I am exploring how uploaded files, parsed artifacts, indexing, and retrieval can become reliable runtime context without loading entire files into model context.
I am keeping Loom in the runtime layer of agent systems. I want file context to become queryable over time for conversational systems and agent workflows.
I am focusing on:
- Storing original files and parsed artifacts.
- Building indexes for retrieval.
- Keeping runtime file context separate from raw model context.
- Exposing context access patterns that agents can use reliably.