Pinecone Vector Database- PAYG

My main use case for Pinecone  is to provide managed vector search for high-dimensional data, ideal for AI apps like semantic search and RAG, where I identify reasons for brand de-growth for big pharma brands through a sales agent and a process agent, utilizing Pinecone  for easy use of RAG and vector embeddings with high-dimensional data.

In my workflow, I have used Pinecone in agentic AI and RAG pipelines that require quick scaling without infrastructure management, aligning well with Python workflows and similar to PGVector extensions.

Pinecone stands out as a fully managed, cloud-native vector database in my brand de-growth analysis, contrasting with libraries such as FAISS or self-hosted options such as Milvus , as it prioritizes ease for production AI apps, allowing easy deployment as a fully managed serverless application with auto-scaling clusters and pay-per-usage cost, making it ideal for production RAG and AI chatbots by using guided search to retrieve outputs from Pinecone vector database.

The best feature Pinecone offers is its scalability since it auto-scales clusters, and its fully managed deployment as a serverless solution is one of the best aspects. Additionally, Pinecone is easily integratable with Python and its ease of use with Python is phenomenal.

Pinecone's scalability allows it to handle billions of vectors with auto-sharding, a capability other databases do not provide. Pinecone is stable, excelling in managed production scaling.

Pinecone has positively impacted my organization by enabling fast similarity searches using metrics such as cosine or Euclidean distance on billions of vectors with low latency around 20 to 100 milliseconds, with key capabilities including hybrid search combining semantic and keyword, real-time updates, filtering, and re-ranking.

The low latency and hybrid search from Pinecone have significantly improved my team's productivity, as when coupled with the RAG pipeline, it has enhanced solution accuracy, reducing query response time to around 10 to 15 seconds compared to 40 to 60 seconds without RAG.

From a cost perspective, I believe Pinecone is a bit expensive compared to other solutions such as FAISS and Milvus , which are free and open source, while Weaviate is more cost-effective at scale, so I would request improvement in Pinecone's pricing structure.

Furthermore, in cases of GPU-accelerated experiments requiring control over indexing strategies, I would prioritize FAISS due to its cost-free prototyping, extreme customization, and high-performance local computation, as Pinecone lacks custom GPU support compared to FAISS and fine-tuned algorithms.

I have been using Pinecone for around two years.

Pinecone is stable, excelling in managed production scaling.

Pinecone's scalability allows it to handle billions of vectors with auto-sharding, a capability other databases do not provide, and I have experienced no issues with scalability.

Customer support for Pinecone is tied to billing plans, generally starting with standard tier access through console tickets, although I feel free support is lacking.

Before adopting Pinecone, we used a Power BI dashboard to identify brand RCA, but it involved many manual and friction points in navigating boards, which did not provide clear insights, while Pinecone's multi-agent architecture has cut down the analysis time from around one week or 10 days to just one day.

I evaluated ChromaDB before implementing Pinecone.

Pinecone is deployed in my organization on a private cloud.

We utilize enterprise licensing for Pinecone.

We have seen a return on investment as we have reduced the work of 10 FTEs, allowing the Salesforce  analytics team to self-serve the data they formerly depended on other business analysts to pull, effectively consolidating the work into one person with the integration of this solution.

We utilize enterprise licensing for Pinecone, and while I cannot specify the exact costs, it should be approximately around $100 to $150 per month.

In cases of GPU-accelerated experiments requiring control over indexing strategies, I would prioritize FAISS due to its cost-free prototyping, extreme customization, and high-performance local computation, as Pinecone lacks custom GPU support compared to FAISS and fine-tuned algorithms.

I advise those looking to use Pinecone to consider it for building a serverless, scalable solution as it achieves millisecond searches across billions of vectors using optimized indexing such as HNSW, with operational simplicity as it is fully managed and serverless, able to be upgraded without infrastructure operations unlike FAISS or ChromaDB.

Overall, I feel Pinecone excels in operational simplicity and scalability, making it a flexible solution ideal for real-time RAG or agentic systems. I would rate this product a 7 out of 10.

I have used Pinecone  for the last five years, when I started my career in generative AI. It is very useful for creating POCs. I created more than 15 POCs on Pinecone  because it is very useful for use and implementation.

I have created many POCs using Pinecone. Let's suppose we have some documents in PDF format. We are getting the data from the text format, chunking and embedding it, and storing it in Pinecone. This is something we do in many applications, mostly in the POCs, because the client is not allowing it to be used on the production server. Mostly we are using the Oracle vector database on the production server. That is the issue from the client side.

I have not used Pinecone in my organization. In most cases, I use Pinecone for small projects as well as POCs. In the small projects, I use private servers for implementation and deployment.

I have not used large data. I use Pinecone for small projects, mostly single files. The file contains more than 100 pages, and it is performing well. There is nothing I'm seeing, such as drawbacks or lagging somewhere. It is working fine for us.

I use it mostly for AI applications, primarily in RAG applications. For the implementation, for the embedding, storing the embedding, and getting the data later, Pinecone works well.

Pinecone is very easy to use and it's very easy to make the connection. I use both cloud-based and local Pinecone, and the performance is much better as compared to other tools for embedding.

Faster retrieval and low latency are significant advantages. The results are mostly correct in most cases.

With Pinecone's features, we can use it both locally and in the cloud. It is a good feature because sometimes we are unable to install Pinecone on a local machine, so we can use the cloud. Pinecone provides credentials so we can directly connect to Pinecone using our script. It is a good feature, so I appreciate what Pinecone company has provided.

It is very fast and it saves us a lot of time for implementation.

Data privacy is important, and there are many layers of security provided by Pinecone.

Pinecone needs to be upgraded because many companies are not using Pinecone for production. I don't know why, but it is very useful for us because my team and I use Pinecone in many POCs. This is very useful for us, but on the production server, the client is not allowing us to use it.

Pinecone should be made ready for production servers. Many companies are not using Pinecone in production. I don't know the reason. We need to work on understanding why companies are not adopting it for production servers.

It would be better to provide better documentation on how to use it, and also provide some videos, because most of the time we are using videos for implementation and use. The documentation is also helpful, but videos are a good option for us.

I have used Pinecone for the last five years, when I started my career in generative AI.

Pinecone is good for POCs and small projects because it's very easy to implement and very easy to use. This is very good for us. I would rate this product a 10 out of 10.

My main use case with Pinecone  involved building a RAG application where we used it for indexing. In our RAG application using Pinecone  for indexing, we convert whatever text chunks are coming in into vector embeddings, which might be of different dimensions. We store these embeddings in Pinecone, and at query time, we perform a nearest neighbor search to find the most relevant text. For example, if you're asking a question about the different OPD expenses, then the embeddings most semantically similar to OPD expenses will come up. It's used for similarity search on text.

The best features that Pinecone offers include the separation via namespaces, which was really good. Another feature is the ability to deploy within different cloud servers of our choice, which was also advantageous, along with the out-of-the-box option that allows you to just use an API key and start using it without needing to set up anything.

Out of the features I mentioned, the most valuable in my workflow is the easy setup because it helps in prototyping easily. There can be different RAG use cases, and for each of these use cases, you just do not have to install a vector database or go through all of that; you can just use the API key.

Pinecone has positively impacted our organization by helping us to land a few clients or at least give a demo of how our application works, becoming a part of the solution we developed.

Pinecone can be improved by having the ability to have multi-modal embeddings out of the box as a good idea. A major reason we did not use Pinecone is that the serverless region was only in the United States; if it were available in India with serverless out-of-the-box implementation, we would have definitely used Pinecone.

Regarding needed improvements, I would like to see more regional endpoints, particularly serverless regional endpoints, as that's the most important one, along with multi-modality support.

I used the solution for about three to six months.

Pinecone is stable in my experience.

Pinecone is highly scalable compared to others, and it's out-of-the-box, although the drawback is the pricing.

Regarding how we set things up with Pinecone, we created different namespaces and within these namespaces, we created the embeddings for different catalogs.

We currently use PG vector with Postgres; it's not something I previously used, but it's beneficial because it operates on disk rather than in memory, which saves a lot of money.

Before choosing Pinecone, I evaluated options including PG vector with Postgres, Qdrant , and Milvus .

My advice for others looking into using Pinecone is to optimize all the embeddings before using it because if you are using it in a highly scalable manner, the pricing can get very high, so you have to be careful.

Pinecone was one of the earliest vector databases I came to know about, and it's the go-to option; I suggest it for anyone new to or learning about vector databases because it's very easy to start and work with without needing complex setups.

For future interviews, before asking deep questions, it's good to clarify things first because there is focus on cloud deployments, but we did not dive deeply into that; we only used Pinecone during the proof of concept and had to switch back due to data residency issues. I rate this product an 8 out of 10.

We are using Pinecone  because we have a good amount of documents which we use on a daily basis from our vendors. Based on those documents, we need to provide information to the end customer for that particular company. We have a UI where customers can ask any questions related to anything in the financial domain. We need to provide the latest information, so we are dynamically doing the chunking with the help of an OpenAI LLM model and then inserting into Pinecone . In Pinecone, we are using a very high dimension vector space, almost more than 3K dimension size. We then perform similarity search and provide the final response to the UI. For our RAG system implementation, we are using Pinecone.

There are multiple factors that impressed me while using Pinecone. I had an option to use Milvus  as well, but I preferred Pinecone. The first is the UI. Pinecone's UI is really strong. If I need to do some debugging on the backend side, I simply log into the UI and can perform operations based on my demand. This is a valuable UI feature.

Second is the scalability option. I can either define my own workers or use the auto-scaling feature. From an enterprise application and scalability perspective, this is very useful. We had an incident during a Black Friday sale and other occasional events that directly impacted our product traffic. Because we selected the auto-scaling feature in Pinecone, it automatically handled all the traffic spikes and we did not face any performance issues.

Ease  of troubleshooting is another valuable feature. If any transaction fails and we need to check and debug each transaction, we can perform a text search on the UI. Based on the text search, we get all the related vectors on the UI. The UI definitely helps us from a troubleshooting perspective. Selecting the infrastructure is also an important option. I can create multiple indexes based on demand so that it will not become messy for our enterprise application. Pinecone is the backbone of the entire system, helping us with cost and time savings.

I have a suggestion to expand the vector size in Pinecone. Right now, whatever the limit Pinecone is suggesting, I would recommend increasing that option. Currently, I believe it is around 3K, but if I want to go to 4K, 5K, or something higher, that would be beneficial. Embedding models are coming into the market and they are providing good amounts of vector sizes. Try to encapsulate all these features into Pinecone.

I have two main suggestions from my side. One is to increase the vector size. Currently, it is supporting only around 3K vector size, and I would recommend increasing that. The second suggestion involves creating a feature similar to LangSmith, which is a monitoring tool. In LangSmith, end-to-end API calls can be analyzed, showing what request came from the customer, what vector search was performed, what prompt was created, what call was given to the LLM, and what response was received from the LLM to the UI. The whole journey can be captured. I would appreciate if Pinecone could provide this capability from their side. I understand that Pinecone cannot capture the LLM call and everything. However, if it is possible, I could use the API key of Pinecone in my code where I can enable these feature logs and see all these things on the Pinecone dashboard.

The major improvement I am expecting from Pinecone is increased vector size. The second improvement would be to provide end-to-end debugging or the whole end-to-end call journey as a GenAI product, showing how the end-to-end journey works for a single request. If I am able to see the whole process on the Pinecone dashboard, it would be really valuable.

I am using Pinecone as an enterprise application in my organization for almost three years.

I do not have specific metrics, but I can give some high-level approximations. The task that was happening before developing this product was taking around one hour, but now it is done in hardly one or two minutes. So from 60 minutes to one or two minutes, you can assume how much cost savings we are achieving. Additionally, we are engaging customers much better through the UI. Previously, customers used to wait for 60 minutes. Now they get results within one or two minutes. We are definitely increasing our customer database.

Milvus  is another contender that we considered while deciding on a vector database. I suggested Pinecone because of its good amount of quality compared to Milvus. However, Milvus has the capability to handle any vector size, which is missing in Pinecone.