mirror of
https://github.com/we-promise/sure.git
synced 2026-09-02 05:11:05 +00:00
Juan José Mata
fd6f4ff078
* Add live AI checks to system health Give super admins a dedicated AI status view with bounded liveness probes for LLMs, vector stores, pgvector, and embedding endpoints. Record sanitized failures in both the system debug log and Rails logger, and document the recommended local configuration.\n\nCloses #3145 * Fix AI health CI checks * Address AI health review feedback * Correct Ollama model preload guidance * Distinguish OpenAI-compatible providers * Make Ollama startup readiness explicit * Recognize Cloudflare AI endpoints
415 lines
16 KiB
YAML
415 lines
16 KiB
YAML
# ===========================================================================
|
|
# Example Docker Compose file with Ollama (local LLM), OpenClaw (external AI
|
|
# assistant), and Pipelock (agent security proxy)
|
|
# ===========================================================================
|
|
#
|
|
# Purpose:
|
|
# --------
|
|
#
|
|
# This file extends the standard Sure setup with optional AI capabilities:
|
|
#
|
|
# Pipelock — agent security proxy
|
|
# (always runs)
|
|
# - Forward proxy (port 8888): applies destination, SSRF, rate, budget,
|
|
# CONNECT-header DLP, and receipt controls to HTTPS tunnels from clients
|
|
# that honor HTTPS_PROXY. HTTPS bodies remain encrypted because this
|
|
# example does not configure TLS interception. HTTPS_PROXY is cooperative;
|
|
# Docker Compose has no egress network policy.
|
|
# - MCP reverse proxy (port 8889): scans inbound AI traffic (DLP,
|
|
# prompt injection, tool poisoning, tool call policy). External AI
|
|
# clients should connect to Pipelock on port 8889 rather than
|
|
# directly to Sure's /mcp endpoint. Note: /mcp is still reachable
|
|
# on web port 3000 (auth token required); Pipelock adds scanning
|
|
# but Docker Compose cannot enforce network-level routing.
|
|
#
|
|
# Ollama + Open WebUI — local LLM inference
|
|
# (optional, --profile local-ai)
|
|
# - Only starts when you run: docker compose --profile local-ai up
|
|
#
|
|
# OpenClaw — external AI assistant
|
|
# (optional, --profile external-assistant)
|
|
# - Starts an OpenClaw locally for Sure's external assistant mode.
|
|
# - Set EXTERNAL_ASSISTANT_URL to http://openclaw:18789/v1/chat/completions
|
|
# for internal routing.
|
|
#
|
|
# Setup:
|
|
# ------
|
|
#
|
|
# 1. Copy pipelock.example.yaml alongside this file (or customize it).
|
|
# 2. Read the full setup guide:
|
|
#
|
|
# https://github.com/we-promise/sure/blob/main/docs/hosting/docker.md
|
|
#
|
|
# Troubleshooting:
|
|
# ----------------
|
|
#
|
|
# Problems? Swing by our Discord or open a Discussion here:
|
|
#
|
|
# https://github.com/we-promise/sure/discussions/categories/general
|
|
#
|
|
# IPv6 Note (Yahoo Finance Sync, others?):
|
|
# ----------------------------------------
|
|
# If you encounter "Failed to open TCP connection" errors to hosts like
|
|
# fc.yahoo.com in your environment, the issue may be that DNS resolves
|
|
# to an IPv6 address first. Since IPv6 may not be configured in the container,
|
|
# the connection hangs. To work around this, you can hardcode the IPv4
|
|
# addresses in your container's /etc/hosts or set custom DNS:
|
|
#
|
|
# extra_hosts:
|
|
# - "fc.yahoo.com:173.223.12.81"
|
|
# - "query1.finance.yahoo.com:98.136.48.169"
|
|
# - "query2.finance.yahoo.com:98.136.48.169"
|
|
#
|
|
# Or use explicit DNS servers that prefer IPv4 (already configured below).
|
|
#
|
|
|
|
x-db-env: &db_env
|
|
POSTGRES_USER: ${POSTGRES_USER:-sure_user}
|
|
POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-sure_password}
|
|
POSTGRES_DB: ${POSTGRES_DB:-sure_production}
|
|
|
|
x-rails-env: &rails_env
|
|
<<: *db_env
|
|
SECRET_KEY_BASE: ${SECRET_KEY_BASE:-a7523c3d0ae56415046ad8abae168d71074a79534a7062258f8d1d51ac2f76d3c3bc86d86b6b0b307df30d9a6a90a2066a3fa9e67c5e6f374dbd7dd4e0778e13}
|
|
SELF_HOSTED: "true"
|
|
RAILS_FORCE_SSL: "false"
|
|
RAILS_ASSUME_SSL: "false"
|
|
DB_HOST: db
|
|
DB_PORT: 5432
|
|
REDIS_URL: redis://redis:6379/1
|
|
# MCP server endpoint — enables /mcp for external AI assistants (e.g. Claude, GPT).
|
|
# Set both values to activate. MCP_USER_EMAIL must match an existing user's email.
|
|
# External AI clients should connect via Pipelock (port 8889) for scanning.
|
|
MCP_API_TOKEN: ${MCP_API_TOKEN:-}
|
|
MCP_USER_EMAIL: ${MCP_USER_EMAIL:-}
|
|
# Route outbound HTTPS through Pipelock for clients that respect HTTPS_PROXY.
|
|
# This applies tunnel-level controls; encrypted HTTPS bodies are not inspected
|
|
# unless you separately configure TLS interception and trust its CA.
|
|
HTTPS_PROXY: "http://pipelock:8888"
|
|
HTTP_PROXY: "http://pipelock:8888"
|
|
# Skip proxy for internal Docker network services (including ollama for local LLM calls)
|
|
NO_PROXY: "db,redis,pipelock,ollama,openclaw,localhost,127.0.0.1"
|
|
AI_DEBUG_MODE: "true" # Useful for debugging, set to "false" in production
|
|
# Ollama using OpenAI API compatible endpoints
|
|
OPENAI_ACCESS_TOKEN: token-can-be-any-value-for-ollama
|
|
OPENAI_MODEL: llama3.1:8b # Note: Use tool-enabled model
|
|
OPENAI_URI_BASE: http://ollama:11434/v1
|
|
# Local models are not streamed, so the chat shows nothing until the whole reply
|
|
# is generated, and tool-call rounds display nothing at all. OPENAI_REQUEST_TIMEOUT
|
|
# bounds each call to the model on its own; AI_RESPONSE_TIMEOUT covers the whole
|
|
# turn, so it has to be sized as a sum rather than simply set higher:
|
|
#
|
|
# (1 + ASSISTANT_MAX_TOOL_CALL_ITERATIONS) * OPENAI_REQUEST_TIMEOUT
|
|
# + tool execution + queue wait
|
|
#
|
|
# Here that is (1 + 2) * 300 = 900s of model time, plus 300s of headroom for tool
|
|
# execution and queue wait, giving 1200.
|
|
#
|
|
# The cap is lowered to 2 rather than left at 5 so the bound is three model calls
|
|
# instead of six — at 300s each, leaving it at 5 would need a timeout beyond half
|
|
# an hour. The trade is that longer tool chains fail earlier, with an explicit
|
|
# tool-call limit error rather than a timeout.
|
|
OPENAI_REQUEST_TIMEOUT: ${OPENAI_REQUEST_TIMEOUT:-300}
|
|
ASSISTANT_MAX_TOOL_CALL_ITERATIONS: ${ASSISTANT_MAX_TOOL_CALL_ITERATIONS:-2}
|
|
AI_RESPONSE_TIMEOUT: ${AI_RESPONSE_TIMEOUT:-1200}
|
|
# Vector store — pgvector keeps all data local (requires pgvector/pgvector Docker image for db)
|
|
VECTOR_STORE_PROVIDER: pgvector
|
|
EMBEDDING_URI_BASE: http://ollama:11434/v1
|
|
EMBEDDING_MODEL: mxbai-embed-large
|
|
EMBEDDING_DIMENSIONS: "1024"
|
|
# NOTE: enabling OpenAI will incur costs when you use AI-related features in the app (chat, rules). Make sure you have set appropriate spend limits on your account before adding this.
|
|
# OPENAI_ACCESS_TOKEN: ${OPENAI_ACCESS_TOKEN}
|
|
# External AI Assistant — delegates chat to a remote AI agent (e.g., OpenClaw).
|
|
# The agent calls back to Sure's /mcp endpoint for financial data.
|
|
# Set EXTERNAL_ASSISTANT_URL + TOKEN to activate, then either set ASSISTANT_TYPE=external
|
|
# here (forces all families) or choose "External" in Settings > Self-Hosting > AI Assistant.
|
|
ASSISTANT_TYPE: ${ASSISTANT_TYPE:-}
|
|
EXTERNAL_ASSISTANT_URL: ${EXTERNAL_ASSISTANT_URL:-http://openclaw:18789/v1/chat/completions}
|
|
EXTERNAL_ASSISTANT_TOKEN: ${EXTERNAL_ASSISTANT_TOKEN:-}
|
|
EXTERNAL_ASSISTANT_AGENT_ID: ${EXTERNAL_ASSISTANT_AGENT_ID:-main}
|
|
EXTERNAL_ASSISTANT_SESSION_KEY: ${EXTERNAL_ASSISTANT_SESSION_KEY:-agent:main:main}
|
|
EXTERNAL_ASSISTANT_ALLOWED_EMAILS: ${EXTERNAL_ASSISTANT_ALLOWED_EMAILS:-}
|
|
|
|
services:
|
|
pipelock:
|
|
image: ghcr.io/luckypipewrench/pipelock:3.4.0@sha256:2e6caa43967a5ef19cacbce00188d829c1b5de1eb0ad304a51d2085f32c5694c
|
|
user: "${PIPELOCK_UID:-1000}:${PIPELOCK_GID:-1000}"
|
|
container_name: pipelock
|
|
hostname: pipelock
|
|
restart: unless-stopped
|
|
volumes:
|
|
- ./pipelock.example.yaml:/etc/pipelock/pipelock.yaml:ro
|
|
# Optional signed receipts (flight recorder). Off until you create the
|
|
# dirs, generate a key, uncomment the two mounts below, and uncomment
|
|
# flight_recorder.dir + signing_key_path in pipelock.example.yaml.
|
|
# Create the dirs yourself first so they are owned by your host user
|
|
# (Docker would otherwise create missing bind-mount paths as root, which
|
|
# the uid-1000 Pipelock process cannot write to):
|
|
# mkdir -p ./pipelock-evidence ./pipelock-keys
|
|
# Generate the receipt-signing key via the image (no host binary needed;
|
|
# --out must be an absolute path inside the container):
|
|
# docker run --rm -v "$PWD/pipelock-keys:/keys" ghcr.io/luckypipewrench/pipelock:3.4.0@sha256:2e6caa43967a5ef19cacbce00188d829c1b5de1eb0ad304a51d2085f32c5694c \
|
|
# signing key generate --purpose receipt-signing --out /keys/flight-recorder-signing.key --id sure-compose
|
|
# The key is written 0600 owned by uid 1000 (Pipelock's default user). If
|
|
# your host user is not uid 1000, use this variant instead:
|
|
# export PIPELOCK_UID="$(id -u)" PIPELOCK_GID="$(id -g)"
|
|
# docker run --rm --user "$PIPELOCK_UID:$PIPELOCK_GID" -v "$PWD/pipelock-keys:/keys" ghcr.io/luckypipewrench/pipelock:3.4.0@sha256:2e6caa43967a5ef19cacbce00188d829c1b5de1eb0ad304a51d2085f32c5694c \
|
|
# signing key generate --purpose receipt-signing --out /keys/flight-recorder-signing.key --id sure-compose
|
|
# That makes the key and evidence directory readable/writable by the same
|
|
# uid/gid the compose service uses.
|
|
# - ./pipelock-evidence:/var/lib/pipelock/evidence
|
|
# - ./pipelock-keys:/etc/pipelock/keys:ro
|
|
command:
|
|
- "run"
|
|
- "--config"
|
|
- "/etc/pipelock/pipelock.yaml"
|
|
- "--listen"
|
|
- "0.0.0.0:8888"
|
|
- "--mcp-listen"
|
|
- "0.0.0.0:8889"
|
|
- "--mcp-upstream"
|
|
- "http://web:3000/mcp"
|
|
ports:
|
|
# MCP reverse proxy — external AI assistants connect here
|
|
- "${MCP_PROXY_PORT:-8889}:8889"
|
|
# Uncomment to expose forward proxy endpoints (/health, /metrics, /stats):
|
|
# - "8888:8888"
|
|
healthcheck:
|
|
test: ["CMD", "/pipelock", "healthcheck", "--addr", "127.0.0.1:8888"]
|
|
interval: 10s
|
|
timeout: 5s
|
|
retries: 3
|
|
start_period: 30s
|
|
networks:
|
|
- sure_net
|
|
|
|
# Ollama's OLLAMA_MODELS setting controls the model storage directory; it is
|
|
# not a preload list. Start Ollama, then pull each model once. The ollama
|
|
# volume below keeps the downloaded models across container restarts:
|
|
#
|
|
# docker compose -f compose.example.ai.yml --profile local-ai up -d --wait ollama
|
|
# docker compose -f compose.example.ai.yml exec ollama ollama pull deepseek-r1:8b
|
|
# docker compose -f compose.example.ai.yml exec ollama ollama pull llama3.1:8b
|
|
# docker compose -f compose.example.ai.yml exec ollama ollama pull mxbai-embed-large
|
|
ollama:
|
|
profiles:
|
|
- local-ai
|
|
volumes:
|
|
- ollama:/root/.ollama
|
|
container_name: ollama
|
|
hostname: ollama
|
|
restart: unless-stopped
|
|
image: docker.io/ollama/ollama:latest
|
|
ports:
|
|
- "11434:11434"
|
|
environment:
|
|
- OLLAMA_KEEP_ALIVE=1h
|
|
healthcheck:
|
|
test: ["CMD", "ollama", "list"]
|
|
interval: 5s
|
|
timeout: 5s
|
|
retries: 12
|
|
start_period: 5s
|
|
networks:
|
|
- sure_net
|
|
# Recommended: Enable GPU support
|
|
# deploy:
|
|
# resources:
|
|
# reservations:
|
|
# devices:
|
|
# - driver: nvidia
|
|
# count: all
|
|
# capabilities: [ gpu ]
|
|
|
|
ollama-webui:
|
|
profiles:
|
|
- local-ai
|
|
image: ghcr.io/open-webui/open-webui
|
|
container_name: ollama-webui
|
|
volumes:
|
|
- ollama-webui:/app/backend/data
|
|
depends_on:
|
|
- ollama
|
|
ports:
|
|
- "8080:8080"
|
|
environment: # https://docs.openwebui.com/getting-started/env-configuration#default_models
|
|
- OLLAMA_BASE_URLS=http://host.docker.internal:11434 #comma separated ollama hosts
|
|
- ENV=dev
|
|
- WEBUI_AUTH=False
|
|
- WEBUI_NAME=AI
|
|
- WEBUI_URL=http://localhost:8080
|
|
- WEBUI_SECRET_KEY=t0p-s3cr3t # pipelock:ignore
|
|
- NO_PROXY=host.docker.internal
|
|
extra_hosts:
|
|
- host.docker.internal:host-gateway
|
|
restart: unless-stopped
|
|
networks:
|
|
- sure_net
|
|
|
|
# OpenClaw gateway for Sure's "external assistant" mode.
|
|
# Based on OpenClaw Docker install docs:
|
|
# https://docs.openclaw.ai/install/docker
|
|
openclaw:
|
|
profiles:
|
|
- external-assistant
|
|
image: ${OPENCLAW_IMAGE:-ghcr.io/openclaw/openclaw:latest}
|
|
hostname: openclaw
|
|
restart: unless-stopped
|
|
init: true
|
|
environment:
|
|
HOME: /home/node
|
|
TERM: xterm-256color
|
|
TZ: ${OPENCLAW_TZ:-UTC}
|
|
OPENCLAW_GATEWAY_TOKEN: ${EXTERNAL_ASSISTANT_TOKEN:-changeme}
|
|
OPENCLAW_ALLOW_INSECURE_PRIVATE_WS: ${OPENCLAW_ALLOW_INSECURE_PRIVATE_WS:-}
|
|
command:
|
|
[
|
|
"node",
|
|
"dist/index.js",
|
|
"gateway",
|
|
# OpenClaw may exit with "Missing config" on first boot in example setups.
|
|
# `--allow-unconfigured` keeps the gateway running until you complete `openclaw setup`.
|
|
"pass",
|
|
"--allow-unconfigured",
|
|
"--bind",
|
|
"${OPENCLAW_GATEWAY_BIND:-lan}",
|
|
"--port",
|
|
"18789",
|
|
]
|
|
healthcheck:
|
|
test:
|
|
[
|
|
"CMD",
|
|
"node",
|
|
"-e",
|
|
"fetch('http://127.0.0.1:18789/healthz').then((r)=>process.exit(r.ok?0:1)).catch(()=>process.exit(1))",
|
|
]
|
|
interval: 30s
|
|
timeout: 5s
|
|
retries: 5
|
|
start_period: 20s
|
|
ports:
|
|
# Expose for local onboarding/debugging. Sure uses the internal service name.
|
|
- "${OPENCLAW_GATEWAY_PORT:-18789}:18789"
|
|
volumes:
|
|
- openclaw-config:/home/node/.openclaw
|
|
- openclaw-workspace:/home/node/.openclaw/workspace
|
|
networks:
|
|
- sure_net
|
|
|
|
web:
|
|
image: ghcr.io/we-promise/sure:stable
|
|
volumes:
|
|
- app-storage:/rails/storage
|
|
ports:
|
|
# Web UI for browser access. Note: /mcp is also reachable on this port,
|
|
# bypassing Pipelock's MCP scanning (auth token is still required).
|
|
# For hardened deployments, use `expose: [3000]` instead and front
|
|
# the web UI with a separate reverse proxy.
|
|
#
|
|
# To also publish on IPv6 (dual-stack), uncomment the line below AND
|
|
# set BINDING=:: in the environment block. See docs/hosting/docker.md
|
|
# "Binding to IPv6" for details.
|
|
# - "[::]:${PORT:-3000}:3000"
|
|
- ${PORT:-3000}:3000
|
|
restart: unless-stopped
|
|
environment:
|
|
<<: *rails_env
|
|
# BINDING: "::" # Uncomment for IPv6 dual-stack inside the container
|
|
depends_on:
|
|
db:
|
|
condition: service_healthy
|
|
redis:
|
|
condition: service_healthy
|
|
pipelock: # Remove this block and unset HTTPS_PROXY/HTTP_PROXY to run without Pipelock
|
|
condition: service_healthy
|
|
dns:
|
|
- 8.8.8.8
|
|
- 1.1.1.1
|
|
networks:
|
|
- sure_net
|
|
|
|
worker:
|
|
image: ghcr.io/we-promise/sure:stable
|
|
command: bundle exec sidekiq
|
|
volumes:
|
|
- app-storage:/rails/storage
|
|
restart: unless-stopped
|
|
depends_on:
|
|
db:
|
|
condition: service_healthy
|
|
redis:
|
|
condition: service_healthy
|
|
pipelock: # Remove this block and unset HTTPS_PROXY/HTTP_PROXY to run without Pipelock
|
|
condition: service_healthy
|
|
dns:
|
|
- 8.8.8.8
|
|
- 1.1.1.1
|
|
environment:
|
|
<<: *rails_env
|
|
networks:
|
|
- sure_net
|
|
|
|
db:
|
|
image: pgvector/pgvector:pg16-trixie
|
|
restart: unless-stopped
|
|
volumes:
|
|
- postgres-data:/var/lib/postgresql/data
|
|
environment:
|
|
<<: *db_env
|
|
healthcheck:
|
|
test: [ "CMD-SHELL", "pg_isready -U $$POSTGRES_USER -d $$POSTGRES_DB" ]
|
|
interval: 5s
|
|
timeout: 5s
|
|
retries: 5
|
|
networks:
|
|
- sure_net
|
|
|
|
backup:
|
|
profiles:
|
|
- backup
|
|
image: prodrigestivill/postgres-backup-local
|
|
restart: unless-stopped
|
|
volumes:
|
|
- /opt/sure-data/backups:/backups # Change this path to your desired backup location on the host machine
|
|
environment:
|
|
- POSTGRES_HOST=db
|
|
- POSTGRES_DB=${POSTGRES_DB:-sure_production}
|
|
- POSTGRES_USER=${POSTGRES_USER:-sure_user}
|
|
- POSTGRES_PASSWORD=${POSTGRES_PASSWORD:-sure_password} # pipelock:ignore
|
|
- SCHEDULE=@daily # Runs once a day at midnight
|
|
- BACKUP_KEEP_DAYS=7 # Keeps the last 7 days of backups
|
|
- BACKUP_KEEP_WEEKS=4 # Keeps 4 weekly backups
|
|
- BACKUP_KEEP_MONTHS=6 # Keeps 6 monthly backups
|
|
depends_on:
|
|
- db
|
|
networks:
|
|
- sure_net
|
|
|
|
redis:
|
|
image: redis:latest
|
|
restart: unless-stopped
|
|
volumes:
|
|
- redis-data:/data
|
|
healthcheck:
|
|
test: [ "CMD", "redis-cli", "ping" ]
|
|
interval: 5s
|
|
timeout: 5s
|
|
retries: 5
|
|
networks:
|
|
- sure_net
|
|
|
|
volumes:
|
|
app-storage:
|
|
postgres-data:
|
|
redis-data:
|
|
ollama:
|
|
ollama-webui:
|
|
openclaw-config:
|
|
openclaw-workspace:
|
|
|
|
networks:
|
|
sure_net:
|
|
driver: bridge
|
|
name: sure_net
|