Merge branch 'master' into feat/hermes43-llamacpp

Conflicts resolved:
- hermes env: keep OLLAMA_HOST=ollama-cpu + master's new vars
- ollama→ollama-cpu rename: keep our rename over master's ollama

ai/compose.yml

@@ -1,4 +1,3 @@
-version: "3.8"
 services:
 
   # webui:
@@ -38,7 +37,7 @@ services:
     # Gateway run enables the internal API server on port 8642
     command: gateway run
     environment:
-      - OLLAMA_HOST=http://ollama:11434
+      - OLLAMA_HOST=http://ollama-cpu:11434
       - HERMES_DASHBOARD=1
       # Multi-profile: comma-separated list of profiles to run as gateways.
       # The entrypoint reads this and starts one gateway per profile.
@@ -50,6 +49,7 @@ services:
       - API_SERVER_KEY=hermes_local_key
       - GATEWAY_ALLOW_ALL_USERS=true
       - OPENROUTER_API_KEY=${OPENROUTER_API_KEY}
+      - OPENCODE_API_KEY=${OPENCODE_API_KEY}
       # ROCm for GPU-accelerated faster-whisper STT
       - HSA_OVERRIDE_GFX_VERSION=9.0.6
       - HCC_AMDGPU_TARGET=gfx906
@@ -127,13 +127,12 @@ services:
       - "traefik.http.routers.syncthing-https.tls.certresolver=njalla"
       - "traefik.http.services.syncthing.loadbalancer.server.port=8384"
 
-
-  ollama:
+  ollama-cpu:
     build:
       context: ./ollama
       dockerfile: Dockerfile
     image: ollama/ollama:rocm-gfx906
-    container_name: ollama
+    container_name: ollama-cpu
     tty: true
     restart: always 
     ports:
@@ -144,41 +143,69 @@ 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
+
+  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:
-      # 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"
+    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
 
   # --- Honcho + OpenConcho combiné: API + Web UI nginx/FastAPI ---
-  honcho-ui:
-    build: ./honcho
-    container_name: honcho-ui
+  honcho:
+    build:
+      context: ./honcho
+      ssh:
+        - default
+    container_name: honcho
     restart: unless-stopped
     environment:
       - DB_CONNECTION_URI=postgresql+psycopg://honcho:honcho_pass@honcho-db:5432/honcho
       - CACHE_URL=redis://honcho-redis:6379/0
       - CACHE_ENABLED=true
-      - EMBEDDING_VECTOR_DIMENSIONS=1536
-      - VECTOR_STORE_DIMENSIONS=1536
+      - EMBEDDING_VECTOR_DIMENSIONS=1024
+      - AUTH_USE_AUTH=true
+      - AUTH_JWT_SECRET=${HONCHO_AUTH_JWT_SECRET}
+      # Needed by deriver/dream to make LLM calls (api_key_env = "HONCHO_OPENAI_API_KEY" in config.toml)
+      - HONCHO_OPENAI_API_KEY=${HONCHO_OPENAI_API_KEY}
     volumes:
       - /mnt/HoardingCow_docker_data/Honcho/data:/app/data
+      - /mnt/HoardingCow_docker_data/Honcho/config.toml:/app/config.toml:ro
     networks:
       - ai_backend
       - ai_net
     labels:
       - "traefik.enable=true"
+      - "traefik.docker.network=ai_net"
 
       # Router for HTTP + redirect to HTTPS
       - "traefik.http.routers.honcho-http.rule=Host(`honcho.lazyworkhorse.net`)"
@@ -246,49 +273,12 @@ networks:
   ai_backend:
     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
 
+volumes:
+  honcho_data:
+    external: true
+    name: honcho_data
+    
   # vllm:
   #   image: nalanzeyu/vllm-gfx906:v0.9.0-rocm6.3
   #   container_name: vllm

ai/honcho/Dockerfile

@@ -2,14 +2,15 @@
 FROM python:3.13-slim-bookworm AS honcho-builder
 
 RUN apt-get update && \
-    apt-get install -y --no-install-recommends git && \
+    apt-get install -y --no-install-recommends git openssh-client && \
     rm -rf /var/lib/apt/lists/*
 
 COPY --from=ghcr.io/astral-sh/uv:0.9.24 /uv /bin/uv
 
-ARG HONCHO_REPO=https://github.com/plastic-labs/honcho
+ARG HONCHO_REPO=ssh://git@code.lazyworkhorse.net:2222/Hermes/honcho.git
 ARG HONCHO_REF=main
-RUN git clone --depth 1 --branch ${HONCHO_REF} ${HONCHO_REPO} /app
+RUN mkdir -p -m 0700 ~/.ssh && ssh-keyscan -p 2222 code.lazyworkhorse.net >> ~/.ssh/known_hosts 2>/dev/null
+RUN --mount=type=ssh git clone --depth 1 --branch ${HONCHO_REF} ${HONCHO_REPO} /app
 
 WORKDIR /app
 
@@ -29,8 +30,8 @@ RUN corepack enable && corepack prepare pnpm@latest --activate
 WORKDIR /app
 RUN apt-get update && apt-get install -y git && rm -rf /var/lib/apt/lists/*
 
-ARG OPENCONCHO_SHA=e490d911fcb27ee193558fd9a28856cde2057665
-RUN git clone --depth 1 https://github.com/offendingcommit/openconcho.git /app && \
+ARG OPENCONCHO_SHA=3b5c3293fc18d768dbe85285264a8d66c896bd81
+RUN --mount=type=ssh git clone --depth 1 ssh://git@code.lazyworkhorse.net:2222/gortium/openconcho.git /app && \
     git -C /app fetch --depth 1 origin ${OPENCONCHO_SHA} && \
     git -C /app checkout ${OPENCONCHO_SHA}
 
@@ -40,9 +41,21 @@ RUN pnpm --filter @openconcho/web build
 # runtime stage — nginx + Honcho FastAPI
 FROM python:3.13-slim-bookworm
 
+# Install nginx and create runtime dirs before dropping permissions
 RUN apt-get update && apt-get install -y --no-install-recommends nginx && \
     rm -rf /var/log/nginx/* && \
-    rm -rf /var/lib/apt/lists/*
+    rm -rf /var/lib/apt/lists/* && \
+    rm -f /etc/nginx/sites-enabled/default
+
+# Patch nginx.conf: comment out "user www-data;" so nginx master stays as root
+# (workers inherit root inside a container — fine for single-service isolation)
+RUN sed -i 's/^user /# user /' /etc/nginx/nginx.conf
+
+# Pre-create nginx runtime directories with proper ownership
+RUN mkdir -p /var/lib/nginx/body /var/lib/nginx/proxy /var/lib/nginx/fastcgi \
+             /var/lib/nginx/uwsgi /var/lib/nginx/scgi /var/lib/nginx/proxy_temp \
+             /var/cache/nginx && \
+    chown -R root:root /var/lib/nginx /var/cache/nginx
 
 # Honcho
 COPY --from=honcho-builder /app /app
@@ -56,17 +69,7 @@ COPY --from=openconcho-builder /app/packages/web/dist /usr/share/nginx/html
 
 # nginx config (proxies /v3/, /v2/ to Honcho on localhost:8000)
 COPY honcho-nginx.conf /etc/nginx/conf.d/default.conf
-RUN rm -f /etc/nginx/sites-enabled/default
-
-RUN groupadd --system app && \
-    useradd --system --gid app --create-home app && \
-    chown -R app:app /app /usr/share/nginx/html && \
-    # nginx runtime dirs need to exist for non-root master? Actually master is root, just ensure /var/lib/nginx exists
-    mkdir -p /var/lib/nginx/body /var/lib/nginx/proxy /var/lib/nginx/fastcgi /var/lib/nginx/uwsgi /var/lib/nginx/scgi && \
-    chown -R root:root /var/lib/nginx && \
-    chmod 755 /var/lib/nginx
 
 EXPOSE 80
 
-# nginx runs as root (needed for port 80 + runtime dirs), Honcho runs as app user
-CMD ["sh", "-c", "nginx && exec su -s /bin/sh app -c 'fastapi run --host 127.0.0.1 --port 8000 src/main.py'"]
+CMD ["bash", "-c", "nginx -g 'daemon off;' & fastapi run --host 127.0.0.1 --port 8000 src/main.py & python3 -m src.deriver & wait -n"]

ai/honcho/config.toml

@@ -29,17 +29,17 @@ URL = "redis://honcho-redis:6379/0"
 [llm]
 DEFAULT_MAX_TOKENS = 4096
 
-# Embeddings via Ollama (nomic-embed-text recommended on this system)
+# Embeddings via Ollama — bge-m3 provides 1024-dim
 [embedding]
-VECTOR_DIMENSIONS = 768
+VECTOR_DIMENSIONS = 1024
 MAX_INPUT_TOKENS = 8192
 
 [embedding.model_config]
 transport = "openai"
-model = "nomic-embed-text"
-base_url = "http://ollama:11434/v1"
+model = "bge-m3"
+overrides = {base_url = "http://ollama:11434/v1", api_key = "ollama"}
 
-# --- Deriver (user representation builder) ---
+# --- Deriver ---
 [deriver]
 ENABLED = true
 WORKERS = 1
@@ -47,9 +47,9 @@ POLLING_SLEEP_INTERVAL_SECONDS = 5.0
 FLUSH_ENABLED = true
 
 [deriver.model_config]
+overrides = {base_url = "https://opencode.ai/zen/go/v1", api_key_env = "HONCHO_OPENAI_API_KEY"}
 transport = "openai"
-model = "hermes-3"
-base_url = "http://ollama:11434/v1"
+model = "deepseek-v4-flash"
 
 # --- Dialectic ---
 [dialectic]
@@ -60,37 +60,37 @@ SESSION_HISTORY_MAX_TOKENS = 8192
 MAX_TOOL_ITERATIONS = 1
 MAX_OUTPUT_TOKENS = 512
 [dialectic.levels.minimal.model_config]
+overrides = {base_url = "https://opencode.ai/zen/go/v1", api_key_env = "HONCHO_OPENAI_API_KEY"}
 transport = "openai"
-model = "hermes-3"
-base_url = "http://ollama:11434/v1"
+model = "deepseek-v4-flash"
 
 [dialectic.levels.low]
 MAX_TOOL_ITERATIONS = 3
 [dialectic.levels.low.model_config]
+overrides = {base_url = "https://opencode.ai/zen/go/v1", api_key_env = "HONCHO_OPENAI_API_KEY"}
 transport = "openai"
-model = "hermes-3"
-base_url = "http://ollama:11434/v1"
+model = "deepseek-v4-flash"
 
 [dialectic.levels.medium]
 MAX_TOOL_ITERATIONS = 2
 [dialectic.levels.medium.model_config]
+overrides = {base_url = "https://opencode.ai/zen/go/v1", api_key_env = "HONCHO_OPENAI_API_KEY"}
 transport = "openai"
-model = "hermes-3"
-base_url = "http://ollama:11434/v1"
+model = "deepseek-v4-flash"
 
 [dialectic.levels.high]
 MAX_TOOL_ITERATIONS = 4
 [dialectic.levels.high.model_config]
+overrides = {base_url = "https://opencode.ai/zen/go/v1", api_key_env = "HONCHO_OPENAI_API_KEY"}
 transport = "openai"
-model = "hermes-3"
-base_url = "http://ollama:11434/v1"
+model = "deepseek-v4-flash"
 
 [dialectic.levels.max]
 MAX_TOOL_ITERATIONS = 10
 [dialectic.levels.max.model_config]
+overrides = {base_url = "https://opencode.ai/zen/go/v1", api_key_env = "HONCHO_OPENAI_API_KEY"}
 transport = "openai"
-model = "hermes-3"
-base_url = "http://ollama:11434/v1"
+model = "deepseek-v4-flash"
 
 # --- Summary ---
 [summary]
@@ -99,13 +99,28 @@ MESSAGES_PER_SHORT_SUMMARY = 20
 MESSAGES_PER_LONG_SUMMARY = 60
 
 [summary.model_config]
+overrides = {base_url = "https://opencode.ai/zen/go/v1", api_key_env = "HONCHO_OPENAI_API_KEY"}
 transport = "openai"
-model = "hermes-3"
-base_url = "http://ollama:11434/v1"
+model = "deepseek-v4-flash"
 
 # --- Dream ---
 [dream]
-ENABLED = false
+ENABLED = true
+
+[dream.model_config]
+overrides = {base_url = "https://opencode.ai/zen/go/v1", api_key_env = "HONCHO_OPENAI_API_KEY"}
+transport = "openai"
+model = "deepseek-v4-flash"
+
+[dream.deduction_model_config]
+overrides = {base_url = "https://opencode.ai/zen/go/v1", api_key_env = "HONCHO_OPENAI_API_KEY"}
+transport = "openai"
+model = "deepseek-v4-flash"
+
+[dream.induction_model_config]
+overrides = {base_url = "https://opencode.ai/zen/go/v1", api_key_env = "HONCHO_OPENAI_API_KEY"}
+transport = "openai"
+model = "deepseek-v4-flash"
 
 # --- Peer Card ---
 [peer_card]
@@ -114,4 +129,4 @@ ENABLED = true
 # --- Vector Store ---
 [vector_store]
 TYPE = "pgvector"
-DIMENSIONS = 768
+# DIMENSIONS is deprecated — EMBEDDING.VECTOR_DIMENSIONS is authoritative

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"]

network/compose.yml

@@ -82,37 +82,37 @@ networks:
     driver: bridge
     name: traefik_backend
   ai_net:
-    external: true
+    driver: bridge
     name: ai_net
   auth_net:
-    external: true
+    driver: bridge
     name: auth_net
   backup_net:
-    external: true
+    driver: bridge
     name: backup_net
   cloud_net:
-    external: true
+    driver: bridge
     name: cloud_net
   coms_net:
-    external: true
+    driver: bridge
     name: coms_net
   finance_net:
-    external: true
+    driver: bridge
     name: finance_net
   home_auto_net:
-    external: true
+    driver: bridge
     name: home_auto_net
   homepage_net:
-    external: true
+    driver: bridge
     name: homepage_net
   passman_net:
-    external: true
+    driver: bridge
     name: passman_net
   tak_net:
-    external: true
+    driver: bridge
     name: tak_net
   vc_net:
-    external: true
+    driver: bridge
     name: vc_net
 
   # duckdns: