"I built an MCP server that connects Claude Code to 2M+ research papers - it stops defaulting to what it "knows" and starts using published, benchmarked and latest methods.

How Claude Code built it:

Embedding pipeline: Qwen3-Embedding on AWS g5 instances, USearch HNSW index, #LMDB cache for 2M+ CS papers

FastAPI MCP server with multi-query generation and synthesis

Elasticsearch BM25 indexing across the full corpus"

https://www.reddit.com/r/ClaudeAI/comments/1ruqw21/i_built_an_mcp_server_that_connects_claude_code/

I Ditched Elasticsearch for Meilisearch. Here's What Nobody Tells You.

https://www.anisafifi.com/en/blog/i-ditched-elasticsearch-for-meilisearch-heres-what-nobody-tells-you/

I spent one afternoon replacing it with Meilisearch. My p99 search latency went from 180ms to 12ms. My infrastructure bill dropped to $14/month (from $120). My configuration went from 340 lines of YAML to about 30.

Meilisearch stores everything in memory-mapped files using #LMDB a high-performance embedded key-value store originally built for the OpenLDAP project. This is why search is fast

Datum Cloud Authoritative DNS Service: Research Report

https://gist.github.com/drewr/2055b97dc8e7518cc298e58413688459

"The standard problem with #LMDB in multi-pod deployments is that it cannot be shared across nodes."

The entire reason you use containers is for isolation. Make up your mind, do you want sharing, or don't you?

You can share LMDB between containers, using a few non-default settings. Mainly, containers must share PID namespace, instead of always running as their own PID 1.