fix(tools): ReadRange resilience for weak model inputs#567
Open
JessicaMulein wants to merge 1 commit into
Open
fix(tools): ReadRange resilience for weak model inputs#567JessicaMulein wants to merge 1 commit into
JessicaMulein wants to merge 1 commit into
Conversation
JessicaMulein
force-pushed
the
pr/readrange-resilience
branch
from
f8ea0d4 to
590d9aa
Compare
- is_provided() now rejects common junk values ('N/A', 'null', 'none')
that small models pass instead of real text markers
- ReadRange auto-fallback: when start_text/end_text are missing or junk
but the file is already in editable or read-only context, show the
whole file (@000 → 000@) instead of erroring. Files not in context
still get the strict error.
Reduces wasted turns on quantized local models (qwen3.6:27b-q4_K_M)
that struggle with ReadRange's start_text/end_text contract.
JessicaMulein
force-pushed
the
pr/readrange-resilience
branch
from
590d9aa to
d97608e
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
is_provided()now rejects common junk values ('N/A','null','none') that small quantized models pass instead of real text markers@000→000@) instead of erroring. Files not in context still get the strict error.Reduces wasted turns on quantized local models (e.g. qwen3.6:27b-q4_K_M) that struggle with ReadRange's start_text/end_text contract.
Test plan
python -m pytest tests/ -x -q(414 passed)