obstacles in reading large files with AI agents

[dan-oak]

Body

What are you proposing?

i just want to discuss the current feature of Zed and propose to add more configurations:

when AI agent tries to read long files, it can get hit with this:

SUCCESS: File outline retrieved. This file is too large to read all at once, so the outline below shows the file's structure with line numbers.

IMPORTANT: Do NOT retry this call without line numbers - you will get the same outline.
Instead, use the line numbers below to read specific sections by calling this tool again with start_line and end_line parameters.

# File outline for ...

... (Markdown outline TOC-style with L##-L## markers)

i've been using Zed for 2 months and my model (DeepSeek) never actually listens to those instructions. most of the time i myself don't want it to follow those instructions, as i just want it to read the whole files anyway.

i suspect this is hardcoded in Zed system prompt because of some (as far as i understand) "context optimization" reasons and to reduce cost for expensive providers. in most of the cases, i do not care about all that. i use an affordable model which is capable of reading huge files and do simple tasks on it.

i am also very surprised the system considers my 500 lines of Markdown text file or 1K lines of C code as "too big too read in full".

Why does this matter?

it just wastes so much time for me. i wait when the LLM iterates throught the file L1-200, L200-400, and so on until the end.

Possible approach

let us configure this?

---

and i am also interested which models actually follow those instructions in your experience? maybe how do you work-around this un-configurable Zed behaviour?