OpenWebUI/open-webui
1
0
Fork 0
mirror of https://github.com/open-webui/open-webui synced 2025-06-26 18:26:48 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

refac

Browse Source
This commit is contained in:
Timothy Jaeryang Baek
2024-12-12 23:31:08 -08:00
parent 9a081c8593
commit f9a05dd1e1
6 changed files with 190 additions and 144 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
backend/open_webui/utils/chat.py
View File

@@ -136,7 +136,7 @@ async def generate_chat_completion(
response = await generate_ollama_chat_completion(
request=request, form_data=form_data, user=user, bypass_filter=bypass_filter
)
if form_data.stream:
if form_data.get("stream"):
response.headers["content-type"] = "text/event-stream"
return StreamingResponse(
convert_streaming_response_ollama_to_openai(response),

10
backend/open_webui/utils/misc.py
View File

@@ -106,7 +106,7 @@ def openai_chat_message_template(model: str):
def openai_chat_chunk_message_template(
model: str, message: Optional[str] = None
model: str, message: Optional[str] = None, usage: Optional[dict] = None
) -> dict:
template = openai_chat_message_template(model)
template["object"] = "chat.completion.chunk"
@@ -114,17 +114,23 @@ def openai_chat_chunk_message_template(
template["choices"][0]["delta"] = {"content": message}
else:
template["choices"][0]["finish_reason"] = "stop"
if usage:
template["usage"] = usage
return template
def openai_chat_completion_message_template(
model: str, message: Optional[str] = None
model: str, message: Optional[str] = None, usage: Optional[dict] = None
) -> dict:
template = openai_chat_message_template(model)
template["object"] = "chat.completion"
if message is not None:
template["choices"][0]["message"] = {"content": message, "role": "assistant"}
template["choices"][0]["finish_reason"] = "stop"
if usage:
template["usage"] = usage
return template

57
backend/open_webui/utils/response.py
View File

@@ -21,8 +21,63 @@ async def convert_streaming_response_ollama_to_openai(ollama_streaming_response)
message_content = data.get("message", {}).get("content", "")
done = data.get("done", False)
usage = None
if done:
usage = {
"response_token/s": (
round(
(
(
data.get("eval_count", 0)
/ ((data.get("eval_duration", 0) / 1_000_000_000))
)
* 100
),
2,
)
if data.get("eval_duration", 0) > 0
else "N/A"
),
"prompt_token/s": (
round(
(
(
data.get("prompt_eval_count", 0)
/ (
(
data.get("prompt_eval_duration", 0)
/ 1_000_000_000
)
)
)
* 100
),
2,
)
if data.get("prompt_eval_duration", 0) > 0
else "N/A"
),
"total_duration": round(
((data.get("total_duration", 0) / 1_000_000) * 100), 2
),
"load_duration": round(
((data.get("load_duration", 0) / 1_000_000) * 100), 2
),
"prompt_eval_count": data.get("prompt_eval_count", 0),
"prompt_eval_duration": round(
((data.get("prompt_eval_duration", 0) / 1_000_000) * 100), 2
),
"eval_count": data.get("eval_count", 0),
"eval_duration": round(
((data.get("eval_duration", 0) / 1_000_000) * 100), 2
),
"approximate_total": (
lambda s: f"{s // 3600}h{(s % 3600) // 60}m{s % 60}s"
)((data.get("total_duration", 0) or 0) // 1_000_000_000),
}
data = openai_chat_chunk_message_template(
model, message_content if not done else None
model, message_content if not done else None, usage
)
line = f"data: {json.dumps(data)}\n\n"