Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-19 09:15:18 +02:00
Code Issues Packages Projects Releases Wiki Activity

mcp media strip normalize

Browse Source
This commit is contained in:
Concedo
2026-08-02 10:46:26 +08:00
parent 4423b3af55
commit 9e64023c7b
1 changed files with 37 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files
koboldcpp.py
+37 -12
View File
@@ -4077,6 +4077,28 @@ def strip_oaicontent_of_media(oaicontent):
return outarr
return oaicontent
def get_base64_from_media_data(data):
if isinstance(data, str) and data.startswith("data:") and "," in data:
return data.split(",", 1)[1]
return data
def sweep_media_from_mcpcontent(mcpcontentstr):
images = []
try:
if isinstance(mcpcontentstr, str):
mcp_pl = json.loads(mcpcontentstr)
if isinstance(mcp_pl, dict) and isinstance(mcp_pl.get("content",None),list):
for item in mcp_pl.get("content",[]):
if not isinstance(item, dict):
continue
if item.get("type","")=="image":
data = get_base64_from_media_data(item.get("data",""))
if data:
images.append(data)
except Exception:
pass
return images
def strip_mcpcontent_of_media(mcpcontentstr):
try:
if isinstance(mcpcontentstr, str):
@@ -4086,7 +4108,7 @@ def strip_mcpcontent_of_media(mcpcontentstr):
if isinstance(mcp_pl, dict) and isinstance(mcp_pl.get("content",None),list):
pl_arr = mcp_pl.get("content",[])
for idx in range(len(pl_arr)):
if pl_arr[idx].get("type","")=="image" and pl_arr[idx].get("data","")!="":
if isinstance(pl_arr[idx], dict) and pl_arr[idx].get("type","")=="image" and pl_arr[idx].get("data","")!="":
pl_arr[idx]["data"] = "(base64 data attached)"
pl_modified = True
if pl_modified:
@@ -4128,7 +4150,10 @@ def determine_tool_json_to_use(genparams, curr_ctx, assistant_message_start, is_
for message in reversed_messages:
if message["role"] == "tool":
toolrespstr = message["content"]
# toolrespstr = strip_mcpcontent_of_media(toolrespstr)
if isinstance(toolrespstr, str):
toolrespstr = strip_mcpcontent_of_media(toolrespstr)
else:
toolrespstr = strip_oaicontent_of_media(toolrespstr)
tool_call_chunk.append(toolrespstr)
else:
break
@@ -4244,20 +4269,13 @@ def sweep_media_from_messages(messages_array):
elif item.get("type") == "image": #handle mcp image content blocks in a list
data = item.get("data", "")
if data:
images.append(data.split(",", 1)[1] if data.startswith("data:") else data)
images.append(get_base64_from_media_data(data))
elif item.get("type") == "input_audio":
data = item.get("input_audio", {}).get("data")
if data:
audio.append(data)
elif message.get("role", "")=="tool" and isinstance(curr_content, str): #handle mcp returned images
try:
mcp_pl = json.loads(curr_content)
if isinstance(mcp_pl, dict) and isinstance(mcp_pl.get("content",None),list):
pl_arr = mcp_pl.get("content",[])
if len(pl_arr)>0 and pl_arr[0].get("type","")=="image" and pl_arr[0].get("data","")!="":
images.append(pl_arr[0].get("data",""))
except Exception:
pass
images.extend(sweep_media_from_mcpcontent(curr_content))
imgs_ollama = message.get("images", None)
if imgs_ollama:
for img in imgs_ollama:
@@ -4460,7 +4478,8 @@ ws ::= | " " | "\n" [ \t]{0,20}
messages_string += "\n(Made a function call)\n"
pass # do nothing
elif isinstance(curr_content, str):
if latest_turn_was_tool and message_index < len(messages_array):
if latest_turn_was_tool:
images_added.extend(sweep_media_from_mcpcontent(curr_content))
curr_content = strip_mcpcontent_of_media(curr_content)
messages_string += curr_content
elif isinstance(curr_content, list): #is an array
@@ -4473,6 +4492,12 @@ ws ::= | " " | "\n" [ \t]{0,20}
images_added.append(item['image_url']['url'].split(",", 1)[1])
attachedimgid += 1
messages_string += f"\n(Attached Image {attachedimgid})\n"
elif item['type']=="image":
data = get_base64_from_media_data(item.get("data", ""))
if data:
images_added.append(data)
attachedimgid += 1
messages_string += f"\n(Attached Image {attachedimgid})\n"
elif item['type']=="input_audio":
if 'input_audio' in item and item['input_audio'] and item['input_audio']['data']:
audio_added.append(item['input_audio']['data'])