ApeRAG: Production-ready GraphRAG with multi-modal indexing and K8s deployment
github.comIn the list of features, it mentions:
> vision-based search for comprehensive document understanding
but it's not clear to me what this means, is it just vector embeddings for each image in every document via a CLIP-like model?
In addition, I'd be curious what's the rationale behind using the plethora of databases, given the docs on running it in production spins them all up, I assume they're all required, for instance I'd be curious on the trade-offs between using postgres with something like pg_search (for bm25 support, which vanilla postgres FTS doesn't have) vs using both postgres and ElasticSearch.
The docs are also very minimal, I'd have loved to see at least 1 example of usage.
Congrats on the launch. How does it compare to HelixDB?
HelixDB is a database. ApeRAG is an application that uses multiple databases (but that not particular one). Hypothetically, you could fork ApeRAG and modify it to use that database.
> ApeRAG requires PostgreSQL, Redis, Qdrant, and Elasticsearch. You have two options:
> bash ./02-install-database.sh # Deploys PostgreSQL, Redis, Qdrant, Elasticsearch
Is this built on top of all databases ? I am just trying to understand.
Yes, ApeRAG uses all these databases.
What does it use for the graph part? Elasticsearch? A Postgres plugin?
> bash ./02-install-database.sh # Deploys PostgreSQL, Redis, Qdrant, Elasticsearch
geez
sorry but, how much SHIT is it going to take to make AI good?
Maintaining databases is painful, so ApeRAG uses kubeblocks for all these databases.
What’s funny is Postgres alone can handle this entire workload decently well.
Postgres isn't a replacement for elastic. You CAN get full text search working in postgres, and for very basic use cases it's good enough, but it's vastly inferior to elastic in terms of features and performance.
Exactly. These "extensions".. they are good, but, far from being replacements for dedicated products, the quality of which customers have come to expect.
This is a very typical, and pretty bare-bones stack. Almost any production grade webapp above a minimal threshold of complexity will have database, cache, and search.
This is just one of a million other wrappers around AI that will be forgotten in a few months.
Yes, and that doesn't change the fact that the stack is typical.