dearwolf/aider
1
0
Fork 0
mirror of https://github.com/Aider-AI/aider.git synced 2025-06-01 10:14:59 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 3

feat: add support for both images and PDFs based on model capabilities

Browse source
This commit is contained in:
Paul Gauthier (aider) 2024-11-26 17:16:43 -08:00
parent aaeaa24153
commit a9c4647461
1 changed files with 14 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

18
aider/coders/base_coder.py
View file

@ -708,8 +708,11 @@ class Coder:
return chat_files_messages return chat_files_messages
def get_images_message(self): def get_images_message(self):
#if not self.main_model.info.get("supports_vision"): supports_images = self.main_model.info.get("supports_vision")
# return None supports_pdfs = self.main_model.info.get("supports_pdf_input")
if not (supports_images or supports_pdfs):
return None
image_messages = [] image_messages = []
for fname, content in self.get_abs_fnames_content(): for fname, content in self.get_abs_fnames_content():
@ -717,14 +720,21 @@ class Coder:
with open(fname, "rb") as image_file: with open(fname, "rb") as image_file:
encoded_string = base64.b64encode(image_file.read()).decode("utf-8") encoded_string = base64.b64encode(image_file.read()).decode("utf-8")
mime_type, _ = mimetypes.guess_type(fname) mime_type, _ = mimetypes.guess_type(fname)
dump(mime_type)
if not mime_type: if not mime_type:
continue continue
if mime_type.startswith("image/") or mime_type == "application/pdf":
if mime_type.startswith("image/") and supports_images:
image_url = f"data:{mime_type};base64,{encoded_string}" image_url = f"data:{mime_type};base64,{encoded_string}"
rel_fname = self.get_rel_fname(fname) rel_fname = self.get_rel_fname(fname)
image_messages += [ image_messages += [
{"type": "text", "text": f"Image file: {rel_fname}"}, {"type": "text", "text": f"Image file: {rel_fname}"},
{"type": "image_url", "image_url": {"url": image_url, "detail": "high"}},
]
elif mime_type == "application/pdf" and supports_pdfs:
image_url = f"data:{mime_type};base64,{encoded_string}"
rel_fname = self.get_rel_fname(fname)
image_messages += [
{"type": "text", "text": f"PDF file: {rel_fname}"},
{"type": "image_url", "image_url": image_url}, {"type": "image_url", "image_url": image_url},
] ]