Semantic cache deploying

[alikhabazian]

I am trying to deploy Bifrost with the semantic cache plugin enabled, using a local LLM and embedding model. However, I have been unable to get the semantic cache working properly.

I tested multiple versions, and after bifrost:v1.4.3, the semantic cache plugin does not seem to be enabled at all. Even with version v1.4.3 and the configuration below, the semantic cache does not appear to trigger any embedding requests.

Here is my configuration:

{
  "$schema":"https://www.getbifrost.ai/schema",
  "client": {  
    "allowed_origins": [  
      "http://78.39.x.x:8080/",  
      "http://localhost:3000/"
    ],
    "drop_excess_requests": false,  
    "enable_logging": true,  
    "enable_governance": true
  },  
  "providers": {

    "tjhs-chat": {
      "keys": [
        {
          "name": "tjhs-key1",
          "value": "sk-vI38Axas3reTqF-zXTlCeQ",
          "models": [
            "Qwen/Qwen3-32B",
            "Qwen/Qwen3-Coder-30B-A3B-Instruct",
            "Qwen/Qwen3-VL-30B-A3B-Instruct-FP8",
            "Qwen3-32B"
          ],
          "weight": 1.0
        }
      ],
      "network_config": {
        "base_url": "https://oldapi.x.x/",
        "default_request_timeout_in_seconds": 120
      },
      "custom_provider_config": {
        "base_provider_type": "openai",
        "allowed_requests": {
          "chat_completion": true,
          "chat_completion_stream": true,
          "list_models": true
        }
      }
    },

    "tjhs-embeddings": {
      "keys": [
        {
          "name": "tjhs-key2",
          "value": "sk-vI38Axas3reTqF-zXTlCeQ",
          "models": [
            "Qwen/Qwen3-Embedding-8B",
            "Qwen3-Embedding-8B"
          ],
          "weight": 1.0
        }
      ],
      "network_config": {
        "base_url": "https://oldapi.x.x/",
        "default_request_timeout_in_seconds": 300
      },
      "custom_provider_config": {
        "base_provider_type": "openai",
        "allowed_requests": {
          "embedding": true,
          "list_models": true
        }
      }
    },

    "tjhs-reranker": {
      "keys": [
        {
          "name": "tjhs-key3",
          "value": "sk-ZEMwkOT1Dv9Jsix4z5BsFg",
          "models": [
            "jinaai/jina-reranker-v2-base-multilingual",
            "Qwen/Qwen3-Reranker-8B"
          ],
          "weight": 1.0
        }
      ],
      "network_config": {
        "base_url": "https://api.x.x/",
        "default_request_timeout_in_seconds": 60
      },
      "custom_provider_config": {
        "base_provider_type": "openai",
        "allowed_requests": {
          "embeddings": true,
          "list_models": true
        }
      }
    },

    "tjhs-audio": {
      "keys": [
        {
          "name": "tjhs-key4",
          "value": "sk-ZEMwkOT1Dv9Jsix4z5BsFg",
          "models": [
            "openai/whisper-large-v3",
            "openai/whisper-large-v3-turbo",
            "openai/whisper-custom"
          ],
          "weight": 1.0
        }
      ],
      "network_config": {
        "base_url": "https://api.x.x/",
        "default_request_timeout_in_seconds": 300
      },
      "custom_provider_config": {
        "base_provider_type": "openai",
        "allowed_requests": {
          "audio_transcription": true,
          "list_models": true
        }
      }
    },

    "tjhs-image": {
      "keys": [
        {
          "name": "tjhs-key5",
          "value": "sk-ZEMwkOT1Dv9Jsix4z5BsFg",
          "models": [
            "openai/sd35"
          ],
          "weight": 1.0
        }
      ],
      "network_config": {
        "base_url": "https://api.x.x/",
        "default_request_timeout_in_seconds": 180
      },
      "custom_provider_config": {
        "base_provider_type": "openai",
        "allowed_requests": {
          "image_generation": true,
          "list_models": true
        }
      }
    }

  },
  "config_store": {
    "enabled": true,
    "type": "sqlite",
    "config": {
      "path": "./config.db"
    }
  },


"vector_store": {
    "enabled": true,
    "type": "weaviate",
    "config": {
      "host": "localhost:9898",
      "scheme": "http"
    }
  },
"plugins": [
    {
      "enabled": true,
      "name": "semantic_cache",
      "config": {        
        "ttl": "5m",
        "threshold": 0.1
      }
    }
  ]


}

When I send the following request:

curl http://78.39.x.x:8080/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "x-bf-cache-key: session-1" \
-H "x-bf-cache-type: semantic" \
-d '{
"model": "tjhs-chat/Qwen3-32B",
"messages": [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Who is Hitler?"}
],
"temperature": 0.7
}'

docker compose

docker-compose.yaml 
services:  

  bifrost:  
    container_name:  bifrost_back
    image: maximhq/bifrost:v1.4.3
    environment:  
      - LOG_LEVEL=debug  
      - BIFROST_CONFIG=/app/data/config.json 
      - HTTP_PROXY=http://127.0.0.1:10809
      - HTTPS_PROXY=http://127.0.0.1:10809
      - NO_PROXY=localhost,127.0.0.1,http://api.x.x,https://api.x.x,https://oldapi.x.x,https://oldapi.x.x
    volumes:  
      - ./config.json:/app/data/config.json

Bifrost does not make any embedding calls, and the semantic cache does not seem to activate.
I've test even qdrant but that even sometimes remove collection itself