feat: add llama-cpp-hermes service with ROCm 6.1 + gfx906 support

- Add custom llama.cpp Dockerfile with ROCm 6.1 + gfx906 (MI50) build
- Add llama-cpp-hermes service serving Hermes 4.3 on dual MI50 GPUs
- Strip GPU devices/ROCm env from ollama service (CPU-only for embeddings)

Hermes 4.3 runs at ~19 t/s on dual MI50s with 160K context.

AGENTS.md

@@ -1,31 +0,0 @@
-# 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

ai/compose.yml

@@ -1,35 +1,4 @@
 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:
@@ -85,6 +54,10 @@ 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
@@ -94,6 +67,35 @@ 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
@@ -110,40 +112,7 @@ services:
       - /mnt/HoardingCow_docker_data/Ollama/ollama:/root/.ollama
     environment:
       - OLLAMA_VULKAN=0
-      - HSA_OVERRIDE_GFX_VERSION=9.0.6
-      - HCC_AMDGPU_TARGET=gfx906
-      - HIP_VISIBLE_DEVICES=0,1
-      - ROCR_VISIBLE_DEVICES=0,1
-      - HSA_ENABLE_SDMA=0 
       - OLLAMA_HOST=0.0.0.0
-      - OLLAMA_DEBUG=1
-      - OLLAMA_FLASH_ATTENTION=1
-      - OLLAMA_NUM_PARALLEL=2
-    devices:
-      # Map the render nodes and KFD for ROCm to work inside the container
-      - /dev/kfd:/dev/kfd
-      - /dev/dri:/dev/dri
-    group_add:
-      - "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:
@@ -153,47 +122,40 @@ networks:
     driver: bridge
     name: ai_backend
     
-  # llama_cpp_devstral:
-  #   image: ghcr.io/ggml-org/llama.cpp:server-rocm
-  #   container_name: llama_cpp_devstral
-  #   restart: unless-stopped
-  #   networks:
-  #     - ai_backend
-  #   ports:
-  #     - "8300:8080"
-  #   ipc: host
-  #   devices:
-  #     - "/dev/kfd:/dev/kfd"
-  #     - "/dev/dri:/dev/dri"
-  #   group_add:
-  #     - "303" # video
-  #     - "26"  # render
-  #   environment:
-  #     HSA_OVERRIDE_GFX_VERSION: 9.0.6
-  #     HIP_VISIBLE_DEVICES: 0,1
-  #     LLAMA_CACHE: /models
-  #   volumes:
-  #     - /mnt/HoardingCow_docker_data/Llama_cpp/models:/models
-  #     - /mnt/HoardingCow_docker_data/Llama_cpp/devstral-agent.jinja:/template.jinja
-  #   command: >
-  #     -hf unsloth/Devstral-Small-2-24B-Instruct-2512-GGUF:Devstral-Small-2-24B-Instruct-2512-Q8_0.gguf
-  #     -a devstral-2-small-llama_cpp
-  #     --chat-template-file /template.jinja
-  #     --host 0.0.0.0
-  #     --port 8080
-  #     --n-gpu-layers 99
-  #     --ctx-size 163840
-  #     --batch-size 4096
-  #     --ubatch-size 4096
-  #     --cache-type-k f16
-  #     --cache-type-v f16
-  #     --cache-reuse 256
-  #     --flash-attn on
-  #     --context-shift
-  #     --split-mode layer
-  #     --no-mmap
-  #     --n-predict -1
-  #     --parallel 2
+  llama-cpp-hermes:
+    image: llama-cpp:rocm-gfx906
+    container_name: llama-cpp-hermes
+    restart: unless-stopped
+    networks:
+      - ai_backend
+    ports:
+      - "127.0.0.1:8300:8080"
+    ipc: host
+    devices:
+      - /dev/kfd:/dev/kfd
+      - /dev/dri:/dev/dri
+    group_add:
+      - "303"
+      - "26"
+    environment:
+      - HSA_OVERRIDE_GFX_VERSION=9.0.6
+      - HSA_ENABLE_SDMA=0
+      - HIP_VISIBLE_DEVICES=0,1
+      - LLAMA_CACHE=/models
+    volumes:
+      - /mnt/HoardingCow_docker_data/Llama_cpp/models:/models
+      - /mnt/HoardingCow_docker_data/Ollama/ollama/models/blobs/sha256-17823599694fa3503ef54bf748d5078c6ce881f4d01616cafa255dc05d215a08:/model.gguf:ro
+    command: >
+      -m /model.gguf
+      --host 0.0.0.0
+      --port 8080
+      --gpu-layers 99
+      --ctx-size 163840
+      -ctk f16 -ctv f16
+      --flash-attn on
+      --split-mode layer
+      --no-mmap
+      --n-predict -1
 
   # vllm:
   #   image: nalanzeyu/vllm-gfx906:v0.9.0-rocm6.3

ai/llama-cpp/Dockerfile

@@ -0,0 +1,30 @@
+# llama-cpp-rocm6/Dockerfile
+# Custom llama.cpp server with ROCm 6.1 + gfx906 (MI50) support.
+# Build: docker build -t llama-cpp:rocm-gfx906 .
+
+FROM rocm/dev-ubuntu-22.04:6.1.2-complete AS builder
+RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y curl git build-essential pkg-config cmake make && rm -rf /var/lib/apt/lists/*
+ARG LLAMACPP_VERSION=b9596
+RUN git clone --depth 1 --branch ${LLAMACPP_VERSION} https://github.com/ggml-org/llama.cpp.git /build
+WORKDIR /build
+ENV HIP_PATH=/opt/rocm ROCM_PATH=/opt/rocm PATH=/opt/rocm/bin:/opt/rocm/llvm/bin:${PATH} CMAKE_PREFIX_PATH=/opt/rocm
+RUN mkdir build && cd build && \
+    cmake .. -DGGML_HIP=ON -DCMAKE_BUILD_TYPE=Release \
+      -DAMDGPU_TARGETS="gfx906:xnack-" \
+      -DCMAKE_POSITION_INDEPENDENT_CODE=ON \
+      -DGGML_CUDA=OFF -DGGML_VULKAN=OFF -DGGML_METAL=OFF \
+      -DBUILD_SHARED_LIBS=OFF && \
+    cmake --build . --target llama-server -- -j $(nproc)
+
+FROM ubuntu:24.04
+RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y \
+    ca-certificates curl libstdc++6 libgomp1 libopenblas0 \
+    libnuma1 libelf1 libdrm2 libdrm-amdgpu1 \
+    && rm -rf /var/lib/apt/lists/*
+COPY --from=builder /opt/rocm/lib/ /opt/rocm/lib/
+COPY --from=builder /opt/rocm/share/ /opt/rocm/share/
+COPY --from=builder /build/build/bin/llama-server /usr/local/bin/llama-server
+RUN echo /opt/rocm/lib > /etc/ld.so.conf.d/rocm.conf && ldconfig
+ENV HSA_OVERRIDE_GFX_VERSION=9.0.6 HCC_AMDGPU_TARGET=gfx906 HSA_ENABLE_SDMA=0
+EXPOSE 8080
+ENTRYPOINT ["/usr/local/bin/llama-server"]

ai/scripts/provision-hermes-worker.sh

@@ -1,135 +0,0 @@
-#!/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 "────────────────────────────────────────────"

versioncontrol/compose.yml

@@ -8,13 +8,10 @@ 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: