Doesnt work

api/api/api.py

@@ -6,7 +6,6 @@
 api.add_router("/", "documents.views.router")
 api.add_router("/", "chunks.views.router")
 api.add_router("/", "bot.views.router")
-
 urlpatterns = [
     path("api/", api.urls),
 ]

api/bot/controllers.py

@@ -1,13 +1,39 @@
 from http import HTTPStatus
 
 import requests
+from llama_index.llms.llama_cpp import LlamaCPP
 from pgvector.django import L2Distance
 
 from bot.schemas import BotIn, BotOut
 from chunks.models import Chunk
 
 
+def messages_to_prompt(messages):
+    prompt = ""
+    for message in messages:
+        if message.role == 'system':
+            prompt += f"<|system|>\n{message.content}</s>\n"
+        elif message.role == 'user':
+            prompt += f"<|user|>\n{message.content}</s>\n"
+        elif message.role == 'assistant':
+            prompt += f"<|assistant|>\n{message.content}</s>\n"
+
+    # ensure we start with a system prompt, insert blank if needed
+    if not prompt.startswith("<|system|>\n"):
+        prompt = "<|system|>\n</s>\n" + prompt
+
+    # add final assistant prompt
+    prompt = prompt + "<|assistant|>\n"
+
+    return prompt
+
+
+def completion_to_prompt(completion):
+    return f"<|system|>\n</s>\n<|user|>\n{completion}</s>\n<|assistant|>\n"
+
+
 def query_llm_controller(payload: BotIn) -> tuple[HTTPStatus, BotOut]:
+    # https: // docs.llamaindex.ai / en / stable / api_reference / llms / llama_cpp /
     # TODO: payload na embeding  -> Vector -> Szukamy w bazie podobne -> dokument do payloada
     embeddings_body = {
         "input": payload.input
@@ -17,12 +43,27 @@ def query_llm_controller(payload: BotIn) -> tuple[HTTPStatus, BotOut]:
     similar_chunk = Chunk.objects.order_by(L2Distance('embedding', input_embedding))[0]
     print(similar_chunk.text)
     llm_body = {
-        "prompt": "\n\n### Instructions:\nOdpowiedz na pytanie "+ payload.input +"\n\nWiedząc że"+similar_chunk.text+"\n\n### Response:\n",
+        "prompt": "\n\n### Instructions:\nOdpowiedz na pytanie używając 5 zdań" + payload.input + "\n\nWiedząc że" + similar_chunk.text + "\n\n### Response:\n",
         "stop": [
             "\n",
             "###"
         ]
     }
-    llm_response = requests.post("http://192.168.0.3:9000/v1/completions/", json=llm_body)
-    llm_response = llm_response.json()['choices'][0]['text']
+    #llm_response = requests.post("", json=llm_body)
+    #llm_response = llm_response.json()['choices'][0]['text']
+
+    model_url = "http://192.168.0.3:9000/v1/completions/"
+
+   # llm = LlamaCPP(
+    #    model_url=model_url,
+     #   model_path=None,
+      #  temperature=0.1,
+       # max_new_tokens=256,
+        #context_window=3900,
+      #  generate_kwargs={},
+       # messages_to_prompt=messages_to_prompt,
+        #completion_to_prompt=completion_to_prompt,
+        #verbose=True,
+    #)
+    llm_response = "SDS" #llm.complete("Odpowiedz na pytanie używając 5 zdań" + payload.input + "\n\nWiedząc że" + similar_chunk.text)
     return HTTPStatus.OK, BotOut(output=str(llm_response))

api/bot/views.py

@@ -4,7 +4,7 @@
 from ninja import Router
 
 from bot.controllers import query_llm_controller
-from bot.schemas import BotIn, BotOut, BotError
+from bot.schemas import BotIn, BotOut
 
 router = Router(tags=["Bot"])