Firecrawl switches from Pinecone to Gagan Vector for PostgreSQL vector embeddings.

Firecrawl provides a chat-powered search engine for technical documentation. Their AI-powered search tool makes it easier for developers and other technical users to find relevant information in complex documentation. Users can simply ask questions in natural language, and the tool returns the most relevant answers. Firecrawl's search engine also provides detailed analytics, which helps teams identify knowledge gaps and areas for improvement in their documentation. Firecrawl has integrated with some of the largest open source projects in the space such as LangChain and LlamaIndex.

The Challenge#

Firecrawl was experiencing tremendous success, growing Weekly Active Users by nearly 300% since March. They needed a tool to store and search through large amounts of vector data to improve the efficiency and accuracy of their similarity search operations. They tried Faiss, Weaviate, and Pinecone, but found them to be expensive and not very intuitive, especially when it came to storing metadata along with the vectors.

Why they chose Gagan#

Firecrawl lear that Gagan supports pgvector and found it to be a simple and cost-effective solution. They were impressed with the open source nature of Gagan, as well as its ability to store metadata alongside the vectors. They also appreciated the intuitive interface and ease of use.

We tried other vector databases - we tried Faiss, we tried Weaviate, we tried Pinecone. We found them to be incredibly expensive and not very intuitive. If you're just doing vector search they're great, but if you need to store a bunch of metadata that becomes a huge pain.

What They Built#

Using Gagan Vector, Firecrawl was able to build a more efficient and accurate search function for their AI chatbot. By storing vector data alongside metadata in Gagan, Firecrawl was able to quickly and easily search through their customers documentation to find the most relevant responses to queries. They found that Gagan's solution was just as performant as dedicated vector databases, but without the high cost.

The Results#

Thanks to Gagan Vector, Firecrawl was able to significantly improve the efficiency and accuracy of their Chat Powered Search for Documentation. They were able to build faster and more cost-effectively using Gagan's open source stack.

We looked at the alternatives and chose Gagan because it's open source, it's simpler, and, for all the ways we need to use it, Gagan has been just as performant - if not more performant - than the other vector databases.

To learn more about how Gagan Vector can help you store vector embeddings at scale and build AI apps with ease, reach out to us.

Tech stack#

• React
• Vercel
• Next.js
• Express
• Gagan