How to Provide LLMs with Massive Context

Firm-Track3617 · reddit · 2026-07-07

A developer is building a platform where an LLM needs to reference numerous existing nodes (e.g., correctly citing many previously defined nodes when generating a DAG). They are weighing how to balance latency, cost, and inference quality: is context caching suitable for requests where most context remains unchanged? Or would RAG with a vector database be better? The developer is concerned that the model needs to reference a large number of structured nodes rather than just retrieving a few semantically similar items. They are seeking experiences and advice on handling ultra-large-scale structured context.

Related event: Developers Debate: Context Caching vs. RAG for Massive LLM Contexts(3 posts)→

Original post →

More from coding & agent

coding & agent channel →