **Language:** Which written languages do you need to support? Many of the smaller models only support English.
 
-The goal is to maximize similarity performance and sequence length, while minimizing model size and dimension size. Notably, if we can find a model that performs well while producing as few dimensions as possible, we can immediately improve many of the above challenges developers are facing with pgvector.
+The goal is to maximize similarity performance and sequence length while minimizing model size and dimension size. Notably, if we can find a model that performs well while producing as few dimensions as possible, we can immediately improve many of the above challenges developers are facing with pgvector.
 
 For example [gte-small](https://huggingface.co/thenlper/gte-small), a model recently trained by the [Alibaba DAMO Academy,](https://damo.alibaba.com/) produces only 384 dimensions while ranking higher than `text-embedding-ada-002` on the MTEB leaderboard.
 
@@ -162,10 +162,18 @@ It's important to keep in mind that using an alternative embedding model doesn't
 
 While dimensionality reduction techniques like Principal Component Analysis ([PCA](https://en.wikipedia.org/wiki/Principal_component_analysis)) or t-Distributed Stochastic Neighbor Embedding ([t-SNE](https://en.wikipedia.org/wiki/T-distributed_stochastic_neighbor_embedding)) may seem like a good alternative to trim down the size of text embedding vectors, there is some risk to consider.
 
-First, these tools may not appropriately model the relationships in high dimensional space. Both PCA and t-SNE are likely to oversimplify the multivariate relationships among different dimensions, leading to loss of important semantic information. This is particularly the case with PCA, which assumes linearity and would fail to preserve nonlinear interactions found in high-dimensional text embeddings.
+First, these tools may not appropriately model the relationships in high-dimensional space. Both PCA and t-SNE are likely to oversimplify the multivariate relationships among different dimensions, leading to the loss of important semantic information. This is particularly the case with PCA, which assumes linearity and would fail to preserve nonlinear interactions found in high-dimensional text embeddings.
 
 Additionally, the validity of the dimensionality reduction models could change over time if the distribution of text data shifts. If the model was trained on older data, it might not accurately reflect or accommodate these shifts. For example, if your company gets a large new customer in the legal field and your embedding dataset contained no legal documents when the dimensionality reduction model was trained, otherwise exploitable structure may be destructively compressed. Both PCA and t-SNE are not inherently adaptable to such changes, which could result in an increasingly poor fit over time. In summary, the derived low-dimensional representations may not maintain semantic coherence of the original high-dimensional vectors.
 
 ## What's next?
 
 At Supabase we are working on ways to make it easy to generate embeddings using open source, high performance, and low dimension embedding models. Stay tuned for [Launch Week 8](/launch-week) next week!
+
+## More pgvector and AI resources
+
+- [Hugging Face is now supported in Supabase](https://supabase.com/blog/hugging-face-supabase)
+- [How to build ChatGPT Plugin from scratch with Supabase Edge Runtime](https://supabase.com/blog/building-chatgpt-plugins-template)
+- [Docs pgvector: Embeddings and vector similarity](https://supabase.com/docs/guides/database/extensions/pgvector)
+- [Choosing Compute Add-on for AI workloads](https://supabase.com/docs/guides/ai/choosing-compute-addon)
+- [pgvector v0.5.0: Faster semantic search with HNSW indexes](https://supabase.com/blog/pgvector-v0-5-0-hnsw)

apps/www/_blog/2023-09-05-pgvector-v0-5-0-hnsw.mdx

@@ -0,0 +1,264 @@
+---
+title: 'pgvector v0.5.0: Faster semantic search with HNSW indexes'
+description: Increase performance in pgvector using HNSW indexes
+author: egor_romanov,gregnr
+categories:
+  - engineering
+tags:
+  - AI
+  - performance
+  - postgres
+  - planetpg
+date: '2023-09-05'
+toc_depth: 3
+image: 2023-09-05-pgvector-v0-5-0-hnsw/pgvector-v0-5-0-og.png
+thumb: 2023-09-05-pgvector-v0-5-0-hnsw/pgvector-v0-5-0-thumb.png
+---
+
+[Supabase Vector](https://supabase.com/vector) is about to get a lot faster. Starting today, new Supabase databases will ship with pgvector v0.5.0 which adds a new type of index: Hierarchical Navigable Small World (HNSW).
+
+HNSW is an algorithm for approximate nearest neighbor search, often used in high-dimensional spaces like those found in embeddings.
+
+With this update, you can take advantage of the new HNSW index on your column using the following:
+
+```sql
+-- Add a HNSW index for the inner product distance function
+CREATE INDEX ON documents
+USING hnsw (embedding vector_ip_ops);
+```
+
+<aside className="bg-gray-300 rounded-lg px-6 py-2 bold">
+
+💡 If you have an existing database that was created before this release, please reach out to [support](https://supabase.com/dashboard/support/new) and we'll assist you in an upgrade to pgvector v0.5.0.
+
+</aside>
+
+## How does HNSW work?
+
+Compared to inverted file (IVF) indexes which use [clusters](https://supabase.com/docs/guides/ai/managing-indexes#ivfflat-indexes) to approximate nearest-neighbor search, HNSW uses proximity graphs (graphs connecting nodes based on distance between them). To understand HNSW, we can break it down into 2 parts:
+
+- **Hierarchical (H):** The algorithm operates over multiple layers
+- **Navigable Small World (NSW):** Each vector is a node within a graph and is connected to several other nodes
+
+### Hierarchical
+
+The hierarchical aspect of HNSW builds off of the idea of skip lists.
+
+Skip lists are multi-layer linked lists. The bottom layer is a regular linked list connecting an ordered sequence of elements. Each new layer above removes some elements from the underlying layer (based on a fixed probability), producing a sparser subsequence that “skips” over elements.
+
+<div>
+  <img
+    alt="visual of an example skip list (light)"
+    className="dark:hidden"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/pgvector-v0-5-0-skip-list--light.png"
+  />
+  <img
+    alt="visual of an example skip list (dark)"
+    className="hidden dark:block"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/pgvector-v0-5-0-skip-list--dark.png"
+  />
+</div>
+
+When searching for an element, the algorithm begins at the top layer and traverses its linked list horizontally. If the target element is found, the algorithm stops and returns it. Otherwise if the next element in the list is greater than the target (or NIL), the algorithm drops down to the next layer below. Since each layer below is less sparse than the layer above (with the bottom layer connecting all elements), the target will eventually be found. Skip lists offer O(log n) average complexity for both search and insertion/deletion.
+
+### Navigable Small World
+
+A navigable small world (NSW) is a special type of proximity graph that also includes long-range connections between nodes. These long-range connections support the “small world” property of the graph, meaning almost every node can be reached from any other node within a few hops. Without these additional long-range connections, many hops would be required to reach a far-away node.
+
+<Img
+  alt="visual of an example navigable small world graph"
+  src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/pgvector-v0-5-0-nsw.png"
+/>
+
+The “navigable” part of NSW specifically refers to the ability to logarithmically scale the greedy search algorithm on the graph, an algorithm that attempts to make only the locally optimal choice at each hop. Without this property, the graph may still be considered a small world with short paths between far-away nodes, but the greedy algorithm tends to miss them. Greedy search is ideal for NSW because it is quick to navigate and has low computational costs.
+
+### **Hierarchical +** Navigable Small World
+
+HNSW combines these two concepts. From the hierarchical perspective, the bottom layer consists of a NSW made up of short links between nodes. Each layer above “skips” elements and creates longer links between nodes further away from each other.
+
+Just like skip lists, search starts at the top layer and works its way down until it finds the target element. However instead of comparing a scalar value at each layer to determine whether or not to descend to the layer below, a multi-dimensional distance measure (such as Euclidean distance) is used.
+
+## HNSW performance: 1536 dimensions
+
+To understand the performance improvements that HNSW offers, we decided to expand upon our previous benchmarks and include results for the HNSW index in addition to IVF and compare the queries per second (QPS) for both.
+
+### wikipedia-en-embeddings
+
+In the first test, we used [224,482 vectors by OpenAI](https://huggingface.co/datasets/Supabase/wikipedia-en-embeddings) (1536 dimensions). You can find our previous benchmark with additional information on how vector dimensions may affect performance in [pgvector: Fewer dimensions are better](https://supabase.com/blog/pgvector-performance).
+
+In this test, we used a Supabase project with a large compute add-on (2-core CPU and 8GB RAM) and built the HNSW index with the following parameters:
+
+- `m`: 32
+- `ef_construction`: 64
+
+<div>
+  <img
+    alt="wikipedia embeddings comparing ivfflat and hnsw queries-per-second (light)"
+    className="dark:hidden"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/wikipedia-ivfflat-vs-hnsw--light.png"
+  />
+  <img
+    alt="wikipedia embeddings comparing ivfflat and hnsw queries-per-second (dark)"
+    className="hidden dark:block"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/wikipedia-ivfflat-vs-hnsw--dark.png"
+  />
+</div>
+
+HNSW has 3 times better performance than IVFFlat and with better precision.
+
+### dbpedia-entities-openai-1M
+
+Next, we took the setup from our benchmarks with [1,000,000 vectors by OpenAI](https://huggingface.co/datasets/KShivendu/dbpedia-entities-openai-1M) (1536 dimensions). If you want to find out more about pgvector 0.4.0 IVFFlat performance and our load testing methodology, check out [pgvector 0.4.0 performance](https://supabase.com/blog/pgvector-performance) blogpost.
+
+Here we used the Supabase project with a 2XL compute add-on (8-core CPU and 32GB RAM) and built the HNSW index with the same parameters:
+
+- `m`: 24
+- `ef_construction`: 56
+
+<div>
+  <img
+    alt="dbpedia embeddings comparing ivfflat and hnsw queries-per-second (light)"
+    className="dark:hidden"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/dbpedia-ivfflat-vs-hnsw--light.png"
+  />
+  <img
+    alt="dbpedia embeddings comparing ivfflat and hnsw queries-per-second (dark)"
+    className="hidden dark:block"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/dbpedia-ivfflat-vs-hnsw--dark.png"
+  />
+</div>
+
+When we maintain fixed HNSW build parameters, we can adjust the select query parameter `ef_search` to balance query speed and precision. To achieve precision@10 of 0.98, we increased it from the default 40 to 100. For precision@10 of 0.99, we further raised it to 250. Remarkably, HNSW demonstrated over six times better performance while maintaining the same level of precision. With higher precision@10 of 0.99 HNSW even outperforms [qdrant](https://nirantk.com/writing/pgvector-vs-qdrant/) on equivalent compute resources.
+
+<div>
+  <img
+    alt="dbpedia embeddings comparing qdrant and hnsw queries-per-second (light)"
+    className="dark:hidden"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/dbpedia-qdrant-vs-hnsw--light.png"
+  />
+  <img
+    alt="dbpedia embeddings comparing qdrant and hnsw queries-per-second (dark)"
+    className="hidden dark:block"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/dbpedia-qdrant-vs-hnsw--dark.png"
+  />
+</div>
+
+### Scaling the database
+
+Performance scales predictably with the size of the database for the HNSW index, just as [it does for the IVFFlat index](https://supabase.com/blog/pgvector-performance#scaling-the-database). The difference in performance between IVFFlat and HNSW remains consistent after a compute upgrade.
+
+<div>
+  <img
+    alt="dbpedia embeddings comparing ivfflat and hnsw queries-per-second using the 4XL compute addon (light)"
+    className="dark:hidden"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/dbpedia-ivfflat-vs-hnsw-4xl--light.png"
+  />
+  <img
+    alt="dbpedia embeddings comparing ivfflat and hnsw queries-per-second using the 4XL compute addon (dark)"
+    className="hidden dark:block"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/dbpedia-ivfflat-vs-hnsw-4xl--dark.png"
+  />
+</div>
+
+Switching to a 4XL compute add-on (with a 16-core CPU and 64GB of RAM) resulted in a 69% increase in QPS for a precision@10 of 0.99 compared to 2XL instance.
+
+<div>
+  <img
+    alt="dbpedia embeddings comparing hnsw queries-per-second using the 2XL vs 4XL compute addon (light)"
+    className="dark:hidden"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/dbpedia-hnsw-2xl-vs-4xl--light.png"
+  />
+  <img
+    alt="dbpedia embeddings comparing hnsw queries-per-second using the 2XL vs 4XL compute addon (dark)"
+    className="hidden dark:block"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/dbpedia-hnsw-2xl-vs-4xl--dark.png"
+  />
+</div>
+
+Furthermore, having 64GB of RAM is better suited for this dataset because Postgres uses approximately 30-35GB of RAM to achieve optimal performance. With a 4XL setup, you not only have the capacity to store 1,000,000 vectors but also the flexibility to accommodate additional data or increase the number of vectors as needed.
+
+### HNSW build parameters
+
+We conducted a small experiment to assess the impact of HNSW index parameters on precision and QPS. In this experiment, we utilized the same 4XL instance as in the previous test but modified the building parameters:
+
+- `m`: 32
+- `ef_construction`: 80
+
+This adjustment enabled us to achieve the same level of precision@10 of 0.99 with an `ef_search` = 100 instead of the previous 250, resulting in a 35% increase in QPS.
+
+<div>
+  <img
+    alt="dbpedia embeddings comparing hnsw queries-per-second using different build parameters (light)"
+    className="dark:hidden"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/dbpedia-hnsw-build-parameters--light.png"
+  />
+  <img
+    alt="dbpedia embeddings comparing hnsw queries-per-second using different build parameters (dark)"
+    className="hidden dark:block"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/dbpedia-hnsw-build-parameters--dark.png"
+  />
+</div>
+
+## Other HNSW features
+
+In addition to the above query performance improvements, HNSW offers another advantage: you don't need to fully fill your table before building the index.
+
+With IVFFlat indexes, the clusters (lists) are constructed based on the distribution of existing data in the table. This means that IVF indexes built on an empty table would produce completely suboptimal centers. This is why pgvector recommends building IVF indexes only once sufficient data exists in the table and rebuilding them any time the distribution of data changes significantly.
+
+HNSW indexes use graphs which inherently are not affected by this limitation, so you are safe to create your HNSW index immediately after the table is created. As new data is added to the table, the index will be filled automatically and the index structure will remain optimal.
+
+## Improvements to IVF indexes
+
+IVFFlat indexes also saw some improvements in v0.5.0. Index build times are now significantly faster for large datasets thanks to 2 new improvements:
+
+- Parallelization during the assignment build step (assigning records to lists)
+- Switching from [double to float](https://github.com/pgvector/pgvector/pull/180) precision for select distance calculations which unlocks the [fused multiply-add](https://en.wikipedia.org/wiki/Multiply%E2%80%93accumulate_operation#Fused_multiply%E2%80%93add) instruction on CPUs
+
+Below we compare the index build times between v0.4.0 and v0.5.0 for the inner product distance measure over 1,000,000 vectors on a Supabase project with 4XL compute add-on (with a 16-core CPU and 64GB of RAM).
+
+- `lists = 1000`
+
+<div>
+  <img
+    alt="dbpedia embeddings comparing ivfflat index build time between v0.4.0 and v0.5.0 using 1000 lists (light)"
+    className="dark:hidden"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/dbpedia-ivfflat-0-4-0-vs-0-5-0-lists-1000--light.png"
+  />
+  <img
+    alt="dbpedia embeddings comparing ivfflat index build time between v0.4.0 and v0.5.0 using 1000 lists (dark)"
+    className="hidden dark:block"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/dbpedia-ivfflat-0-4-0-vs-0-5-0-lists-1000--dark.png"
+  />
+</div>
+
+- `lists = 5000`
+
+<div>
+  <img
+    alt="dbpedia embeddings comparing ivfflat index build time between v0.4.0 and v0.5.0 using 5000 lists (light)"
+    className="dark:hidden"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/dbpedia-ivfflat-0-4-0-vs-0-5-0-lists-5000--light.png"
+  />
+  <img
+    alt="dbpedia embeddings comparing ivfflat index build time between v0.4.0 and v0.5.0 using 5000 lists (dark)"
+    className="hidden dark:block"
+    src="/images/blog/2023-09-05-pgvector-v0-5-0-hnsw/dbpedia-ivfflat-0-4-0-vs-0-5-0-lists-5000--dark.png"
+  />
+</div>
+
+The index build time has decreased by over 50%, and this ratio remains consistent when you adjust the index build parameters ([specifically, the `lists` value](https://supabase.com/blog/pgvector-performance#other-performance-factors)).
+
+## When should you use HNSW vs IVF?
+
+In most cases today, HNSW offers a more performant and robust index over IVFFlat. It's worth noting though that HNSW indexes will almost always be slower to build and use more memory than IVFFlat, so if your system is memory-constrained and you don't foresee the need to rebuild your index often, you may find IVFFlat to be more suitable. It's also worth noting that product quantization (compressing index entries for vectors) is expected for IVF in the next versions of pgvector which should significantly improve performance and lower resource requirements. Regardless of the type of index you use, we still recommend [reducing the number of dimensions](https://supabase.com/blog/fewer-dimensions-are-better-pgvector) in your embeddings when possible.
+
+## Start using HNSW
+
+All [new Supabase databases](https://database.new/) will automatically ship with pgvector v0.5.0 which includes the new HNSW indexes. If you have an existing database, please reach out to [support](https://supabase.com/dashboard/support/new) and we'll be more than happy to assist you in an upgrade.
+
+## More pgvector and AI resources
+
+- [How to build ChatGPT Plugin from scratch with Supabase Edge Runtime](https://supabase.com/blog/building-chatgpt-plugins-template)
+- [Docs pgvector: Embeddings and vector similarity](https://supabase.com/docs/guides/database/extensions/pgvector)
+- [Choosing Compute Add-on for AI workloads](https://supabase.com/docs/guides/ai/choosing-compute-addon)
+- [pgvector: Fewer dimensions are better](https://supabase.com/blog/fewer-dimensions-are-better-pgvector)

apps/www/_customers/chatbase.mdx

@@ -0,0 +1,78 @@
+---
+name: Chatbase
+title: Bootstrapped founder builds an AI app with Supabase and scales to $1M in 5 months
+# Use meta_title to add a custom meta title. Otherwise it defaults to '{name} | Supabase Customer Stories':
+meta_title: Bootstrapped founder builds an AI app with Supabase and scales to $1M in 5 months
+description: How Yasser leveraged Supabase to build Chatbase and became one of the most successful single-founder AI products.
+# Use meta_description to add a custom meta description. Otherwise it defaults to {description}:
+meta_description: How Yasser leveraged Supabase to build Chatbase and became one of the most successful single-founder AI products.
+author: copple
+author_title: Supabase
+author_url: https://github.com/kiwicopple
+author_image_url: https://avatars2.githubusercontent.com/u/10214025?s=400&u=c6775be2ae667e2acae3ccd347fed62bb3f5b3e7&v=4
+authorURL: https://github.com/kiwicopple
+logo: /images/customers/logos/chatbase.png
+logo_inverse: /images/customers/logos/light/chatbase.png
+og_image: /images/customers/og/chatbase.jpg
+tags:
+  - supabase
+date: '2023-10-06'
+company_url: https://www.chatbase.co/
+stats: []
+misc:
+  [
+    { label: 'Use case', text: 'AI chatbot builder' },
+    {
+      label: 'Solutions',
+      text: 'Supabase Database, Supabase Auth, Supabase Storage, Supabase Realtime',
+    },
+  ]
+about: 'Chatbase is an AI chatbot builder. It trains ChatGPT on your data and lets you add a chat widget to your website. Just upload a document or add a link to your website and get a chatbot that can answer any question about their content.'
+---
+
+After getting a new grad offer from his “dream company” rescinded with tech layoffs, Yasser Elsaid decided to embark on a different path and bootstrap his own venture.
+
+Two months before the release of the ChatGPT API, Yasser explored the GPT3 API and saw its immense potential. Leveraging Supabase as the backend, he built Chatbase in two weeks and launched it in February.
+
+Just five months after launching, Chatbase reach $1,000,000 annualized revenue, making it one of the most successful single-founder AI products in the industry.
+
+## The Challenge
+
+Fascinated by the possibilities of GPT3, Yasser envisioned Chatbase as an AI-driven chatbot capable of handling complex customer queries in real time. However, building a cost-effective and reliable solution to ingest and store data and manage large-scale customer interactions by himself wasn’t trivial.
+
+Yasser faced another challenge: he needed to rapidly transform his idea into an actual product to maintain a competitive edge, but he didn’t have a team or funding.
+
+## Choosing Supabase
+
+Using Supabase for the first time, Yasser was immediately impressed by the developer experience and ease of use. He effortlessly implemented secure user authentication, relieving the pain of building such a system from scratch.
+
+But the most crucial aspect for us is Supabase's all-in-one solution. Chatbase relies on Supabase for [database](https://supabase.com/database), [authentication](https://supabase.com/auth), [storage](https://supabase.com/storage), and [real-time](https://supabase.com/realtime) functionality. Yasser believes he saved between 100 to 150 hours by using only 1 solution to build almost the entire backend, instead of spending that researching, learning, and implementing individual solutions for each component.
+
+With Supabase, everything seamlessly came together and he was able to go from idea to MVP in two weeks and launch soon after that.
+
+<Quote img="yasser-elsaid-chatbase.jpeg" caption="Yasser Elsaid, Founder of Chatbase.">
+  <p>
+    Supabase is great because it has everything. I don’t need a different solution for
+    authentication, a different solution for database, or a different solution for storage.
+  </p>
+</Quote>
+
+## What he built
+
+Chatbase is a powerful AI chatbot builder that enables users to train ChatGPT on their own data, revolutionizing the way businesses and individuals interact with customers. With seamless integration to over 5000 apps through Zapier, including popular platforms like Excel, Notion, Whatsapp, Discord, and Slack, Chatbase empowers everyone to deploy custom AI for their specific workflows.
+
+By simply uploading a PDF containing the desired information, Chatbase creates intelligent chatbots that automate customer support, handle frequently asked questions, and streamline communication channels, ultimately enhancing overall efficiency and user experience.
+
+Chatbase became a sensation, crossing 2,000,000 visitors in a matter of months and now has 2,200 paying customers.
+
+## The Results
+
+Leveraging Supabase as the backend gave Yasser the speed and efficiency to turn into a significant advantage over the competition, enabling him to gain a strong advantage in the market and establish Chatbase as one of the pioneering AI chatbot builders.
+
+As Chatbase gained traction and acquired users, Supabase's robust infrastructure seamlessly accommodated the surging user base, ensuring smooth scalability and uninterrupted growth.
+
+Within just five months, Chatbase grew to an astonishing $1,000,000 in annualize revenue.
+
+Chatbase is still bootstrapped, but now Yasser has a team of 5 who help with customer support, marketing, and development. He is now exploring Supabase's capabilities for achieving GDPR compliance and implementing Vault for end-to-end encryption.
+
+> To learn more about how Supabase can help you build and scale AI apps with ease, [reach out to us](https://forms.supabase.com/enterprise).

apps/www/_customers/quivr.mdx

@@ -11,9 +11,9 @@ author_title: Supabase
 author_url: https://github.com/kiwicopple
 author_image_url: https://avatars2.githubusercontent.com/u/10214025?s=400&u=c6775be2ae667e2acae3ccd347fed62bb3f5b3e7&v=4
 authorURL: https://github.com/kiwicopple
-logo: /images/customers/logos/xendit.png
-logo_inverse: /images/customers/logos/light/xendit.png
-og_image: /images/customers/og/xendit.jpg
+logo: /images/customers/logos/quivr.png
+logo_inverse: /images/customers/logos/light/quivr.png
+og_image: /images/customers/og/quivr.jpg
 tags:
   - supabase
 date: '2023-10-05'
@@ -27,8 +27,6 @@ misc:
 about: "Quivr is an open source 'second brain'. It's like a private ChatGPT, personalized with your own data: you upload your documents and you can then search and ask questions using generative AI."
 ---
 
-# Quivr launch 5,000 Vector databases on Supabase
-
 ## The challenge: Building a second brain
 
 In May of 2023, [Stan Girard's](https://twitter.com/_StanGirard) started building small prototypes that allowed him to “chat with documents”. After 2 weeks of research, he settled on an idea - build a “second brain” where a user could dump all their digital knowledge (audio, URLs, text, and code) into a vector store and query it with GPT4.
@@ -39,7 +37,7 @@ He built the first version in a single afternoon, pushed it to GitHub, and then
 
 A critical piece of the tech stack was the vector store. Stan needed a place to store millions of embeddings. After comparing between Supabase, Pinecone, and Chroma, he settled on [Supabase Vector](https://supabase.com/vector), our open source vector offering for developing AI applications. The decision was driven largely by his familiarity with Postgres, and the tight integration with Vercel.
 
-<Quote img="caryn-marooney.jpeg" caption="Stan Girard, Founder of Quivr.">
+<Quote img="stan-girard-avatar.jpeg" caption="Stan Girard, Founder of Quivr.">
   <p>
     Supabase Vector powered by pgvector allowed us to create a simple and efficient product. We are
     storing over 1.6 million embeddings and the performance and results are great. Open source
@@ -51,7 +49,7 @@ A critical piece of the tech stack was the vector store. Stan needed a place to
 
 It didn't take long for the Quivr community to grow. After the viral launch, the [Quivr repo](https://github.com/StanGirard/quivr) stayed at number 1 on [GitHub Trending](https://github.com/trending) for 4 days. Today, it has over 22,000 GitHub stars and 67 contributors. Supabase has been a key part of the open source stack since the beginning.
 
-<Quote img="caryn-marooney.jpeg" caption="Stan Girard, Founder of Quivr.">
+<Quote img="stan-girard-avatar.jpeg" caption="Stan Girard, Founder of Quivr.">
   <p>
     Because Supabase is open source, the possibility of running it locally made it a better choice
     compared with other products like Auth0. Since Auth is integrated with the Vector database it

apps/www/pages/customers/[slug].tsx

@@ -66,7 +66,8 @@ function CaseStudyPage(props: any) {
   const content = props.blog.content
 
   const meta = {
-    title: props.blog.meta_title ?? `${props.blog.name} | Supabase Customer Stories`,
+    title:
+      props.blog.meta_title ?? props.blog.title ?? `${props.blog.name} | Supabase Customer Stories`,
     description: props.blog.meta_description ?? props.blog.description,
     image:
       `${SITE_ORIGIN}${props.blog.og_image}` ??

apps/www/public/customers-rss.xml

@@ -5,9 +5,16 @@
       <link>https://supabase.com</link>
       <description>Latest news from Supabase</description>
       <language>en</language>
-      <lastBuildDate>Wed, 04 Oct 2023 22:00:00 GMT</lastBuildDate>
+      <lastBuildDate>Thu, 05 Oct 2023 22:00:00 GMT</lastBuildDate>
       <atom:link href="https://supabase.com/rss.xml" rel="self" type="application/rss+xml"/>
       <item>
+  <guid>https://supabase.com/customers/chatbase</guid>
+  <title>Bootstrapped founder builds an AI app with Supabase and scales to $1M in 5 months</title>
+  <link>https://supabase.com/customers/chatbase</link>
+  <description>How Yasser leveraged Supabase to build Chatbase and became one of the most successful single-founder AI products.</description>
+  <pubDate>Thu, 05 Oct 2023 22:00:00 GMT</pubDate>
+</item>
+<item>
   <guid>https://supabase.com/customers/quivr</guid>
   <title>Quivr launch 5,000 Vector databases on Supabase.</title>
   <link>https://supabase.com/customers/quivr</link>

apps/www/public/planetpg-angelico_de_los_reyes-rss.xml

@@ -5,35 +5,35 @@
       <link>https://supabase.com/blog</link>
       <description>Latest Postgres news from Angelico de los Reyes at Supabase</description>
       <language>en</language>
-      <lastBuildDate>Thu, 15 Dec 2022 22:00:00 GMT</lastBuildDate>
+      <lastBuildDate>Thu, 15 Dec 2022 23:00:00 GMT</lastBuildDate>
       <atom:link href="https://supabase.com/planetpg-angelico_de_los_reyes-rss.xml" rel="self" type="application/rss+xml"/>
       <item>
   <guid>https://supabase.com/blog/postgres-point-in-time-recovery</guid>
   <title>Point in Time Recovery is now available for Pro projects</title>
   <link>https://supabase.com/blog/postgres-point-in-time-recovery</link>
   <description>We&apos;re making PITR available for more projects, with a new Dashboard UI that makes it simple to use.</description>
-  <pubDate>Thu, 15 Dec 2022 22:00:00 GMT</pubDate>
+  <pubDate>Thu, 15 Dec 2022 23:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/continuous-postgresql-backup-walg</guid>
   <title>Continuous PostgreSQL Backups using WAL-G</title>
   <link>https://supabase.com/blog/continuous-postgresql-backup-walg</link>
   <description>Have you ever wanted to restore your database&apos;s state to a particular moment in time? This post explains how, using WAL-G.</description>
-  <pubDate>Sat, 01 Aug 2020 21:00:00 GMT</pubDate>
+  <pubDate>Sat, 01 Aug 2020 22:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/postgresql-templates</guid>
   <title>What are PostgreSQL Templates?</title>
   <link>https://supabase.com/blog/postgresql-templates</link>
   <description>What are PostgreSQL templates and what are they used for?</description>
-  <pubDate>Wed, 08 Jul 2020 21:00:00 GMT</pubDate>
+  <pubDate>Wed, 08 Jul 2020 22:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/postgresql-physical-logical-backups</guid>
   <title>Physical vs Logical Backups in PostgreSQL</title>
   <link>https://supabase.com/blog/postgresql-physical-logical-backups</link>
   <description>What are physical and logical backups in Postgres?</description>
-  <pubDate>Mon, 06 Jul 2020 21:00:00 GMT</pubDate>
+  <pubDate>Mon, 06 Jul 2020 22:00:00 GMT</pubDate>
 </item>
 
     </channel>

apps/www/public/planetpg-ant_wilson-rss.xml

@@ -5,14 +5,14 @@
       <link>https://supabase.com/blog</link>
       <description>Latest Postgres news from Ant Wilson at Supabase</description>
       <language>en</language>
-      <lastBuildDate>Fri, 26 Feb 2021 22:00:00 GMT</lastBuildDate>
+      <lastBuildDate>Fri, 26 Feb 2021 23:00:00 GMT</lastBuildDate>
       <atom:link href="https://supabase.com/planetpg-ant_wilson-rss.xml" rel="self" type="application/rss+xml"/>
       <item>
   <guid>https://supabase.com/blog/cracking-postgres-interview</guid>
   <title>Cracking PostgreSQL Interview Questions</title>
   <link>https://supabase.com/blog/cracking-postgres-interview</link>
   <description>Understand the top PostgreSQL Interview Questions</description>
-  <pubDate>Fri, 26 Feb 2021 22:00:00 GMT</pubDate>
+  <pubDate>Fri, 26 Feb 2021 23:00:00 GMT</pubDate>
 </item>
 
     </channel>

apps/www/public/planetpg-bo_lu-rss.xml

@@ -5,14 +5,14 @@
       <link>https://supabase.com/blog</link>
       <description>Latest Postgres news from Bo Lu at Supabase</description>
       <language>en</language>
-      <lastBuildDate>Wed, 14 Dec 2022 22:00:00 GMT</lastBuildDate>
+      <lastBuildDate>Wed, 14 Dec 2022 23:00:00 GMT</lastBuildDate>
       <atom:link href="https://supabase.com/planetpg-bo_lu-rss.xml" rel="self" type="application/rss+xml"/>
       <item>
   <guid>https://supabase.com/blog/postgres-foreign-data-wrappers-rust</guid>
   <title>Supabase Wrappers, a Postgres FDW framework written in Rust</title>
   <link>https://supabase.com/blog/postgres-foreign-data-wrappers-rust</link>
   <description>A framework for building Postgres Foreign Data Wrappers which connects to Stripe, Firebase, Clickhouse, and more.</description>
-  <pubDate>Wed, 14 Dec 2022 22:00:00 GMT</pubDate>
+  <pubDate>Wed, 14 Dec 2022 23:00:00 GMT</pubDate>
 </item>
 
     </channel>

apps/www/public/planetpg-burggraf-rss.xml

@@ -5,21 +5,21 @@
       <link>https://supabase.com/blog</link>
       <description>Latest Postgres news from Mark Burggraf at Supabase</description>
       <language>en</language>
-      <lastBuildDate>Wed, 23 Nov 2022 22:00:00 GMT</lastBuildDate>
+      <lastBuildDate>Wed, 23 Nov 2022 23:00:00 GMT</lastBuildDate>
       <atom:link href="https://supabase.com/planetpg-burggraf-rss.xml" rel="self" type="application/rss+xml"/>
       <item>
   <guid>https://supabase.com/blog/sql-or-nosql-both-with-postgresql</guid>
   <title>SQL or NoSQL? Why not use both (with PostgreSQL)?</title>
   <link>https://supabase.com/blog/sql-or-nosql-both-with-postgresql</link>
   <description>How to turn Postgres into an easy-to-use NoSQL database that retains all the power of SQL</description>
-  <pubDate>Wed, 23 Nov 2022 22:00:00 GMT</pubDate>
+  <pubDate>Wed, 23 Nov 2022 23:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/postgres-wasm</guid>
   <title>Postgres WASM by Snaplet and Supabase</title>
   <link>https://supabase.com/blog/postgres-wasm</link>
   <description>We&apos;re open sourcing postgres-wasm, a PostgresQL server that runs inside a browser, with our friends at Snaplet.</description>
-  <pubDate>Sun, 02 Oct 2022 21:00:00 GMT</pubDate>
+  <pubDate>Sun, 02 Oct 2022 22:00:00 GMT</pubDate>
 </item>
 
     </channel>

apps/www/public/planetpg-egor_romanov-rss.xml

@@ -5,21 +5,28 @@
       <link>https://supabase.com/blog</link>
       <description>Latest Postgres news from Egor Romanov at Supabase</description>
       <language>en</language>
-      <lastBuildDate>Wed, 02 Aug 2023 21:00:00 GMT</lastBuildDate>
+      <lastBuildDate>Mon, 04 Sep 2023 22:00:00 GMT</lastBuildDate>
       <atom:link href="https://supabase.com/planetpg-egor_romanov-rss.xml" rel="self" type="application/rss+xml"/>
       <item>
+  <guid>https://supabase.com/blog/pgvector-v0-5-0-hnsw</guid>
+  <title>pgvector v0.5.0: Faster semantic search with HNSW indexes</title>
+  <link>https://supabase.com/blog/pgvector-v0-5-0-hnsw</link>
+  <description>Increase performance in pgvector using HNSW indexes</description>
+  <pubDate>Mon, 04 Sep 2023 22:00:00 GMT</pubDate>
+</item>
+<item>
   <guid>https://supabase.com/blog/fewer-dimensions-are-better-pgvector</guid>
   <title>pgvector: Fewer dimensions are better</title>
   <link>https://supabase.com/blog/fewer-dimensions-are-better-pgvector</link>
   <description>Increase performance in pgvector by using embedding vectors with fewer dimensions</description>
-  <pubDate>Wed, 02 Aug 2023 21:00:00 GMT</pubDate>
+  <pubDate>Wed, 02 Aug 2023 22:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/pgvector-performance</guid>
   <title>pgvector 0.4.0 performance</title>
   <link>https://supabase.com/blog/pgvector-performance</link>
   <description>There&apos;s been lot of talk about pgvector performance lately, so we took some datasets and pushed pgvector to the limits to find out its strengths and limitations.</description>
-  <pubDate>Wed, 12 Jul 2023 21:00:00 GMT</pubDate>
+  <pubDate>Wed, 12 Jul 2023 22:00:00 GMT</pubDate>
 </item>
 
     </channel>

apps/www/public/planetpg-gregnr-rss.xml

@@ -5,21 +5,28 @@
       <link>https://supabase.com/blog</link>
       <description>Latest Postgres news from Greg Richardson at Supabase</description>
       <language>en</language>
-      <lastBuildDate>Wed, 02 Aug 2023 21:00:00 GMT</lastBuildDate>
+      <lastBuildDate>Mon, 04 Sep 2023 22:00:00 GMT</lastBuildDate>
       <atom:link href="https://supabase.com/planetpg-gregnr-rss.xml" rel="self" type="application/rss+xml"/>
       <item>
+  <guid>https://supabase.com/blog/pgvector-v0-5-0-hnsw</guid>
+  <title>pgvector v0.5.0: Faster semantic search with HNSW indexes</title>
+  <link>https://supabase.com/blog/pgvector-v0-5-0-hnsw</link>
+  <description>Increase performance in pgvector using HNSW indexes</description>
+  <pubDate>Mon, 04 Sep 2023 22:00:00 GMT</pubDate>
+</item>
+<item>
   <guid>https://supabase.com/blog/fewer-dimensions-are-better-pgvector</guid>
   <title>pgvector: Fewer dimensions are better</title>
   <link>https://supabase.com/blog/fewer-dimensions-are-better-pgvector</link>
   <description>Increase performance in pgvector by using embedding vectors with fewer dimensions</description>
-  <pubDate>Wed, 02 Aug 2023 21:00:00 GMT</pubDate>
+  <pubDate>Wed, 02 Aug 2023 22:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/openai-embeddings-postgres-vector</guid>
   <title>Storing OpenAI embeddings in Postgres with pgvector</title>
   <link>https://supabase.com/blog/openai-embeddings-postgres-vector</link>
   <description>An example of how to build an AI-powered search engine using OpenAI&apos;s embeddings and PostgreSQL.</description>
-  <pubDate>Sun, 05 Feb 2023 22:00:00 GMT</pubDate>
+  <pubDate>Sun, 05 Feb 2023 23:00:00 GMT</pubDate>
 </item>
 
     </channel>

apps/www/public/planetpg-michel-rss.xml

@@ -5,21 +5,21 @@
       <link>https://supabase.com/blog</link>
       <description>Latest Postgres news from Michel Pelletier at Supabase</description>
       <language>en</language>
-      <lastBuildDate>Thu, 15 Dec 2022 22:00:00 GMT</lastBuildDate>
+      <lastBuildDate>Thu, 15 Dec 2022 23:00:00 GMT</lastBuildDate>
       <atom:link href="https://supabase.com/planetpg-michel-rss.xml" rel="self" type="application/rss+xml"/>
       <item>
   <guid>https://supabase.com/blog/vault-now-in-beta</guid>
   <title>Supabase Vault is now in Beta</title>
   <link>https://supabase.com/blog/vault-now-in-beta</link>
   <description>A Postgres extension to store encrypted secrets and encrypt data.</description>
-  <pubDate>Thu, 15 Dec 2022 22:00:00 GMT</pubDate>
+  <pubDate>Thu, 15 Dec 2022 23:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/transparent-column-encryption-with-postgres</guid>
   <title>Transparent Column Encryption with Postgres</title>
   <link>https://supabase.com/blog/transparent-column-encryption-with-postgres</link>
   <description>Using pgsodium&apos;s Transparent Column Encryption to encrypt data and provide your users with row-level encryption.</description>
-  <pubDate>Wed, 30 Nov 2022 22:00:00 GMT</pubDate>
+  <pubDate>Wed, 30 Nov 2022 23:00:00 GMT</pubDate>
 </item>
 
     </channel>

apps/www/public/planetpg-oli_rice-rss.xml

@@ -5,56 +5,56 @@
       <link>https://supabase.com/blog</link>
       <description>Latest Postgres news from Oliver Rice at Supabase</description>
       <language>en</language>
-      <lastBuildDate>Wed, 02 Aug 2023 21:00:00 GMT</lastBuildDate>
+      <lastBuildDate>Wed, 02 Aug 2023 22:00:00 GMT</lastBuildDate>
       <atom:link href="https://supabase.com/planetpg-oli_rice-rss.xml" rel="self" type="application/rss+xml"/>
       <item>
   <guid>https://supabase.com/blog/fewer-dimensions-are-better-pgvector</guid>
   <title>pgvector: Fewer dimensions are better</title>
   <link>https://supabase.com/blog/fewer-dimensions-are-better-pgvector</link>
   <description>Increase performance in pgvector by using embedding vectors with fewer dimensions</description>
-  <pubDate>Wed, 02 Aug 2023 21:00:00 GMT</pubDate>
+  <pubDate>Wed, 02 Aug 2023 22:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/whats-new-in-pg-graphql-v1-2</guid>
   <title>What&apos;s New in pg_graphql v1.2</title>
   <link>https://supabase.com/blog/whats-new-in-pg-graphql-v1-2</link>
   <description>New Features in the v1.2 release of pg_graphql</description>
-  <pubDate>Thu, 20 Apr 2023 21:00:00 GMT</pubDate>
+  <pubDate>Thu, 20 Apr 2023 22:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/type-constraints-in-65-lines-of-sql</guid>
   <title>Type Constraints in 65 lines of SQL</title>
   <link>https://supabase.com/blog/type-constraints-in-65-lines-of-sql</link>
   <description>Creating validated data types in Postgres</description>
-  <pubDate>Thu, 16 Feb 2023 22:00:00 GMT</pubDate>
+  <pubDate>Thu, 16 Feb 2023 23:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/pg-graphql-v1</guid>
   <title>pg_graphql v1.0</title>
   <link>https://supabase.com/blog/pg-graphql-v1</link>
   <description>Announcing the v1.0 release of pg_graphql</description>
-  <pubDate>Thu, 15 Dec 2022 22:00:00 GMT</pubDate>
+  <pubDate>Thu, 15 Dec 2022 23:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/postgres-foreign-data-wrappers-rust</guid>
   <title>Supabase Wrappers, a Postgres FDW framework written in Rust</title>
   <link>https://supabase.com/blog/postgres-foreign-data-wrappers-rust</link>
   <description>A framework for building Postgres Foreign Data Wrappers which connects to Stripe, Firebase, Clickhouse, and more.</description>
-  <pubDate>Wed, 14 Dec 2022 22:00:00 GMT</pubDate>
+  <pubDate>Wed, 14 Dec 2022 23:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/pg-jsonschema-a-postgres-extension-for-json-validation</guid>
   <title>pg_jsonschema: JSON Schema support for Postgres</title>
   <link>https://supabase.com/blog/pg-jsonschema-a-postgres-extension-for-json-validation</link>
   <description>Today we&apos;re releasing pg_jsonschema, a Postgres extension for JSON validation.</description>
-  <pubDate>Thu, 18 Aug 2022 21:00:00 GMT</pubDate>
+  <pubDate>Thu, 18 Aug 2022 22:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/postgres-audit</guid>
   <title>Postgres Auditing in 150 lines of SQL</title>
   <link>https://supabase.com/blog/postgres-audit</link>
   <description>PostgreSQL has a robust set of features which we can leverage to create a generic auditing solution in 150 lines of SQL.</description>
-  <pubDate>Mon, 07 Mar 2022 22:00:00 GMT</pubDate>
+  <pubDate>Mon, 07 Mar 2022 23:00:00 GMT</pubDate>
 </item>
 
     </channel>

apps/www/public/planetpg-paul_copplestone-rss.xml

@@ -5,49 +5,49 @@
       <link>https://supabase.com/blog</link>
       <description>Latest Postgres news from Paul Copplestone at Supabase</description>
       <language>en</language>
-      <lastBuildDate>Sun, 30 Apr 2023 21:00:00 GMT</lastBuildDate>
+      <lastBuildDate>Sun, 30 Apr 2023 22:00:00 GMT</lastBuildDate>
       <atom:link href="https://supabase.com/planetpg-paul_copplestone-rss.xml" rel="self" type="application/rss+xml"/>
       <item>
   <guid>https://supabase.com/blog/postgres-pluggable-strorage</guid>
   <title>Next steps for Postgres pluggable storage</title>
   <link>https://supabase.com/blog/postgres-pluggable-strorage</link>
   <description>Exploring history of Postgres pluggable storage and the possibility of landing it in the Postgres core.</description>
-  <pubDate>Sun, 30 Apr 2023 21:00:00 GMT</pubDate>
+  <pubDate>Sun, 30 Apr 2023 22:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/postgres-foreign-data-wrappers-rust</guid>
   <title>Supabase Wrappers, a Postgres FDW framework written in Rust</title>
   <link>https://supabase.com/blog/postgres-foreign-data-wrappers-rust</link>
   <description>A framework for building Postgres Foreign Data Wrappers which connects to Stripe, Firebase, Clickhouse, and more.</description>
-  <pubDate>Wed, 14 Dec 2022 22:00:00 GMT</pubDate>
+  <pubDate>Wed, 14 Dec 2022 23:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/postgres-crdt</guid>
   <title>pg_crdt - an experimental CRDT extension for Postgres</title>
   <link>https://supabase.com/blog/postgres-crdt</link>
   <description>Embedding Yjs and Automerge into Postgres for collaborative applications.</description>
-  <pubDate>Fri, 09 Dec 2022 22:00:00 GMT</pubDate>
+  <pubDate>Fri, 09 Dec 2022 23:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/partial-postgresql-data-dumps-with-rls</guid>
   <title>Partial data dumps using Postgres Row Level Security</title>
   <link>https://supabase.com/blog/partial-postgresql-data-dumps-with-rls</link>
   <description>Using RLS to create seed files for local PostgreSQL testing.</description>
-  <pubDate>Mon, 27 Jun 2022 21:00:00 GMT</pubDate>
+  <pubDate>Mon, 27 Jun 2022 22:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/postgres-as-a-cron-server</guid>
   <title>Postgres as a CRON Server</title>
   <link>https://supabase.com/blog/postgres-as-a-cron-server</link>
   <description>Running repetitive tasks with your Postgres database.</description>
-  <pubDate>Thu, 04 Mar 2021 22:00:00 GMT</pubDate>
+  <pubDate>Thu, 04 Mar 2021 23:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/postgresql-views</guid>
   <title>Postgres Views</title>
   <link>https://supabase.com/blog/postgresql-views</link>
   <description>Creating and using a view in PostgreSQL.</description>
-  <pubDate>Tue, 17 Nov 2020 22:00:00 GMT</pubDate>
+  <pubDate>Tue, 17 Nov 2020 23:00:00 GMT</pubDate>
 </item>
 
     </channel>

apps/www/public/planetpg-pavel-rss.xml

@@ -5,28 +5,28 @@
       <link>https://supabase.com/blog</link>
       <description>Latest Postgres news from Pavel Borisov at Supabase</description>
       <language>en</language>
-      <lastBuildDate>Wed, 12 Jul 2023 21:00:00 GMT</lastBuildDate>
+      <lastBuildDate>Wed, 12 Jul 2023 22:00:00 GMT</lastBuildDate>
       <atom:link href="https://supabase.com/planetpg-pavel-rss.xml" rel="self" type="application/rss+xml"/>
       <item>
   <guid>https://supabase.com/blog/pgvector-performance</guid>
   <title>pgvector 0.4.0 performance</title>
   <link>https://supabase.com/blog/pgvector-performance</link>
   <description>There&apos;s been lot of talk about pgvector performance lately, so we took some datasets and pushed pgvector to the limits to find out its strengths and limitations.</description>
-  <pubDate>Wed, 12 Jul 2023 21:00:00 GMT</pubDate>
+  <pubDate>Wed, 12 Jul 2023 22:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/new-in-postgres-15</guid>
   <title>What&apos;s new in Postgres 15?</title>
   <link>https://supabase.com/blog/new-in-postgres-15</link>
   <description>Describes the release of Postgres 15, new features and reasons to use it</description>
-  <pubDate>Thu, 15 Dec 2022 22:00:00 GMT</pubDate>
+  <pubDate>Thu, 15 Dec 2022 23:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/postgresql-commitfest</guid>
   <title>What is PostgreSQL commitfest and how to contribute</title>
   <link>https://supabase.com/blog/postgresql-commitfest</link>
   <description>A time-tested method for contributing to the core Postgres code</description>
-  <pubDate>Wed, 26 Oct 2022 21:00:00 GMT</pubDate>
+  <pubDate>Wed, 26 Oct 2022 22:00:00 GMT</pubDate>
 </item>
 
     </channel>

apps/www/public/planetpg-steve_chavez-rss.xml

@@ -5,21 +5,21 @@
       <link>https://supabase.com/blog</link>
       <description>Latest Postgres news from Steve Chavez at Supabase</description>
       <language>en</language>
-      <lastBuildDate>Thu, 15 Dec 2022 22:00:00 GMT</lastBuildDate>
+      <lastBuildDate>Thu, 15 Dec 2022 23:00:00 GMT</lastBuildDate>
       <atom:link href="https://supabase.com/planetpg-steve_chavez-rss.xml" rel="self" type="application/rss+xml"/>
       <item>
   <guid>https://supabase.com/blog/postgrest-11-prerelease</guid>
   <title>PostgREST 11 pre-release</title>
   <link>https://supabase.com/blog/postgrest-11-prerelease</link>
   <description>Describes new features of PostgREST 11 pre-release</description>
-  <pubDate>Thu, 15 Dec 2022 22:00:00 GMT</pubDate>
+  <pubDate>Thu, 15 Dec 2022 23:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/roles-postgres-hooks</guid>
   <title>Protecting reserved roles with PostgreSQL Hooks</title>
   <link>https://supabase.com/blog/roles-postgres-hooks</link>
   <description>Using Postgres Hooks to protect functionality in your Postgres database.</description>
-  <pubDate>Thu, 01 Jul 2021 21:00:00 GMT</pubDate>
+  <pubDate>Thu, 01 Jul 2021 22:00:00 GMT</pubDate>
 </item>
 
     </channel>

apps/www/public/planetpg-victor-rss.xml

@@ -5,28 +5,28 @@
       <link>https://supabase.com/blog</link>
       <description>Latest Postgres news from Victor at Supabase</description>
       <language>en</language>
-      <lastBuildDate>Thu, 13 Oct 2022 21:00:00 GMT</lastBuildDate>
+      <lastBuildDate>Thu, 13 Oct 2022 22:00:00 GMT</lastBuildDate>
       <atom:link href="https://supabase.com/planetpg-victor-rss.xml" rel="self" type="application/rss+xml"/>
       <item>
   <guid>https://supabase.com/blog/postgres-full-text-search-vs-the-rest</guid>
   <title>Postgres Full Text Search vs the rest</title>
   <link>https://supabase.com/blog/postgres-full-text-search-vs-the-rest</link>
   <description>Comparing one of the most popular Postgres features against alternatives</description>
-  <pubDate>Thu, 13 Oct 2022 21:00:00 GMT</pubDate>
+  <pubDate>Thu, 13 Oct 2022 22:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/choosing-a-postgres-primary-key</guid>
   <title>Choosing a Postgres Primary Key</title>
   <link>https://supabase.com/blog/choosing-a-postgres-primary-key</link>
   <description>Turns out the question of which identifier to use as a Primary Key is complicated -- we&apos;re going to dive into some of the complexity and inherent trade-offs, and figure things out</description>
-  <pubDate>Wed, 07 Sep 2022 21:00:00 GMT</pubDate>
+  <pubDate>Wed, 07 Sep 2022 22:00:00 GMT</pubDate>
 </item>
 <item>
   <guid>https://supabase.com/blog/seen-by-in-postgresql</guid>
   <title>Implementing &quot;seen by&quot; functionality with Postgres</title>
   <link>https://supabase.com/blog/seen-by-in-postgresql</link>
   <description>Different approaches for tracking visitor counts with PostgreSQL.</description>
-  <pubDate>Sun, 17 Jul 2022 21:00:00 GMT</pubDate>
+  <pubDate>Sun, 17 Jul 2022 22:00:00 GMT</pubDate>
 </item>
 
     </channel>