LiteLLM Gateway ArchitectureΒΆ

Compose 🐳¢

flowchart LR
    subgraph role [svc-ai-litellm]
        svc_litellm["litellm<br>ghcr.io/berriai/litellm:main-v1.77.7-stable<br>http 4000"]
    end

Swarm 🐝¢

flowchart LR
    svc_litellm["litellm"]
    svc_litellm --> rep_litellm["replicas: from the host group"]
    classDef pinned fill:#fef7e0,stroke:#b06000,color:#3d2600;

VolumesΒΆ

flowchart LR
    none["svc-ai-litellm declares no volume"]

Condition

Effect

swarm, storage.backend: nfs, no nfs: false, role not pinned

shared over NFS

any one of those missing

node-local

All four conditions must hold, so pinning a role takes every one of its volumes node-local. Only the role’s primary entry counts as pinned here: a placement on any other service pins that service alone and leaves the backing of every volume unchanged. The chain is drawn rather than its result, because that coupling is what makes a placement change move data without saying so.

DependenciesΒΆ

flowchart LR
    subgraph deps [Dependencies]
        dep_svc_ai_lmstudio["svc-ai-lmstudio 🐳🐝"]
        dep_svc_ai_ollama["svc-ai-ollama 🐳🐝"]
        dep_svc_ai_s1["svc-ai-s1 🐳🐝"]
        dep_svc_db_postgres["svc-db-postgres 🐳🐝"]
        dep_web_app_docs["web-app-docs 🐳🐝"]
    end
    subgraph role [svc-ai-litellm 🐳🐝]
        svc_litellm["litellm"]
        svc_postgres["postgres"]
        svc_s1["s1"]
        svc_ollama["ollama"]
        svc_lmstudio["lmstudio"]
        svc_docs["docs"]
    end
    subgraph dependents [Dependents]
        dpt_dsk_gnt_claude["dsk-gnt-claude πŸ’»"]
        dpt_dsk_gnt_codex["dsk-gnt-codex πŸ’»"]
        dpt_dsk_gnt_pi["dsk-gnt-pi πŸ’»"]
        dpt_svc_ai_agent_broker["svc-ai-agent-broker 🐳🐝"]
        dpt_web_app_discourse["web-app-discourse 🐳🐝"]
        dpt_web_app_flowise["web-app-flowise 🐳🐝"]
        dpt_web_app_hermes["web-app-hermes 🐳🐝"]
        dpt_web_app_litellm["web-app-litellm πŸ’»"]
        dpt_web_app_matrix["web-app-matrix 🐳🐝"]
        dpt_web_app_mattermost["web-app-mattermost 🐳🐝"]
        dpt_web_app_mediawiki["web-app-mediawiki 🐳🐝"]
        dpt_web_app_moodle["web-app-moodle 🐳🐝"]
        dpt_web_app_n8n["web-app-n8n 🐳🐝"]
        dpt_web_app_nextcloud["web-app-nextcloud 🐳🐝"]
        dpt_web_app_openclaw["web-app-openclaw 🐳🐝"]
        dpt_web_app_openwebui["web-app-openwebui 🐳🐝"]
        dpt_web_app_wordpress["web-app-wordpress 🐳🐝"]
        dpt_web_app_xwiki["web-app-xwiki 🐳🐝"]
        dpt_web_app_zammad["web-app-zammad 🐳🐝"]
        dpt_web_svc_translate["web-svc-translate 🐳🐝"]
    end
    dep_svc_ai_lmstudio -. "0..1" .-> svc_lmstudio
    dep_svc_ai_ollama -. "0..1" .-> svc_ollama
    dep_svc_ai_s1 -. "0..1" .-> svc_s1
    dep_svc_db_postgres -. "0..1" .-> svc_postgres
    dep_web_app_docs -. "0..1" .-> svc_docs
    svc_litellm -. "0..1" .-> dpt_dsk_gnt_claude
    svc_litellm -. "0..1" .-> dpt_dsk_gnt_codex
    svc_litellm -. "0..1" .-> dpt_dsk_gnt_pi
    svc_litellm -- "1:1" --> dpt_svc_ai_agent_broker
    svc_litellm -. "0..1" .-> dpt_web_app_discourse
    svc_litellm -. "0..1" .-> dpt_web_app_flowise
    svc_litellm -. "0..1" .-> dpt_web_app_hermes
    svc_litellm -- "1:1" --> dpt_web_app_litellm
    svc_litellm -. "0..1" .-> dpt_web_app_matrix
    svc_litellm -. "0..1" .-> dpt_web_app_mattermost
    svc_litellm -. "0..1" .-> dpt_web_app_mediawiki
    svc_litellm -. "0..1" .-> dpt_web_app_moodle
    svc_litellm -. "0..1" .-> dpt_web_app_n8n
    svc_litellm -. "0..1" .-> dpt_web_app_nextcloud
    svc_litellm -. "0..1" .-> dpt_web_app_openclaw
    svc_litellm -. "0..1" .-> dpt_web_app_openwebui
    svc_litellm -. "0..1" .-> dpt_web_app_wordpress
    svc_litellm -. "0..1" .-> dpt_web_app_xwiki
    svc_litellm -. "0..1" .-> dpt_web_app_zammad
    svc_litellm -. "0..1" .-> dpt_web_svc_translate

Symbol

Meaning

1:1 solid

fixed relationship

0..1 dashed

conditional, enabled only in matching deployments

0..0 red

declared but turned off in this role

πŸ’»

deploys to the host

🐳

deploys under compose

🐝

deploys under swarm

❌

service explicitly turned off

βš™οΈ

Ansible role dependency from meta/main.yml

Every dependency and dependent is listed; the graph is not truncated.