Patryk nie patrz

api/bot/controllers.py

@@ -12,12 +12,12 @@ def query_llm_controller(payload: BotIn) -> tuple[HTTPStatus, BotOut]:
     embeddings_body = {
         "input": payload.input
     }
-    response = requests.post("http://192.168.0.3:9000/v1/embeddings/", json=embeddings_body)
+    response = requests.post("http://llm-embedding-cpu:9000/v1/embeddings/", json=embeddings_body)
     input_embedding = response.json()['data'][0]['embedding']
     similar_chunks = Chunk.objects.order_by(L2Distance('embedding', input_embedding).desc())[:5]
     similar_chunks_text = "".join([chunk.text for chunk in similar_chunks])
     print(similar_chunks_text)
-    model_url = "http://192.168.0.3:9000/v1/"
+    model_url = "http://llm-embedding-cpu:9000/v1/"
     llm = OpenAI(temperature=0.5,
                  openai_api_key="XD",
                  openai_api_base=model_url

discord_bot/discord_bot/bot.py

@@ -44,7 +44,7 @@ async def sync(ctx) -> None:
 
 def query_llm(prompt):
     """Returns llm's response."""
-    url = "http://192.168.0.1:8000/api/bot/"
+    url = "http://api:8000/api/bot/"
     headers = {"Content-Type": "application/json"}
     data = {"input": prompt}
 

docker-compose.yml

@@ -6,17 +6,13 @@ services:
 
   db:
     profiles: [ "dev", "prod" ]
-    build:
-      context: ./postgres
-      dockerfile: postgres.Dockerfile
+    image: pgvector/pgvector:0.7.4-pg17
     volumes:
       - ./db:/var/lib/postgresql/data
       - ./postgres/vector_extension.sql:/docker-entrypoint-initdb.d/0-vector_extension.sql
     env_file:
       - .env
-    networks:
-      inner:
-        ipv4_address: 192.168.0.4
+
   api:
     profiles: [ "dev", "prod" ]
     build:
@@ -29,9 +25,6 @@ services:
       - "8000:8000"
     depends_on:
       - db
-    networks:
-      inner:
-        ipv4_address: 192.168.0.2
 
   llm-embedding-cpu:
     profiles: [ "cpu" ]
@@ -41,9 +34,6 @@ services:
       - ./llm/models:/models
     ports:
       - "9000:9000"
-    networks:
-      inner:
-        ipv4_address: 192.168.0.3
 
   llm-embedding-gpu:
     profiles: [ "gpu" ]
@@ -60,13 +50,4 @@ services:
             - driver: nvidia
               count: 1
               capabilities: [ gpu ]
-    networks:
-      inner:
-        ipv4_address: 192.168.0.3
 
-networks:
-  inner:
-    ipam:
-      driver: default
-      config:
-        - subnet: 192.168.0.0/30

postgres/postgres.Dockerfile → postgres/Dockerfile

@@ -11,4 +11,4 @@ RUN git clone https://github.com/pgvector/pgvector.git
 
 WORKDIR /tmp/pgvector
 RUN make
-RUN make install
+RUN make install