mirrors/hermes-agent
1
0
Fork 0
mirror of https://github.com/NousResearch/hermes-agent.git synced 2026-06-12 08:51:53 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 14

fix(agent): steer GPT/Codex family to V4A for single-file edits too (#44411)

Browse source 
The coding-posture brief told GPT/Codex models to use patch mode='patch'
(V4A) for structured/multi-file changes but mode='replace' "for a single
small swap". That second nudge points those models at a format their
first-party harness never taught them.

Verified against openai/codex (current main): apply_patch is the ONLY file
editor in codex-rs — zero occurrences of str_replace/old_string anywhere in
the repo; the grammar (core/src/tools/handlers/apply_patch.lark) is exactly
the V4A dialect our patch_parser implements; the shipped model prompts
(gpt_5_codex, gpt-5.2-codex, gpt-5.1-codex-max + instruction templates)
explicitly say to use apply_patch "for single file edits"; and the tool is
gated per model via ModelInfo.apply_patch_tool_type, i.e. OpenAI ships
V4A-for-everything as model metadata.

The GPT-family line now steers to mode='patch' for all edits, single-file
included. The replace-family line (Claude + open-weight) is unchanged —
Claude Code's FileEdit is old_string/new_string/replace_all exact string
replacement (confirmed from Anthropic's shipped sdk-tools.d.ts, the only
file editor in its tool union), matching our mode='replace'.
This commit is contained in:
brooklyn! 2026-06-11 12:52:52 -05:00 • committed by GitHub
parent cb29e8a82e
commit a4f179c509
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 13 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

12
agent/coding_context.py
View file

@ -106,13 +106,19 @@ _GIT_TIMEOUT = 2.5
# multi-file) and mode="replace" (find-and-swap). We nudge each family toward
# its native format. Unknown families get nothing (the brief's neutral wording
# stands). Substrings match the model id; aligned with TOOL_USE_ENFORCEMENT_MODELS.
#
# GPT/Codex get V4A for ALL edits, single-file included: in codex-rs,
# apply_patch (V4A — apply_patch.lark) is the ONLY file editor, no
# str_replace-style tool exists, and the shipped model prompts say to use
# apply_patch even "for single file edits" — so a replace-mode nudge would
# steer those models toward a format their first-party harness never taught
# them.
_EDIT_FORMAT_GUIDANCE: dict[str, tuple[tuple[str, ...], str]] = {
"patch": (
("gpt", "codex"),
"- Edit format: author new files with `write_file`; for edits to "
"existing code prefer `patch` with `mode='patch'` (V4A multi-file diff) "
"for structured or multi-file changes — it's the diff format you handle "
"most reliably. Use `mode='replace'` for a single small swap.",
"existing code use `patch` with `mode='patch'` (V4A diff) — including "
"single-file edits. It's the edit format you handle most reliably.",
),
"replace": (
("claude", "sonnet", "opus", "haiku",

4
tests/agent/test_coding_context.py
View file

@ -312,6 +312,10 @@ class TestEditFormatSteering:
assert "mode='patch'" in brief
assert "V4A" in brief
assert "write_file" in brief # new files authored, not patched
# Codex-family harnesses ship apply_patch (V4A) as the ONLY editor and
# instruct it even for single-file edits — never nudge replace mode.
assert "single-file" in brief
assert "mode='replace'" not in brief
def test_anthropic_family_gets_replace_nudge(self, tmp_path):
_git_init(tmp_path)