ask-source-chunks-persist-across-tool-iterations

IN premise

Source document content included in the initial ask prompt is re-included in all subsequent prompts after tool-call round-trips, ensuring the LLM retains source context across the entire multi-turn loop.

Summary

When source documents are provided at the start of a conversation, they stay visible to the model throughout the entire tool-use loop rather than being dropped after the first turn. This means the LLM can reference the original source material at every step, so it won't lose track of what it was asked to work with even after multiple rounds of calling tools and processing results.

Details

Sourceentries/2026/05/11/tests-test_ask.md