Context Engineering: Phase 2

## What problem does this solve?
What is the right information to solve a problem that we need to provide to the agent?

### Goal:
Sweet spot between the following 3 params. We need multiple datapoints to identify this.

### Solution Space:
Chat History
Files
RAG/external sources
Other

### Resolve Rate
Number of Turns
Cache hits/misses (prefix caching)

## How will it work?
By collecting and refining the precise context sent to them model there by improving the generated output.
Metrics we will be evaluating against are resolve rate and number of turns for model quality. For efficiency we will measure input context size and number of API requests/task

### Why is this needed?

Addressed by #4081

### Additional context

Over the past quarter we have made multiple improvements in this space and plan to continue implementing improvements in the coming months.