palinode/docker-compose.yml at main · TLNing260310/palinode · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
# Palinode — one-command install: API + watcher + Ollama (+ bge-m3 pull).
#
#   docker compose up -d
#
# First run pulls the bge-m3 embedding model (≈1.2 GB) — saves made before the
# pull finishes are keyword-searchable immediately and embed once the model is
# up (the watcher retries). Check readiness with:
#
#   curl http://127.0.0.1:6340/status        # embed_functional: true when ready
#
# Your memory lives on the HOST at ${PALINODE_DATA_DIR:-$HOME/.palinode} —
# the bind mount below is load-bearing: files are the source of truth and must
# survive the containers. On Windows, set PALINODE_DATA_DIR explicitly
# (e.g. PALINODE_DATA_DIR=C:\Users\you\.palinode).
#
# Already running Ollama on the host? Skip the bundled one:
#
#   OLLAMA_URL=http://host.docker.internal:11434 \
#     docker compose up -d palinode-api palinode-watcher
#
# Containers run as root by default, so files in the bind mount are root-owned
# on Linux hosts; add `user: "1000:1000"` to both palinode services (with a
# pre-created, owned data dir) if that matters to you. macOS/Windows Docker
# Desktop maps ownership transparently.

services:
  palinode-api:
    build: .
    image: palinode:local
    restart: unless-stopped
    environment:
      PALINODE_DIR: /data
      OLLAMA_URL: ${OLLAMA_URL:-http://ollama:11434}
      EMBEDDING_MODEL: ${EMBEDDING_MODEL:-bge-m3}
    ports:
      # Loopback-only by default. Set PALINODE_BIND=0.0.0.0 to expose on the
      # network — and set PALINODE_API_TOKEN when you do. The container always
      # binds 0.0.0.0 internally; the app's non-loopback bind gate keys on
      # PALINODE_API_HOST (unset in the image) and cannot see this host-side
      # port mapping, so this line is the boundary and the token is on you.
      # See SECURITY.md#api-authentication for the contract.
      - "${PALINODE_BIND:-127.0.0.1}:${PALINODE_API_PORT:-6340}:6340"
    volumes:
      - ${PALINODE_DATA_DIR:-${HOME}/.palinode}:/data
    # Lets the host-Ollama override (OLLAMA_URL=http://host.docker.internal:11434)
    # work on Linux, where Docker doesn't provide host.docker.internal natively.
    extra_hosts:
      - "host.docker.internal:host-gateway"
    healthcheck:
      test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://127.0.0.1:6340/status', timeout=5)"]
      interval: 30s
      timeout: 10s
      start_period: 15s

  palinode-watcher:
    build: .
    image: palinode:local
    restart: unless-stopped
    command: ["python", "-m", "palinode.indexer.watcher"]
    environment:
      PALINODE_DIR: /data
      OLLAMA_URL: ${OLLAMA_URL:-http://ollama:11434}
      EMBEDDING_MODEL: ${EMBEDDING_MODEL:-bge-m3}
    volumes:
      - ${PALINODE_DATA_DIR:-${HOME}/.palinode}:/data
    extra_hosts:
      - "host.docker.internal:host-gateway"
    depends_on:
      - palinode-api

  ollama:
    image: ollama/ollama
    restart: unless-stopped
    volumes:
      - ollama-models:/root/.ollama
    healthcheck:
      test: ["CMD", "ollama", "list"]
      interval: 15s
      timeout: 10s
      start_period: 10s

  # One-shot: pulls the embedding model into the ollama service's volume,
  # then exits. Idempotent — an already-pulled model is a fast no-op.
  ollama-init:
    image: ollama/ollama
    restart: "no"
    environment:
      OLLAMA_HOST: http://ollama:11434
    entrypoint: ["ollama", "pull"]
    command: ["${EMBEDDING_MODEL:-bge-m3}"]
    depends_on:
      ollama:
        condition: service_healthy

volumes:
  ollama-models: