mirror of
https://github.com/infiniflow/ragflow.git
synced 2026-01-30 07:06:39 +08:00
c8bd413e4c
### What problem does this PR solve? During figure enhancement, some cropped figure images are extremely small. Sending these to the Image2Text/VLM provider fails with a 400 invalid_parameter_error because the image width/height must be >10px. This aborts the enhancement step. This PR adds a minimal size guard to skip tiny crops and continue processing. <img width="1084" height="494" alt="image" src="https://github.com/user-attachments/assets/ad074270-94e6-4571-91c8-37df85212639" /> ### Type of change - [x] Bug Fix (non-breaking change which fixes an issue)
131 lines
4.3 KiB
Python
131 lines
4.3 KiB
Python
#
|
|
# Copyright 2025 The InfiniFlow Authors. All Rights Reserved.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
#
|
|
|
|
import asyncio
|
|
import io
|
|
import re
|
|
|
|
import numpy as np
|
|
from PIL import Image
|
|
|
|
from api.db.services.llm_service import LLMBundle
|
|
from common.constants import LLMType
|
|
from common.string_utils import clean_markdown_block
|
|
from deepdoc.vision import OCR
|
|
from rag.nlp import attach_media_context, rag_tokenizer, tokenize
|
|
|
|
ocr = OCR()
|
|
|
|
# Gemini supported MIME types
|
|
VIDEO_EXTS = [".mp4", ".mov", ".avi", ".flv", ".mpeg", ".mpg", ".webm", ".wmv", ".3gp", ".3gpp", ".mkv"]
|
|
|
|
|
|
def chunk(filename, binary, tenant_id, lang, callback=None, **kwargs):
|
|
doc = {
|
|
"docnm_kwd": filename,
|
|
"title_tks": rag_tokenizer.tokenize(re.sub(r"\.[a-zA-Z]+$", "", filename)),
|
|
}
|
|
eng = lang.lower() == "english"
|
|
|
|
parser_config = kwargs.get("parser_config", {}) or {}
|
|
image_ctx = max(0, int(parser_config.get("image_context_size", 0) or 0))
|
|
|
|
if any(filename.lower().endswith(ext) for ext in VIDEO_EXTS):
|
|
try:
|
|
doc.update(
|
|
{
|
|
"doc_type_kwd": "video",
|
|
}
|
|
)
|
|
cv_mdl = LLMBundle(tenant_id, llm_type=LLMType.IMAGE2TEXT, lang=lang)
|
|
ans = asyncio.run(
|
|
cv_mdl.async_chat(system="", history=[], gen_conf={}, video_bytes=binary, filename=filename))
|
|
callback(0.8, "CV LLM respond: %s ..." % ans[:32])
|
|
ans += "\n" + ans
|
|
tokenize(doc, ans, eng)
|
|
return [doc]
|
|
except Exception as e:
|
|
callback(prog=-1, msg=str(e))
|
|
else:
|
|
img = Image.open(io.BytesIO(binary)).convert("RGB")
|
|
doc.update(
|
|
{
|
|
"image": img,
|
|
"doc_type_kwd": "image",
|
|
}
|
|
)
|
|
bxs = ocr(np.array(img))
|
|
txt = "\n".join([t[0] for _, t in bxs if t[0]])
|
|
callback(0.4, "Finish OCR: (%s ...)" % txt[:12])
|
|
if (eng and len(txt.split()) > 32) or len(txt) > 32:
|
|
tokenize(doc, txt, eng)
|
|
callback(0.8, "OCR results is too long to use CV LLM.")
|
|
return attach_media_context([doc], 0, image_ctx)
|
|
|
|
try:
|
|
callback(0.4, "Use CV LLM to describe the picture.")
|
|
cv_mdl = LLMBundle(tenant_id, LLMType.IMAGE2TEXT, lang=lang)
|
|
img_binary = io.BytesIO()
|
|
img.save(img_binary, format="JPEG")
|
|
img_binary.seek(0)
|
|
ans = cv_mdl.describe(img_binary.read())
|
|
callback(0.8, "CV LLM respond: %s ..." % ans[:32])
|
|
txt += "\n" + ans
|
|
tokenize(doc, txt, eng)
|
|
return attach_media_context([doc], 0, image_ctx)
|
|
except Exception as e:
|
|
callback(prog=-1, msg=str(e))
|
|
|
|
return []
|
|
|
|
|
|
def vision_llm_chunk(binary, vision_model, prompt=None, callback=None):
|
|
"""
|
|
A simple wrapper to process image to markdown texts via VLM.
|
|
|
|
Returns:
|
|
Simple markdown texts generated by VLM.
|
|
"""
|
|
callback = callback or (lambda prog, msg: None)
|
|
|
|
img = binary
|
|
txt = ""
|
|
|
|
try:
|
|
# Skip tiny crops that fail provider image-size limits.
|
|
if hasattr(img, "size"):
|
|
min_side = 11
|
|
if img.size[0] < min_side or img.size[1] < min_side:
|
|
callback(0.0, f"Skip tiny image for VLM: {img.size[0]}x{img.size[1]}")
|
|
return ""
|
|
with io.BytesIO() as img_binary:
|
|
try:
|
|
img.save(img_binary, format="JPEG")
|
|
except Exception:
|
|
img_binary.seek(0)
|
|
img_binary.truncate()
|
|
img.save(img_binary, format="PNG")
|
|
|
|
img_binary.seek(0)
|
|
ans = clean_markdown_block(vision_model.describe_with_prompt(img_binary.read(), prompt))
|
|
txt += "\n" + ans
|
|
return txt
|
|
|
|
except Exception as e:
|
|
callback(-1, str(e))
|
|
|
|
return ""
|