nodes/olivv-cs-ComfyUI-FunPack
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

feat: chat runs show the whole message sent to the enhancer, instructions included

Browse Source 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
digital-garbage
2026-09-27 19:44:17 +03:00
co-authored by Claude Opus 5.5
parent eeddf46f71
commit 178e025a64
3 changed files with 24 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
conditioning.py
+7 -4
View File
@@ -9149,12 +9149,14 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
before answering, if it did, is left on `self._v2_last_thinking` for the readout.
"""
self._v2_last_thinking = ""
self._v2_last_sent = "" # chat runs: the whole message, for the readout
original = str(text or "").strip()
if not original:
return text, "Prompt enhancer: skipped; prompt empty."
key = (str(system_prompt), original, str(reference or ""), str(chat or ""))
if cache is not None and key in cache:
self._v2_last_thinking = cache.get(("thinking",) + key, "")
self._v2_last_sent = cache.get(("sent",) + key, "")
return cache[key], "Prompt enhancer: reused."
if isinstance(clip, _FunPackAdvisorLLMWrapper):
clip.last_thinking = "" # a failed call must not show the last one's
@@ -9164,6 +9166,7 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
# simply appended, a model rewrote it as more prompt text and echoed it back.
system_prompt = str(system_prompt) + "\n\n" + V2_PROMPT_ENHANCER_CHAT_SYSTEM
user_turn = "ORIGINAL PROMPT:\n" + user_turn + str(chat)
self._v2_last_sent = f"SYSTEM:\n{system_prompt}\n\nMESSAGE:\n{user_turn}"
raw, status = self._v2_generate_advisor_text(
clip, system_prompt, user_turn, seed=seed, image=image, thinking=bool(thinking),
max_length=max_length, temperature=temperature, top_p=top_p,
@@ -9185,6 +9188,7 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
if cache is not None:
cache[key] = enhanced
cache[("thinking",) + key] = self._v2_last_thinking
cache[("sent",) + key] = self._v2_last_sent
# `status` carries where it ran, how long it took and whether the model stopped on its
# own — all invisible from the text alone, and all of it is what a slow or runaway run
# needs in order to be diagnosed instead of guessed at.
@@ -10818,12 +10822,11 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
thinking=prompt_enhance_thinking, image=prompt_enhance_image,
**_enhance_sampling,
)
_chat_sent = self._v2_enhancer_chat(prompt_enhance_chat)
print(f"[FunPackVideoRefinerV2] {_enhance_status}")
self._v2_note_enhanced({
"scene": None, "before": _before, "after": prompt_to_encode,
"status": _enhance_status, "image": prompt_enhance_image is not None,
"reference": _ref.strip(), "chat": _chat_sent.strip(),
"reference": _ref.strip(), "chat": self._v2_last_sent,
"thinking": self._v2_last_thinking})
# Studio's `enhanced_prompt` output, for a node encoding on its own (Editor link
@@ -10845,7 +10848,7 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
self._v2_note_enhanced({
"scene": None, "before": _whole, "after": _after, "status": _st,
"image": prompt_enhance_image is not None, "reference": _ref.strip(),
"chat": self._v2_enhancer_chat(prompt_enhance_chat).strip(),
"chat": self._v2_last_sent,
"thinking": self._v2_last_thinking})
_whole = _after
self._v2_enhanced_output = _whole
@@ -11101,7 +11104,7 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
"scene": _i, "before": t, "after": _after, "status": _st,
"image": prompt_enhance_image is not None,
"reference": _ref.strip(),
"chat": self._v2_enhancer_chat(prompt_enhance_chat, _i).strip(),
"chat": self._v2_last_sent,
"thinking": self._v2_last_thinking})
_enhanced_texts.append(_after)
split_scene_texts = _enhanced_texts
movie_editor/frontend/composer.js
+3 -2
View File
@@ -1077,11 +1077,12 @@
el("div", "enh-before-text enh-ref-sent", ref));
card.append(det);
}
// The Chat tab's conversation exactly as the model received it (empty = none sent).
// Chat runs: the whole message exactly as the model received it, system instructions
// included, so what it was told can be checked rather than assumed.
const chat = String(it.chat || "").trim();
if (chat) {
const det = el("details", "enh-before");
det.append(el("summary", null, "Chat sent"), el("div", "enh-before-text enh-ref-sent", chat));
det.append(el("summary", null, "Sent to the model (chat)"), el("div", "enh-before-text enh-ref-sent", chat));
card.append(det);
}
if (it.status) card.append(el("div", "enh-status", String(it.status)));
tests/test_prompt_enhancer.py
+14
View File
@@ -674,3 +674,17 @@ def test_every_readout_says_what_chat_the_model_was_sent():
import inspect
src = inspect.getsource(C.FunPackVideoRefinerV2.refine_v2)
assert src.count('"chat": ') == 3
def test_a_chat_run_records_the_whole_message_for_the_readout(studio):
chat = C.FunPackVideoRefinerV2._v2_enhancer_chat([{"rewrites": {"whole": "a fox at dusk"},
"comment": "make it night"}])
cache = {}
studio._v2_enhance_prompt(FakeClip(), "a fox", "my rules", chat=chat, cache=cache)
sent = studio._v2_last_sent
for part in ("SYSTEM:", "my rules", "REVISION", "MESSAGE:", "ORIGINAL PROMPT:", "a fox at dusk", "make it night"):
assert part in sent, part
studio._v2_enhance_prompt(FakeClip(), "a fox", "my rules", chat=chat, cache=cache)
assert studio._v2_last_sent == sent # a reused answer reports the same
studio._v2_enhance_prompt(FakeClip(), "a fox", "my rules")
assert studio._v2_last_sent == ""