Compare commits
2 Commits
feat/herme
...
1ec22ff3fc
| Author | SHA1 | Date | |
|---|---|---|---|
| 1ec22ff3fc | |||
| b1dbdb9f2d |
31
AGENTS.md
Normal file
31
AGENTS.md
Normal file
@@ -0,0 +1,31 @@
|
||||
# AGENTS.md
|
||||
|
||||
Development conventions for the compose repository (Docker Compose stacks).
|
||||
|
||||
## Build & Deploy
|
||||
|
||||
- The ai stack is managed via `systemctl restart ai_stack.service` on the NixOS host
|
||||
- Compose files are built from the git repo; apply via `nh os switch` or `systemctl restart ai_stack.service`
|
||||
- Never `docker compose up -d` directly — it bypasses the systemd env file and breaks secrets
|
||||
|
||||
## Hermes Workers
|
||||
|
||||
- Paperclip Hermes workers are added via `ai/scripts/provision-hermes-worker.sh`
|
||||
- The script appends only — never deletes or modifies existing content
|
||||
- Workers are CPU-only containers on the `ai_backend` network with no GPU passthrough
|
||||
|
||||
## Workflow
|
||||
|
||||
- New feature → clean branch from `origin/master` → push → PR on Gitea
|
||||
- Branch naming: `feat/description` (features), `fix/description` (bugs/docs)
|
||||
- Always branch from `origin/master`, never from another feature branch
|
||||
- Submodule changes (when this repo is consumed as a submodule): commit the submodule update in the parent repo
|
||||
- PR title should describe the change; body should explain motivation + summary
|
||||
- After PR merge, delete the feature branch
|
||||
|
||||
## YAML Conventions
|
||||
|
||||
- Use `x-*` extension fields for reusable anchors
|
||||
- Comment out inactive services rather than deleting them
|
||||
- Environment variables in dictionary format (`KEY: value`) for anchor compatibility
|
||||
- List format (`- KEY=value`) works for standard services but doesn't merge with anchors
|
||||
@@ -1,4 +1,35 @@
|
||||
version: "3.8"
|
||||
|
||||
# ── Hermes Worker Template ──────────────────────────────────
|
||||
# Used by paperclip-worker-* Hermes containers via YAML anchor.
|
||||
# Each worker = one isolated Hermes agent for a Paperclip employee.
|
||||
# Override at service level: container_name, API_SERVER_PORT,
|
||||
# API_SERVER_KEY, DISCORD_BOT_TOKEN, volumes.
|
||||
# Workers have NO GPU — they use OpenCode Go or remote providers.
|
||||
x-hermes-worker: &hermes-worker
|
||||
build:
|
||||
context: ./hermes
|
||||
ssh:
|
||||
- default
|
||||
entrypoint: ["/bin/bash", "-c",
|
||||
"bash /opt/data/hermes-tools/install.sh && exec /usr/bin/tini -g -- /opt/hermes/docker/entrypoint.sh \"$@\"",
|
||||
"hermes-entrypoint"]
|
||||
command: gateway run
|
||||
restart: always
|
||||
environment:
|
||||
API_SERVER_ENABLED: "true"
|
||||
API_SERVER_HOST: "0.0.0.0"
|
||||
OLLAMA_HOST: "http://ollama:11434"
|
||||
OPENROUTER_API_KEY: ${OPENROUTER_API_KEY}
|
||||
# Each worker needs its own OpenCode Go API key in .env
|
||||
OPENCODE_GO_API_KEY: ${OPENCODE_GO_API_KEY}
|
||||
GATEWAY_ALLOW_ALL_USERS: "true"
|
||||
TZ: "America/Montreal"
|
||||
networks:
|
||||
ai_backend:
|
||||
# NO devices — workers are CPU-only, no GPU passthrough
|
||||
# ─────────────────────────────────────────────────────────────
|
||||
|
||||
services:
|
||||
|
||||
# webui:
|
||||
@@ -54,10 +85,6 @@ services:
|
||||
- TZ=America/Montreal
|
||||
volumes:
|
||||
- /mnt/HoardingCow_docker_data/Hermes/data:/opt/data
|
||||
# Syncthing-shared org files — read-only view of user's agenda
|
||||
- /mnt/HoardingCow_docker_data/Syncthing/telos-ro:/opt/data/telos-ro:ro
|
||||
# Syncthing-shared inbox — write tasks here, they sync to user's laptop
|
||||
- /mnt/HoardingCow_docker_data/Syncthing/telos-rw:/opt/data/telos-rw:rw
|
||||
devices:
|
||||
- /dev/kfd:/dev/kfd
|
||||
- /dev/dri:/dev/dri
|
||||
@@ -67,35 +94,6 @@ services:
|
||||
networks:
|
||||
- ai_backend
|
||||
|
||||
syncthing:
|
||||
image: syncthing/syncthing:latest
|
||||
container_name: syncthing
|
||||
hostname: syncthing
|
||||
restart: always
|
||||
ports:
|
||||
- "8384:8384"
|
||||
- "22000:22000"
|
||||
- "21027:21027/udp"
|
||||
environment:
|
||||
- TZ=America/Montreal
|
||||
volumes:
|
||||
- /mnt/HoardingCow_docker_data/Syncthing/config:/var/syncthing/config
|
||||
- /mnt/HoardingCow_docker_data/Syncthing/telos-ro:/telos-ro
|
||||
- /mnt/HoardingCow_docker_data/Syncthing/telos-rw:/telos-rw
|
||||
networks:
|
||||
- ai_backend
|
||||
- ai_net
|
||||
labels:
|
||||
- "traefik.enable=true"
|
||||
- "traefik.http.routers.syncthing-http.rule=Host(`syncthing.lazyworkhorse.net`)"
|
||||
- "traefik.http.routers.syncthing-http.entrypoints=web"
|
||||
- "traefik.http.routers.syncthing-http.middlewares=redirect-to-https"
|
||||
- "traefik.http.routers.syncthing-https.rule=Host(`syncthing.lazyworkhorse.net`)"
|
||||
- "traefik.http.routers.syncthing-https.entrypoints=websecure"
|
||||
- "traefik.http.routers.syncthing-https.tls=true"
|
||||
- "traefik.http.routers.syncthing-https.tls.certresolver=njalla"
|
||||
- "traefik.http.services.syncthing.loadbalancer.server.port=8384"
|
||||
|
||||
ollama:
|
||||
build:
|
||||
context: ./ollama
|
||||
@@ -129,6 +127,24 @@ services:
|
||||
- "303"
|
||||
- "26"
|
||||
|
||||
# ── Paperclip Worker Hermes Agents ──────────────────────────
|
||||
# Each worker is an isolated Hermes agent for a Paperclip employee.
|
||||
# Add new workers with: ./scripts/provision-hermes-worker.sh <name> <discord-token>
|
||||
# The API server key and port are generated automatically.
|
||||
# Workers are CPU-only — they use OpenCode Go or remote providers.
|
||||
|
||||
# ── Worker Template (commented — uncomment + configure to activate) ──
|
||||
# hermes-worker-1:
|
||||
# <<: *hermes-worker
|
||||
# container_name: hermes-worker-1
|
||||
# environment:
|
||||
# API_SERVER_PORT: "8651"
|
||||
# API_SERVER_KEY: "generated-by-provision-script"
|
||||
# DISCORD_BOT_TOKEN: ${WORKER_1_DISCORD_BOT_TOKEN}
|
||||
# volumes:
|
||||
# - /mnt/HoardingCow_docker_data/Hermes/worker-1:/opt/data
|
||||
# ─────────────────────────────────────────────────────────────
|
||||
|
||||
networks:
|
||||
ai_net:
|
||||
external: true
|
||||
|
||||
@@ -1,154 +0,0 @@
|
||||
#!/bin/bash
|
||||
# ── Hermes Workspace Combined Entrypoint ──
|
||||
# Waits for the Hermes gateway container (hermes:8642) to become healthy,
|
||||
# then starts the Hermes Workspace web UI in the foreground.
|
||||
# Supports graceful shutdown via SIGTERM/SIGINT.
|
||||
# ──────────────────────────────────────────
|
||||
|
||||
set -euo pipefail
|
||||
|
||||
# ── Configuration ──────────────────────────────────────────────
|
||||
GATEWAY_HOST="${GATEWAY_HOST:-hermes}"
|
||||
GATEWAY_PORT="${GATEWAY_PORT:-8642}"
|
||||
GATEWAY_URL="http://${GATEWAY_HOST}:${GATEWAY_PORT}"
|
||||
|
||||
HEALTH_ENDPOINT="${HEALTH_ENDPOINT:-/health}"
|
||||
MAX_RETRIES="${HEALTH_MAX_RETRIES:-60}"
|
||||
RETRY_INTERVAL="${HEALTH_RETRY_INTERVAL:-2}"
|
||||
|
||||
WORKSPACE_DIR="${WORKSPACE_DIR:-/workspace}"
|
||||
WORKSPACE_ENTRY="${WORKSPACE_ENTRY:-server-entry.js}"
|
||||
|
||||
PID_FILE="${PID_FILE:-/tmp/workspace.pid}"
|
||||
|
||||
# ── Logging ────────────────────────────────────────────────────
|
||||
log_info() { echo "[$(date '+%Y-%m-%d %H:%M:%S')] [INFO] $*"; }
|
||||
log_warn() { echo "[$(date '+%Y-%m-%d %H:%M:%S')] [WARN] $*"; }
|
||||
log_error() { echo "[$(date '+%Y-%m-%d %H:%M:%S')] [ERROR] $*"; }
|
||||
|
||||
# ── Graceful Shutdown ──────────────────────────────────────────
|
||||
_workspace_pid=""
|
||||
_shutting_down=false
|
||||
|
||||
cleanup() {
|
||||
if [ "$_shutting_down" = true ]; then
|
||||
return
|
||||
fi
|
||||
_shutting_down=true
|
||||
|
||||
log_info "Shutdown signal received, cleaning up..."
|
||||
|
||||
# Stop workspace process if running
|
||||
if [ -n "$_workspace_pid" ] && kill -0 "$_workspace_pid" 2>/dev/null; then
|
||||
log_info "Stopping workspace (PID: $_workspace_pid)..."
|
||||
kill -TERM "$_workspace_pid" 2>/dev/null || true
|
||||
|
||||
# Give it time to shut down gracefully
|
||||
local wait_sec=10
|
||||
while kill -0 "$_workspace_pid" 2>/dev/null && [ "$wait_sec" -gt 0 ]; do
|
||||
sleep 1
|
||||
wait_sec=$((wait_sec - 1))
|
||||
done
|
||||
|
||||
# Force kill if still running
|
||||
if kill -0 "$_workspace_pid" 2>/dev/null; then
|
||||
log_warn "Workspace did not shut down gracefully, force killing..."
|
||||
kill -KILL "$_workspace_pid" 2>/dev/null || true
|
||||
fi
|
||||
fi
|
||||
|
||||
# Clean up PID file
|
||||
[ -f "$PID_FILE" ] && rm -f "$PID_FILE"
|
||||
|
||||
log_info "Shutdown complete."
|
||||
exit 0
|
||||
}
|
||||
|
||||
# Trap termination signals for graceful shutdown
|
||||
trap cleanup SIGTERM SIGINT
|
||||
|
||||
# ── Gateway Health Check ───────────────────────────────────────
|
||||
wait_for_gateway() {
|
||||
local url="${GATEWAY_URL}${HEALTH_ENDPOINT}"
|
||||
local retries="$MAX_RETRIES"
|
||||
local interval="$RETRY_INTERVAL"
|
||||
local attempt=0
|
||||
|
||||
log_info "Waiting for Hermes gateway at ${GATEWAY_URL}..."
|
||||
log_info "Max retries: ${retries}, interval: ${interval}s"
|
||||
|
||||
while [ "$attempt" -lt "$retries" ]; do
|
||||
attempt=$((attempt + 1))
|
||||
|
||||
if curl -fsS "${url}" >/dev/null 2>&1; then
|
||||
log_info "Gateway is healthy after ${attempt} attempt(s) (${GATEWAY_URL})"
|
||||
return 0
|
||||
fi
|
||||
|
||||
if [ "$attempt" -lt "$retries" ]; then
|
||||
log_info "Gateway not ready yet (attempt ${attempt}/${retries}), retrying in ${interval}s..."
|
||||
sleep "$interval"
|
||||
fi
|
||||
done
|
||||
|
||||
log_error "Gateway did not become healthy after ${retries} attempts ($((retries * interval))s)"
|
||||
return 1
|
||||
}
|
||||
|
||||
# ── Workspace Startup ──────────────────────────────────────────
|
||||
start_workspace() {
|
||||
local entry="${WORKSPACE_DIR}/${WORKSPACE_ENTRY}"
|
||||
|
||||
if [ ! -d "$WORKSPACE_DIR" ]; then
|
||||
log_error "Workspace directory not found: ${WORKSPACE_DIR}"
|
||||
return 1
|
||||
fi
|
||||
|
||||
if [ ! -f "$entry" ]; then
|
||||
log_error "Workspace entry point not found: ${entry}"
|
||||
return 1
|
||||
fi
|
||||
|
||||
log_info "Starting Hermes Workspace web UI..."
|
||||
log_info " Directory: ${WORKSPACE_DIR}"
|
||||
log_info " Entry: ${entry}"
|
||||
|
||||
cd "$WORKSPACE_DIR"
|
||||
|
||||
# Start workspace in background so we can trap signals
|
||||
node --max-old-space-size=2048 "${entry}" &
|
||||
_workspace_pid=$!
|
||||
echo "$_workspace_pid" > "$PID_FILE"
|
||||
|
||||
log_info "Workspace started (PID: ${_workspace_pid})"
|
||||
|
||||
# Wait for workspace process
|
||||
wait "$_workspace_pid"
|
||||
local exit_code=$?
|
||||
|
||||
log_info "Workspace exited with code ${exit_code}"
|
||||
return "$exit_code"
|
||||
}
|
||||
|
||||
# ── Main ───────────────────────────────────────────────────────
|
||||
main() {
|
||||
log_info "=== Hermes Workspace Combined Entrypoint ==="
|
||||
log_info "Gateway: ${GATEWAY_URL}"
|
||||
log_info "Workspace: ${WORKSPACE_DIR}/${WORKSPACE_ENTRY}"
|
||||
log_info "PID file: ${PID_FILE}"
|
||||
|
||||
# Wait for gateway to be healthy
|
||||
if ! wait_for_gateway; then
|
||||
log_warn "Proceeding without confirmed gateway health..."
|
||||
fi
|
||||
|
||||
# Start the workspace
|
||||
start_workspace
|
||||
local exit_code=$?
|
||||
|
||||
log_info "Entrypoint exiting with code ${exit_code}"
|
||||
return "$exit_code"
|
||||
}
|
||||
|
||||
# Run main; exit with its return code
|
||||
main "$@"
|
||||
135
ai/scripts/provision-hermes-worker.sh
Executable file
135
ai/scripts/provision-hermes-worker.sh
Executable file
@@ -0,0 +1,135 @@
|
||||
#!/usr/bin/env bash
|
||||
set -euo pipefail
|
||||
|
||||
# ── Hermes Worker Provisioner ──────────────────────────────
|
||||
# Adds a new Paperclip Hermes worker to the ai compose stack.
|
||||
#
|
||||
# Usage:
|
||||
# ./provision-hermes-worker.sh <name> <discord_bot_token_var>
|
||||
#
|
||||
# Example:
|
||||
# ./provision-hermes-worker.sh worker-1 WORKER_1_DISCORD_BOT_TOKEN
|
||||
#
|
||||
# The script APPENDS only — never modifies or removes existing
|
||||
# content, even commented lines.
|
||||
#
|
||||
# Post-provision steps (manual):
|
||||
# 1. Add secrets to agenix .env file
|
||||
# 2. systemctl restart ai_stack.service
|
||||
# 3. Configure Paperclip agent
|
||||
# ─────────────────────────────────────────────────────────────
|
||||
|
||||
NAME="${1:?Usage: $0 <name> <discord_bot_token_var>}"
|
||||
TOKEN_VAR="${2:?Usage: $0 <name> <discord_bot_token_var>}"
|
||||
|
||||
# ── Paths ───────────────────────────────────────────────────
|
||||
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
|
||||
COMPOSE_DIR="$(cd "$SCRIPT_DIR/.." && pwd)"
|
||||
COMPOSE_FILE="${COMPOSE_DIR}/compose.yml"
|
||||
|
||||
# Each Hermes worker gets its own volume on the NFS HoardingCow
|
||||
VOLUME_BASE="/mnt/HoardingCow_docker_data/Hermes"
|
||||
VOLUME_DIR="${VOLUME_BASE}/${NAME}"
|
||||
|
||||
# The Hermes container runs as UID 10000 (hermes user from Dockerfile)
|
||||
HERMES_UID=10000
|
||||
|
||||
# ── Validation ──────────────────────────────────────────────
|
||||
if ! [ -f "$COMPOSE_FILE" ]; then
|
||||
echo "❌ compose.yml not found at $COMPOSE_FILE"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
if grep -q "^ ${NAME}:" "$COMPOSE_FILE"; then
|
||||
echo "❌ Service '${NAME}' already exists in ${COMPOSE_FILE}"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
# ── Generate unique API key ─────────────────────────────────
|
||||
# Used by Paperclip to authenticate against this worker's
|
||||
# Hermes API server (/v1/chat/completions)
|
||||
API_KEY="pc_worker_$(openssl rand -hex 16)"
|
||||
|
||||
# ── Find next available API port ────────────────────────────
|
||||
# Workers get sequential ports starting at 8650.
|
||||
# Scans compose.yml for existing API_SERVER_PORT values and
|
||||
# picks the next one.
|
||||
BASE_PORT=8650
|
||||
MAX_PORT=0
|
||||
while IFS= read -r line; do
|
||||
port="${line#*API_SERVER_PORT: \"}"
|
||||
port="${port%%\"*}"
|
||||
if [ -n "$port" ] && [ "$port" -gt "$MAX_PORT" ]; then
|
||||
MAX_PORT="$port"
|
||||
fi
|
||||
done < <(grep -oP 'API_SERVER_PORT:\s*"\d+"' "$COMPOSE_FILE" 2>/dev/null)
|
||||
|
||||
NEW_PORT=$((MAX_PORT + 1))
|
||||
if [ "$NEW_PORT" -lt "$BASE_PORT" ]; then
|
||||
NEW_PORT=$BASE_PORT
|
||||
fi
|
||||
|
||||
# ── Create volume directory (on NFS) ────────────────────────
|
||||
echo "📁 Creating volume directory: ${VOLUME_DIR}"
|
||||
mkdir -p "$VOLUME_DIR"
|
||||
|
||||
# Hermes container runs as UID 10000 — set ownership so the
|
||||
# container can write its config, sessions, skills
|
||||
if command -v chown &>/dev/null; then
|
||||
chown -R "${HERMES_UID}:${HERMES_UID}" "$VOLUME_DIR" 2>/dev/null || \
|
||||
echo "⚠ Could not chown ${VOLUME_DIR} — run with sudo if needed"
|
||||
fi
|
||||
|
||||
# Make it group-readable for debugging
|
||||
chmod 755 "$VOLUME_DIR" 2>/dev/null || true
|
||||
|
||||
# ── Append service to compose.yml ───────────────────────────
|
||||
echo "📝 Appending service '${NAME}' to compose.yml ..."
|
||||
|
||||
TMPFILE=$(mktemp)
|
||||
|
||||
awk -v name="$NAME" \
|
||||
-v port="$NEW_PORT" \
|
||||
-v api_key="$API_KEY" \
|
||||
-v token_var="$TOKEN_VAR" \
|
||||
'
|
||||
# Insert new worker service block just before the networks: section
|
||||
/^networks:/ {
|
||||
print ""
|
||||
print " " name ":"
|
||||
print " <<: *hermes-worker"
|
||||
print " container_name: " name
|
||||
print " environment:"
|
||||
print " API_SERVER_PORT: \"" port "\""
|
||||
print " API_SERVER_KEY: \"" api_key "\""
|
||||
print " DISCORD_BOT_TOKEN: ${" token_var "}"
|
||||
print " volumes:"
|
||||
print " - /mnt/HoardingCow_docker_data/Hermes/" name ":/opt/data"
|
||||
print ""
|
||||
}
|
||||
{ print }
|
||||
' "$COMPOSE_FILE" > "$TMPFILE" && mv "$TMPFILE" "$COMPOSE_FILE"
|
||||
|
||||
# ── Done ────────────────────────────────────────────────────
|
||||
echo ""
|
||||
echo "✅ Worker '${NAME}' provisioned successfully"
|
||||
echo ""
|
||||
echo "────────────────────────────────────────────"
|
||||
echo " NEXT STEPS"
|
||||
echo "────────────────────────────────────────────"
|
||||
echo ""
|
||||
echo "1. Add secrets to the agenix .env stack file:"
|
||||
echo ""
|
||||
echo " # ${NAME}"
|
||||
echo " ${TOKEN_VAR}=<paste-discord-bot-token-here>"
|
||||
echo ""
|
||||
echo "2. Restart the AI stack:"
|
||||
echo ""
|
||||
echo " systemctl restart ai_stack.service"
|
||||
echo ""
|
||||
echo "3. In Paperclip, create an agent with HTTP adapter:"
|
||||
echo ""
|
||||
echo " Endpoint: http://${NAME}:${NEW_PORT}/v1/chat/completions"
|
||||
echo " API Key: ${API_KEY}"
|
||||
echo ""
|
||||
echo "────────────────────────────────────────────"
|
||||
@@ -8,10 +8,13 @@ services:
|
||||
- USER_GID=1000
|
||||
- GITEA__server__ROOT_URL=https://code.lazyworkhorse.net
|
||||
- GITEA__actions__ENABLED=true
|
||||
- GITEA__actions__DEFAULT_ACTIONS_URL=off
|
||||
- SSH_PORT=2222
|
||||
- SSH_LISTEN_PORT=2222
|
||||
# Enable Gitea Actions (act_runner required on host)
|
||||
- GITEA__actions__ENABLED=true
|
||||
# Don't fetch actions from GitHub (offline mode + local only)
|
||||
- GITEA__actions__DEFAULT_ACTIONS_URL=off
|
||||
volumes:
|
||||
- /mnt/HoardingCow_docker_data/Gitea:/data
|
||||
networks:
|
||||
|
||||
Reference in New Issue
Block a user