Need help with LiteLLM proxy server in Docker

[cvsickle]

The problem I'm having

I have a LiteLLM proxy running in docker, and I'm trying to add the Headroom integration. It's not working and I'm trying to track down why.

I created an image using this Dockerfile, and pushed it to our local registry. Our apps guy uses that image to spin up an instance on our Swarm server, and we connect to that instance via Open WebUI and other tools. You can see the Dockerfile, config, and compose below.

The problem is that the HeadroomCallback isn't doing anything, other than throwing some errors on the Open WebUI interface calls that say:

Request Failed
Message: type object 'HeadroomCallback' has no attribute 'async_post_call_success_hook'

The input tokens aren't being compressed, and the callback_duration_ms is always 0.

However, I can see via the LiteLLM API that the callback is available. You can see that in the JSON below.

I feel like I'm missing something. Any help or guidance would be greatly appreciated.

The setup I'm using

Dockerfile

FROM ghcr.io/berriai/litellm:v1.87.1

# Install headroom
RUN /app/.venv/bin/python -m ensurepip --upgrade \
    && /app/.venv/bin/python -m pip install ast-grep-cli \
    && /app/.venv/bin/python -m pip install headroom-ai==0.23.0 --no-deps \
    && ln -s /app/.venv/lib/python3.13/site-packages/headroom /app/headroom

# Copy LiteLLM config
COPY config.yaml /app/config.yaml

CMD ["--config", "/app/config.yaml", "--port", "4000", "--detailed_debug"]

config.yaml

general_settings:
  store_model_in_db: true
  user_header_mappings:
    - header_name: X-OpenWebUI-User-Id
      litellm_user_role: internal_user
    - header_name: X-OpenWebUI-User-Email
      litellm_user_role: customer
  default_team_disabled: true
  use_redis_transaction_buffer: true
  token_rate_limit_type: input
  store_prompts_in_spend_logs: true
  turn_off_message_logging: false
litellm_settings:
  drop_params: true
  modify_params: true
  max_end_user_budget_id: "max_end_user_budget"
  stream_timeout: 15.0
  turn_off_message_logging: true
  cache: true
  cache_params:
    type: redis
    supported_call_types: []
  callbacks:
    - headroom.integrations.litellm_callback.HeadroomCallback
router_settings:
  routing_strategy: simple-shuffle
  redis_host: redis
  redis_port: 6379
  redis_password: <redacted>

docker-compose.yaml

services:
  app:
    image: git.alliedmachine.com/library/litellm-headroom:1.87.1
    user: ${PUID}:${PGID}
    command: --config /app/config.yaml
    environment:
      - DATABASE_URL=postgresql://llmproxy:<redacted>@db:5432/litellm
      - REDIS_URL=redis://:<redacted>@redis:6379/0
      - POSTGRES_DB=litellm
      - POSTGRES_USER=llmproxy
      - POSTGRES_PASSWORD=<redacted>
      - REDIS_PASSWORD=<redacted>
      - LITELLM_MASTER_KEY=<redacted>
      - PROMETHEUS_ENABLED=true
      - STORE_MODEL_IN_DB=true
      - AUTO_REDIRECT_UI_LOGIN_TO_SSO=true
      - LITELLM_DONT_SHOW_FEEDBACK_BOX=true
      - GOOGLE_APPLICATION_CREDENTIALS=/app/credentials/open-webui.json
      - HEADROOM_MODE=local
    ports:
      - "4001:4000"
    networks:
      - network
    volumes:
      - ${CEPH_DEV_DOCKER_DIR}/litellm/data/litellm/config.yaml:/app/config.yaml:ro
      - ${CEPH_DEV_DOCKER_DIR}/litellm/data/litellm/credentials/open-webui.json:/app/credentials/open-webui.json:ro
    depends_on:
      - db
      - redis
    deploy:
      mode: replicated
      replicas: 1
    healthcheck:
      test:
        - CMD-SHELL
        - python3 -c "import urllib.request; urllib.request.urlopen('http://localhost:4000/health/liveliness')"
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 40s

  db:
    image: postgres:16
    user: ${PUID}:${PGID}
    environment:
      - POSTGRES_DB=litellm
      - POSTGRES_USER=llmproxy
      - POSTGRES_PASSWORD=<redacted>
    ports:
      - "5441:5432"
    networks:
      - network
    volumes:
      - ${CEPH_DEV_DOCKER_DIR}/litellm/data/postgres:/var/lib/postgresql/data
    deploy:
      mode: replicated
      replicas: 1
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U llmproxy -d litellm"]
      interval: 5s
      timeout: 5s
      retries: 20

  redis:
    image: redis:8.6.3-alpine
    user: ${PUID}:${PGID}
    command: sh -c "redis-server --appendonly yes --requirepass '<redacted>'"
    ports:
      - "6384:6379"
    networks:
      - network
    volumes:
      - ${CEPH_DEV_DOCKER_DIR}/litellm/data/redis:/data
    deploy:
      mode: replicated
      replicas: 1
    healthcheck:
      test: ["CMD-SHELL", "redis-cli -a \"<redacted>\" PING | grep PONG"]
      interval: 5s
      timeout: 5s
      retries: 20

networks:
  network:
    driver: overlay
    attachable: true

LiteLLM API - /active/callbacks/ endpoint

{
  "alerting": "None",
  "litellm.callbacks": [
    "<litellm.proxy.hooks.litellm_skills.main.SkillsInjectionHook object at 0x7d961922fcb0>",
    "<litellm.proxy.hooks.model_max_budget_limiter._PROXY_VirtualKeyModelMaxBudgetLimiter object at 0x7d960e2e4d70>",
    "<class 'headroom.integrations.litellm_callback.HeadroomCallback'>",
    "<litellm.proxy.hooks.proxy_track_cost_callback._ProxyDBLogger object at 0x7d960e8b96a0>",
    "<litellm.proxy.hooks.max_budget_limiter._PROXY_MaxBudgetLimiter object at 0x7d960ea374d0>",
    "<litellm.proxy.hooks.parallel_request_limiter_v3._PROXY_MaxParallelRequestsHandler_v3 object at 0x7d960e8b9940>",
    "<litellm.proxy.hooks.cache_control_check._PROXY_CacheControlCheck object at 0x7d960e8a4910>",
    "<litellm.proxy.hooks.responses_id_security.ResponsesIDSecurity object at 0x7d960e8b9be0>",
    "<litellm.proxy.hooks.max_iterations_limiter._PROXY_MaxIterationsHandler object at 0x7d960e8b9d30>",
    "<litellm.proxy.hooks.max_budget_per_session_limiter._PROXY_MaxBudgetPerSessionHandler object at 0x7d960e8b9e80>",
    "<litellm_enterprise.proxy.hooks.managed_files._PROXY_LiteLLMManagedFiles object at 0x7d960e8b9fd0>",
    "<litellm_enterprise.proxy.hooks.managed_vector_stores._PROXY_LiteLLMManagedVectorStores object at 0x7d960e8ba120>",
    "<litellm._service_logger.ServiceLogging object at 0x7d9608dfdcd0>",
    "headroom.integrations.litellm_callback.HeadroomCallback"
  ],
  "litellm.input_callback": [
    "cache",
    "<litellm.proxy.hooks.litellm_skills.main.SkillsInjectionHook object at 0x7d961922fcb0>",
    "<litellm.proxy.hooks.model_max_budget_limiter._PROXY_VirtualKeyModelMaxBudgetLimiter object at 0x7d960e2e4d70>",
    "<class 'headroom.integrations.litellm_callback.HeadroomCallback'>",
    "<litellm.proxy.hooks.proxy_track_cost_callback._ProxyDBLogger object at 0x7d960e8b96a0>",
    "<litellm.proxy.hooks.max_budget_limiter._PROXY_MaxBudgetLimiter object at 0x7d960ea374d0>",
    "<litellm.proxy.hooks.parallel_request_limiter_v3._PROXY_MaxParallelRequestsHandler_v3 object at 0x7d960e8b9940>",
    "<litellm.proxy.hooks.cache_control_check._PROXY_CacheControlCheck object at 0x7d960e8a4910>",
    "<litellm.proxy.hooks.responses_id_security.ResponsesIDSecurity object at 0x7d960e8b9be0>",
    "<litellm.proxy.hooks.max_iterations_limiter._PROXY_MaxIterationsHandler object at 0x7d960e8b9d30>",
    "<litellm.proxy.hooks.max_budget_per_session_limiter._PROXY_MaxBudgetPerSessionHandler object at 0x7d960e8b9e80>",
    "<litellm_enterprise.proxy.hooks.managed_files._PROXY_LiteLLMManagedFiles object at 0x7d960e8b9fd0>",
    "<litellm_enterprise.proxy.hooks.managed_vector_stores._PROXY_LiteLLMManagedVectorStores object at 0x7d960e8ba120>",
    "<litellm._service_logger.ServiceLogging object at 0x7d9608dfdcd0>"
  ],
  "litellm.failure_callback": [
    "<bound method Router.deployment_callback_on_failure of <litellm.router.Router object at 0x7d960ea2e510>>",
    "<litellm.proxy.hooks.litellm_skills.main.SkillsInjectionHook object at 0x7d961922fcb0>",
    "<litellm.proxy.hooks.model_max_budget_limiter._PROXY_VirtualKeyModelMaxBudgetLimiter object at 0x7d960e2e4d70>",
    "<class 'headroom.integrations.litellm_callback.HeadroomCallback'>",
    "<litellm.proxy.hooks.proxy_track_cost_callback._ProxyDBLogger object at 0x7d960e8b96a0>",
    "<litellm.proxy.hooks.max_budget_limiter._PROXY_MaxBudgetLimiter object at 0x7d960ea374d0>",
    "<litellm.proxy.hooks.parallel_request_limiter_v3._PROXY_MaxParallelRequestsHandler_v3 object at 0x7d960e8b9940>",
    "<litellm.proxy.hooks.cache_control_check._PROXY_CacheControlCheck object at 0x7d960e8a4910>",
    "<litellm.proxy.hooks.responses_id_security.ResponsesIDSecurity object at 0x7d960e8b9be0>",
    "<litellm.proxy.hooks.max_iterations_limiter._PROXY_MaxIterationsHandler object at 0x7d960e8b9d30>",
    "<litellm.proxy.hooks.max_budget_per_session_limiter._PROXY_MaxBudgetPerSessionHandler object at 0x7d960e8b9e80>",
    "<litellm_enterprise.proxy.hooks.managed_files._PROXY_LiteLLMManagedFiles object at 0x7d960e8b9fd0>",
    "<litellm_enterprise.proxy.hooks.managed_vector_stores._PROXY_LiteLLMManagedVectorStores object at 0x7d960e8ba120>",
    "<litellm._service_logger.ServiceLogging object at 0x7d9608dfdcd0>"
  ],
  "litellm.success_callback": [
    "cache",
    "<bound method Router.sync_deployment_callback_on_success of <litellm.router.Router object at 0x7d960ea2e510>>",
    "<litellm.proxy.hooks.litellm_skills.main.SkillsInjectionHook object at 0x7d961922fcb0>",
    "<litellm.proxy.hooks.model_max_budget_limiter._PROXY_VirtualKeyModelMaxBudgetLimiter object at 0x7d960e2e4d70>",
    "<class 'headroom.integrations.litellm_callback.HeadroomCallback'>",
    "<litellm.proxy.hooks.proxy_track_cost_callback._ProxyDBLogger object at 0x7d960e8b96a0>",
    "<litellm.proxy.hooks.max_budget_limiter._PROXY_MaxBudgetLimiter object at 0x7d960ea374d0>",
    "<litellm.proxy.hooks.parallel_request_limiter_v3._PROXY_MaxParallelRequestsHandler_v3 object at 0x7d960e8b9940>",
    "<litellm.proxy.hooks.cache_control_check._PROXY_CacheControlCheck object at 0x7d960e8a4910>",
    "<litellm.proxy.hooks.responses_id_security.ResponsesIDSecurity object at 0x7d960e8b9be0>",
    "<litellm.proxy.hooks.max_iterations_limiter._PROXY_MaxIterationsHandler object at 0x7d960e8b9d30>",
    "<litellm.proxy.hooks.max_budget_per_session_limiter._PROXY_MaxBudgetPerSessionHandler object at 0x7d960e8b9e80>",
    "<litellm_enterprise.proxy.hooks.managed_files._PROXY_LiteLLMManagedFiles object at 0x7d960e8b9fd0>",
    "<litellm_enterprise.proxy.hooks.managed_vector_stores._PROXY_LiteLLMManagedVectorStores object at 0x7d960e8ba120>",
    "<litellm._service_logger.ServiceLogging object at 0x7d9608dfdcd0>"
  ],
  "litellm._async_success_callback": [
    "cache",
    "<bound method Router.deployment_callback_on_success of <litellm.router.Router object at 0x7d960ea2e510>>",
    "<litellm.proxy.hooks.proxy_track_cost_callback._ProxyDBLogger object at 0x7d960ea35450>",
    "<litellm.proxy.hooks.litellm_skills.main.SkillsInjectionHook object at 0x7d961922fcb0>",
    "<litellm.proxy.hooks.model_max_budget_limiter._PROXY_VirtualKeyModelMaxBudgetLimiter object at 0x7d960e2e4d70>",
    "<class 'headroom.integrations.litellm_callback.HeadroomCallback'>",
    "<litellm.proxy.hooks.max_budget_limiter._PROXY_MaxBudgetLimiter object at 0x7d960ea374d0>",
    "<litellm.proxy.hooks.parallel_request_limiter_v3._PROXY_MaxParallelRequestsHandler_v3 object at 0x7d960e8b9940>",
    "<litellm.proxy.hooks.cache_control_check._PROXY_CacheControlCheck object at 0x7d960e8a4910>",
    "<litellm.proxy.hooks.responses_id_security.ResponsesIDSecurity object at 0x7d960e8b9be0>",
    "<litellm.proxy.hooks.max_iterations_limiter._PROXY_MaxIterationsHandler object at 0x7d960e8b9d30>",
    "<litellm.proxy.hooks.max_budget_per_session_limiter._PROXY_MaxBudgetPerSessionHandler object at 0x7d960e8b9e80>",
    "<litellm_enterprise.proxy.hooks.managed_files._PROXY_LiteLLMManagedFiles object at 0x7d960e8b9fd0>",
    "<litellm_enterprise.proxy.hooks.managed_vector_stores._PROXY_LiteLLMManagedVectorStores object at 0x7d960e8ba120>",
    "<litellm._service_logger.ServiceLogging object at 0x7d9608dfdcd0>"
  ],
  "litellm._async_failure_callback": [
    "<bound method Router.async_deployment_callback_on_failure of <litellm.router.Router object at 0x7d960ea2e510>>",
    "<litellm.proxy.hooks.litellm_skills.main.SkillsInjectionHook object at 0x7d961922fcb0>",
    "<litellm.proxy.hooks.model_max_budget_limiter._PROXY_VirtualKeyModelMaxBudgetLimiter object at 0x7d960e2e4d70>",
    "<class 'headroom.integrations.litellm_callback.HeadroomCallback'>",
    "<litellm.proxy.hooks.proxy_track_cost_callback._ProxyDBLogger object at 0x7d960e8b96a0>",
    "<litellm.proxy.hooks.max_budget_limiter._PROXY_MaxBudgetLimiter object at 0x7d960ea374d0>",
    "<litellm.proxy.hooks.parallel_request_limiter_v3._PROXY_MaxParallelRequestsHandler_v3 object at 0x7d960e8b9940>",
    "<litellm.proxy.hooks.cache_control_check._PROXY_CacheControlCheck object at 0x7d960e8a4910>",
    "<litellm.proxy.hooks.responses_id_security.ResponsesIDSecurity object at 0x7d960e8b9be0>",
    "<litellm.proxy.hooks.max_iterations_limiter._PROXY_MaxIterationsHandler object at 0x7d960e8b9d30>",
    "<litellm.proxy.hooks.max_budget_per_session_limiter._PROXY_MaxBudgetPerSessionHandler object at 0x7d960e8b9e80>",
    "<litellm_enterprise.proxy.hooks.managed_files._PROXY_LiteLLMManagedFiles object at 0x7d960e8b9fd0>",
    "<litellm_enterprise.proxy.hooks.managed_vector_stores._PROXY_LiteLLMManagedVectorStores object at 0x7d960e8ba120>",
    "<litellm._service_logger.ServiceLogging object at 0x7d9608dfdcd0>"
  ],
  "litellm._async_input_callback": [],
  "all_litellm_callbacks": [
    "<litellm.proxy.hooks.litellm_skills.main.SkillsInjectionHook object at 0x7d961922fcb0>",
    "<litellm.proxy.hooks.model_max_budget_limiter._PROXY_VirtualKeyModelMaxBudgetLimiter object at 0x7d960e2e4d70>",
    "<class 'headroom.integrations.litellm_callback.HeadroomCallback'>",
    "<litellm.proxy.hooks.proxy_track_cost_callback._ProxyDBLogger object at 0x7d960e8b96a0>",
    "<litellm.proxy.hooks.max_budget_limiter._PROXY_MaxBudgetLimiter object at 0x7d960ea374d0>",
    "<litellm.proxy.hooks.parallel_request_limiter_v3._PROXY_MaxParallelRequestsHandler_v3 object at 0x7d960e8b9940>",
    "<litellm.proxy.hooks.cache_control_check._PROXY_CacheControlCheck object at 0x7d960e8a4910>",
    "<litellm.proxy.hooks.responses_id_security.ResponsesIDSecurity object at 0x7d960e8b9be0>",
    "<litellm.proxy.hooks.max_iterations_limiter._PROXY_MaxIterationsHandler object at 0x7d960e8b9d30>",
    "<litellm.proxy.hooks.max_budget_per_session_limiter._PROXY_MaxBudgetPerSessionHandler object at 0x7d960e8b9e80>",
    "<litellm_enterprise.proxy.hooks.managed_files._PROXY_LiteLLMManagedFiles object at 0x7d960e8b9fd0>",
    "<litellm_enterprise.proxy.hooks.managed_vector_stores._PROXY_LiteLLMManagedVectorStores object at 0x7d960e8ba120>",
    "<litellm._service_logger.ServiceLogging object at 0x7d9608dfdcd0>",
    "headroom.integrations.litellm_callback.HeadroomCallback",
    "cache",
    "<litellm.proxy.hooks.litellm_skills.main.SkillsInjectionHook object at 0x7d961922fcb0>",
    "<litellm.proxy.hooks.model_max_budget_limiter._PROXY_VirtualKeyModelMaxBudgetLimiter object at 0x7d960e2e4d70>",
    "<class 'headroom.integrations.litellm_callback.HeadroomCallback'>",
    "<litellm.proxy.hooks.proxy_track_cost_callback._ProxyDBLogger object at 0x7d960e8b96a0>",
    "<litellm.proxy.hooks.max_budget_limiter._PROXY_MaxBudgetLimiter object at 0x7d960ea374d0>",
    "<litellm.proxy.hooks.parallel_request_limiter_v3._PROXY_MaxParallelRequestsHandler_v3 object at 0x7d960e8b9940>",
    "<litellm.proxy.hooks.cache_control_check._PROXY_CacheControlCheck object at 0x7d960e8a4910>",
    "<litellm.proxy.hooks.responses_id_security.ResponsesIDSecurity object at 0x7d960e8b9be0>",
    "<litellm.proxy.hooks.max_iterations_limiter._PROXY_MaxIterationsHandler object at 0x7d960e8b9d30>",
    "<litellm.proxy.hooks.max_budget_per_session_limiter._PROXY_MaxBudgetPerSessionHandler object at 0x7d960e8b9e80>",
    "<litellm_enterprise.proxy.hooks.managed_files._PROXY_LiteLLMManagedFiles object at 0x7d960e8b9fd0>",
    "<litellm_enterprise.proxy.hooks.managed_vector_stores._PROXY_LiteLLMManagedVectorStores object at 0x7d960e8ba120>",
    "<litellm._service_logger.ServiceLogging object at 0x7d9608dfdcd0>",
    "<bound method Router.deployment_callback_on_failure of <litellm.router.Router object at 0x7d960ea2e510>>",
    "<litellm.proxy.hooks.litellm_skills.main.SkillsInjectionHook object at 0x7d961922fcb0>",
    "<litellm.proxy.hooks.model_max_budget_limiter._PROXY_VirtualKeyModelMaxBudgetLimiter object at 0x7d960e2e4d70>",
    "<class 'headroom.integrations.litellm_callback.HeadroomCallback'>",
    "<litellm.proxy.hooks.proxy_track_cost_callback._ProxyDBLogger object at 0x7d960e8b96a0>",
    "<litellm.proxy.hooks.max_budget_limiter._PROXY_MaxBudgetLimiter object at 0x7d960ea374d0>",
    "<litellm.proxy.hooks.parallel_request_limiter_v3._PROXY_MaxParallelRequestsHandler_v3 object at 0x7d960e8b9940>",
    "<litellm.proxy.hooks.cache_control_check._PROXY_CacheControlCheck object at 0x7d960e8a4910>",
    "<litellm.proxy.hooks.responses_id_security.ResponsesIDSecurity object at 0x7d960e8b9be0>",
    "<litellm.proxy.hooks.max_iterations_limiter._PROXY_MaxIterationsHandler object at 0x7d960e8b9d30>",
    "<litellm.proxy.hooks.max_budget_per_session_limiter._PROXY_MaxBudgetPerSessionHandler object at 0x7d960e8b9e80>",
    "<litellm_enterprise.proxy.hooks.managed_files._PROXY_LiteLLMManagedFiles object at 0x7d960e8b9fd0>",
    "<litellm_enterprise.proxy.hooks.managed_vector_stores._PROXY_LiteLLMManagedVectorStores object at 0x7d960e8ba120>",
    "<litellm._service_logger.ServiceLogging object at 0x7d9608dfdcd0>",
    "cache",
    "<bound method Router.sync_deployment_callback_on_success of <litellm.router.Router object at 0x7d960ea2e510>>",
    "<litellm.proxy.hooks.litellm_skills.main.SkillsInjectionHook object at 0x7d961922fcb0>",
    "<litellm.proxy.hooks.model_max_budget_limiter._PROXY_VirtualKeyModelMaxBudgetLimiter object at 0x7d960e2e4d70>",
    "<class 'headroom.integrations.litellm_callback.HeadroomCallback'>",
    "<litellm.proxy.hooks.proxy_track_cost_callback._ProxyDBLogger object at 0x7d960e8b96a0>",
    "<litellm.proxy.hooks.max_budget_limiter._PROXY_MaxBudgetLimiter object at 0x7d960ea374d0>",
    "<litellm.proxy.hooks.parallel_request_limiter_v3._PROXY_MaxParallelRequestsHandler_v3 object at 0x7d960e8b9940>",
    "<litellm.proxy.hooks.cache_control_check._PROXY_CacheControlCheck object at 0x7d960e8a4910>",
    "<litellm.proxy.hooks.responses_id_security.ResponsesIDSecurity object at 0x7d960e8b9be0>",
    "<litellm.proxy.hooks.max_iterations_limiter._PROXY_MaxIterationsHandler object at 0x7d960e8b9d30>",
    "<litellm.proxy.hooks.max_budget_per_session_limiter._PROXY_MaxBudgetPerSessionHandler object at 0x7d960e8b9e80>",
    "<litellm_enterprise.proxy.hooks.managed_files._PROXY_LiteLLMManagedFiles object at 0x7d960e8b9fd0>",
    "<litellm_enterprise.proxy.hooks.managed_vector_stores._PROXY_LiteLLMManagedVectorStores object at 0x7d960e8ba120>",
    "<litellm._service_logger.ServiceLogging object at 0x7d9608dfdcd0>",
    "cache",
    "<bound method Router.deployment_callback_on_success of <litellm.router.Router object at 0x7d960ea2e510>>",
    "<litellm.proxy.hooks.proxy_track_cost_callback._ProxyDBLogger object at 0x7d960ea35450>",
    "<litellm.proxy.hooks.litellm_skills.main.SkillsInjectionHook object at 0x7d961922fcb0>",
    "<litellm.proxy.hooks.model_max_budget_limiter._PROXY_VirtualKeyModelMaxBudgetLimiter object at 0x7d960e2e4d70>",
    "<class 'headroom.integrations.litellm_callback.HeadroomCallback'>",
    "<litellm.proxy.hooks.max_budget_limiter._PROXY_MaxBudgetLimiter object at 0x7d960ea374d0>",
    "<litellm.proxy.hooks.parallel_request_limiter_v3._PROXY_MaxParallelRequestsHandler_v3 object at 0x7d960e8b9940>",
    "<litellm.proxy.hooks.cache_control_check._PROXY_CacheControlCheck object at 0x7d960e8a4910>",
    "<litellm.proxy.hooks.responses_id_security.ResponsesIDSecurity object at 0x7d960e8b9be0>",
    "<litellm.proxy.hooks.max_iterations_limiter._PROXY_MaxIterationsHandler object at 0x7d960e8b9d30>",
    "<litellm.proxy.hooks.max_budget_per_session_limiter._PROXY_MaxBudgetPerSessionHandler object at 0x7d960e8b9e80>",
    "<litellm_enterprise.proxy.hooks.managed_files._PROXY_LiteLLMManagedFiles object at 0x7d960e8b9fd0>",
    "<litellm_enterprise.proxy.hooks.managed_vector_stores._PROXY_LiteLLMManagedVectorStores object at 0x7d960e8ba120>",
    "<litellm._service_logger.ServiceLogging object at 0x7d9608dfdcd0>",
    "<bound method Router.async_deployment_callback_on_failure of <litellm.router.Router object at 0x7d960ea2e510>>",
    "<litellm.proxy.hooks.litellm_skills.main.SkillsInjectionHook object at 0x7d961922fcb0>",
    "<litellm.proxy.hooks.model_max_budget_limiter._PROXY_VirtualKeyModelMaxBudgetLimiter object at 0x7d960e2e4d70>",
    "<class 'headroom.integrations.litellm_callback.HeadroomCallback'>",
    "<litellm.proxy.hooks.proxy_track_cost_callback._ProxyDBLogger object at 0x7d960e8b96a0>",
    "<litellm.proxy.hooks.max_budget_limiter._PROXY_MaxBudgetLimiter object at 0x7d960ea374d0>",
    "<litellm.proxy.hooks.parallel_request_limiter_v3._PROXY_MaxParallelRequestsHandler_v3 object at 0x7d960e8b9940>",
    "<litellm.proxy.hooks.cache_control_check._PROXY_CacheControlCheck object at 0x7d960e8a4910>",
    "<litellm.proxy.hooks.responses_id_security.ResponsesIDSecurity object at 0x7d960e8b9be0>",
    "<litellm.proxy.hooks.max_iterations_limiter._PROXY_MaxIterationsHandler object at 0x7d960e8b9d30>",
    "<litellm.proxy.hooks.max_budget_per_session_limiter._PROXY_MaxBudgetPerSessionHandler object at 0x7d960e8b9e80>",
    "<litellm_enterprise.proxy.hooks.managed_files._PROXY_LiteLLMManagedFiles object at 0x7d960e8b9fd0>",
    "<litellm_enterprise.proxy.hooks.managed_vector_stores._PROXY_LiteLLMManagedVectorStores object at 0x7d960e8ba120>",
    "<litellm._service_logger.ServiceLogging object at 0x7d9608dfdcd0>"
  ],
  "num_callbacks": 86,
  "num_alerting": 0,
  "litellm.request_timeout": 6000
}

[yudin-s]

The error points to a LiteLLM callback API mismatch, not to the compression logic itself.

In the current Headroom callback code, HeadroomCallback implements the pre-call hook (async_pre_call_hook) and the older success/failure handlers (async_success_handler, async_failure_handler). The stack trace suggests the LiteLLM proxy is trying to call async_post_call_success_hook, which that installed class does not expose, so the callback is loaded but LiteLLM cannot finish the callback lifecycle.

I would isolate it this way inside the exact image you deployed:

/app/.venv/bin/python - <<'PY'
from headroom.integrations.litellm_callback import HeadroomCallback
cb = HeadroomCallback()
for name in [
    'async_pre_call_hook',
    'async_post_call_success_hook',
    'async_success_handler',
    'async_failure_handler',
]:
    print(name, hasattr(cb, name))
PY

If that prints False for async_post_call_success_hook, you have three practical paths:

1. Use a Headroom release that matches the LiteLLM callback API used by ghcr.io/berriai/litellm:v1.87.1.
2. Pin LiteLLM to a version whose proxy still calls the hook names implemented by headroom-ai==0.23.0.
3. As a temporary test only, subclass/wrap HeadroomCallback and add a no-op/delegating async_post_call_success_hook so LiteLLM stops failing after the provider call.

I would also avoid --no-deps until the hook mismatch is resolved. It is not the cause of this specific attribute error, but it can hide the next failure once the callback actually starts compressing.

If that introspection matches, this is probably worth marking answered as a version/API mismatch rather than a compression config issue.