Add a /v1/internal/chat-prompt endpoint

extensions/openai/completions.py

@@ -135,6 +135,7 @@ def convert_history(history):
     current_message = ""
     current_reply = ""
     user_input = ""
+    user_input_last = True
     system_message = ""
 
     # Multimodal: convert OpenAI format to multimodal extension format
@@ -188,13 +189,15 @@ def convert_history(history):
 
         if role == "user":
             user_input = content
+            user_input_last = True
             if current_message:
                 chat_dialogue.append([current_message, ''])
                 current_message = ""
 
             current_message = content
         elif role == "assistant":
             current_reply = content
+            user_input_last = False
             if current_message:
                 chat_dialogue.append([current_message, current_reply])
                 current_message = ""
@@ -204,13 +207,13 @@ def convert_history(history):
         elif role == "system":
             system_message = content
 
-    # if current_message:
-    #     chat_dialogue.append([current_message, ''])
+    if not user_input_last:
+        user_input = ""
 
     return user_input, system_message, {'internal': chat_dialogue, 'visible': copy.deepcopy(chat_dialogue)}
 
 
-def chat_completions_common(body: dict, is_legacy: bool = False, stream=False) -> dict:
+def chat_completions_common(body: dict, is_legacy: bool = False, stream=False, prompt_only=False) -> dict:
     if body.get('functions', []):
         raise InvalidRequestError(message="functions is not supported.", param='functions')
 
@@ -310,14 +313,18 @@ def chat_streaming_chunk(content):
         #    chunk[resp_list][0]["logprobs"] = None
         return chunk
 
-    if stream:
-        yield chat_streaming_chunk('')
-
     # generate reply #######################################
-    prompt = generate_chat_prompt(user_input, generate_params)
+    prompt = generate_chat_prompt(user_input, generate_params, _continue=continue_)
+    if prompt_only:
+        yield {'prompt': prompt}
+        return
+
     token_count = len(encode(prompt)[0])
     debug_msg({'prompt': prompt, 'generate_params': generate_params})
 
+    if stream:
+        yield chat_streaming_chunk('')
+
     generator = generate_chat_reply(
         user_input, generate_params, regenerate=False, _continue=continue_, loading_message=False)
 

extensions/openai/models.py

@@ -9,7 +9,8 @@
 def get_current_model_info():
     return {
         'model_name': shared.model_name,
-        'lora_names': shared.lora_names
+        'lora_names': shared.lora_names,
+        'loader': shared.args.loader
     }
 
 

extensions/openai/script.py

@@ -3,6 +3,7 @@
 import logging
 import os
 import traceback
+from collections import deque
 from threading import Thread
 
 import speech_recognition as sr
@@ -31,6 +32,7 @@
 from .typing import (
     ChatCompletionRequest,
     ChatCompletionResponse,
+    ChatPromptResponse,
     CompletionRequest,
     CompletionResponse,
     DecodeRequest,
@@ -259,6 +261,15 @@ async def handle_logits(request_data: LogitsRequest):
     return JSONResponse(response)
 
 
+@app.post('/v1/internal/chat-prompt', response_model=ChatPromptResponse, dependencies=check_key)
+async def handle_chat_prompt(request: Request, request_data: ChatCompletionRequest):
+    path = request.url.path
+    is_legacy = "/generate" in path
+    generator = OAIcompletions.chat_completions_common(to_dict(request_data), is_legacy=is_legacy, prompt_only=True)
+    response = deque(generator, maxlen=1).pop()
+    return JSONResponse(response)
+
+
 @app.post("/v1/internal/stop-generation", dependencies=check_key)
 async def handle_stop_generation(request: Request):
     stop_everything_event()

extensions/openai/typing.py

@@ -124,6 +124,10 @@ class ChatCompletionResponse(BaseModel):
     usage: dict
 
 
+class ChatPromptResponse(BaseModel):
+    prompt: str
+
+
 class EmbeddingsRequest(BaseModel):
     input: str | List[str] | List[int] | List[List[int]]
     model: str | None = Field(default=None, description="Unused parameter. To change the model, set the OPENEDAI_EMBEDDING_MODEL and OPENEDAI_EMBEDDING_DEVICE environment variables before starting the server.")

modules/models_settings.py

@@ -136,9 +136,6 @@ def get_model_metadata(model):
     if 'instruction_template' not in model_settings:
         model_settings['instruction_template'] = 'Alpaca'
 
-    if model_settings['instruction_template'] != 'Custom (obtained from model metadata)':
-        model_settings['instruction_template_str'] = chat.load_instruction_template(model_settings['instruction_template'])
-
     # Ignore rope_freq_base if set to the default value
     if 'rope_freq_base' in model_settings and model_settings['rope_freq_base'] == 10000:
         model_settings.pop('rope_freq_base')
@@ -150,6 +147,10 @@ def get_model_metadata(model):
             for k in settings[pat]:
                 model_settings[k] = settings[pat][k]
 
+    # Load instruction template if defined by name rather than by value
+    if model_settings['instruction_template'] != 'Custom (obtained from model metadata)':
+        model_settings['instruction_template_str'] = chat.load_instruction_template(model_settings['instruction_template'])
+
     return model_settings