feat: image generation service — ComfyUI cluster integration

- POST /api/image/generate proxy endpoint (admin required)
- GET /api/image/status lists available image gen nodes
- Cluster AMQP protocol: cmd.image_generate, image_generated, image_failed
- Node agent auto-detects ComfyUI, registers image_gen capability
- ComfyUI workflow builder: CheckpointLoader → KSampler → VAEDecode → SaveImage
- Hardware probe checks ComfyUI reachability + checkpoint model list
- 27 tests covering cluster handlers, router, node agent, hardware, capabilities
- Config: CAIC_COMFYUI_BASE, CAIC_COMFYUI_TIMEOUT, comfyui_port in agent.ini
- Version bump to v1.1.0
- Documentation: ai.md, wiki/Developer-Architecture.md, current-wip.md, README.md, .env.example

docs/wiki/Developer-Architecture.md

@@ -22,9 +22,9 @@ Refactored from single-file (`app.py`) into modules under project root:
 | `rag.py` | Qdrant vector search, system prompt assembly, chunk_text() helper, collection stats |
 | `eviction.py` | Score-based RAG eviction engine (extracted from rag.py) |
 | `gpu.py` | AMD GPU stats via rocm-smi |
-| `hardware.py` | Hardware self-assessment — CPU, RAM, VRAM, service health probes |
+| `hardware.py` | Hardware self-assessment — CPU, RAM, VRAM, service health probes (llama-server, Qdrant, SearXNG, ComfyUI) |
 | `amqp.py` | aio-pika connection manager for RabbitMQ (connect, disconnect, publish, subscribe, auto-reconnect) |
-| `cluster.py` | Cluster node registry, event log, coordinator election, ping/pong, model swap handlers |
+| `cluster.py` | Cluster node registry, event log, coordinator election, ping/pong, model swap handlers, image generation request/response |
 | `triage.py` | Phi-4-mini query classification + `select_node()` for cluster routing |
 | `routers/` | One module per endpoint group |
 
@@ -36,6 +36,7 @@ Refactored from single-file (`app.py`) into modules under project root:
 | SearXNG | No | 8888 | Privacy-respecting web search |
 | Qdrant (coordinator) | No | 6333 | Vector database for RAG |
 | Ollama (worker) | No | 11434 | Embeddings for RAG chunk vectors |
+| ComfyUI (worker) | No | 8188 | Image generation (Stable Diffusion / Flux) |
 | RabbitMQ (coordinator) | No | 5672 | AMQP broker for cluster messaging |
 | rocm-smi | No | — | AMD GPU stats (host-level) |
 
@@ -50,6 +51,8 @@ Key base URLs are configured via environment variables with sensible defaults:
 | `SEARXNG_BASE` | `http://localhost:8888` | SearXNG |
 | `QDRANT_URL` | `http://192.168.50.108:6333` | Qdrant on coordinator |
 | `CAIC_AMQP_URL` | `amqp://caic:password@localhost:5672/caic` | RabbitMQ |
+| `CAIC_COMFYUI_BASE` | `http://192.168.50.115:8188` | ComfyUI on worker |
+| `CAIC_COMFYUI_TIMEOUT` | `120` | ComfyUI generation timeout (seconds) |
 
 ## 2. Request/Response Architecture
 
@@ -87,6 +90,17 @@ Key base URLs are configured via environment variables with sensible defaults:
 4. Three modes: `context` (SQLite with 1hr expiry), `ingest` (RAG/Qdrant), `both`
 5. Trigger `maybe_evict()` if ingest mode
 
+### 2.5 Image Generation Pipeline (`POST /api/image/generate`)
+
+1. Admin required, JSON body with prompt and optional params (width, height, steps, seed, model)
+2. Find active node with `image_gen` capability via `_find_image_node()`
+3. Publish `cmd.image_generate` via AMQP to selected worker node
+4. Worker node agent builds ComfyUI workflow (CheckpointLoader → KSampler → VAEDecode → SaveImage)
+5. Worker polls ComfyUI `/history/{prompt_id}` until image is ready
+6. Worker fetches PNG from ComfyUI `/view` endpoint, base64-encodes, publishes `image_generated` on `jc.system`
+7. Coordinator receives response, decodes base64, returns `image/png` to client
+8. `GET /api/image/status` returns available image gen nodes and their status
+
 ## 3. Data Model (SQLite)
 
 Key tables:
@@ -242,8 +256,8 @@ Every RabbitMQ server belongs to a cluster. Currently only the coordinator runs
 
 | Exchange | Type | Purpose |
 |----------|------|---------|
-| `jc.admin` | topic | Lifecycle commands: register, deregister, ping, pong, admitted, rejected; model commands: cmd.swap_model |
-| `jc.system` | topic | Events: model_ready, model_failed, node.*.heartbeat, event; coordinator queries: coord_query, coord_response |
+| `jc.admin` | topic | Lifecycle commands: register, deregister, ping, pong, admitted, rejected; model commands: cmd.swap_model; image commands: cmd.image_generate |
+| `jc.system` | topic | Events: model_ready, model_failed, image_generated, image_failed, node.*.heartbeat, event; coordinator queries: coord_query, coord_response |
 
 All exchanges, queues, and bindings are declared by `amqp.py` at startup. Worker runs `node_agent/agent.py` which connects as an AMQP client, registers, responds to ping, and handles model swap commands.
 
@@ -265,7 +279,7 @@ All streaming endpoints yield `data: {json}\n\n`:
 - No live external services required
 - Test factories reset `SESSIONS`, `PIN_ATTEMPTS`, `RATE_EVENTS` globals per test
 
-### 8.2 Test Coverage Areas (200 tests)
+### 8.2 Test Coverage Areas (228 tests)
 
 | Test file | Coverage |
 |-----------|----------|
@@ -277,6 +291,8 @@ All streaming endpoints yield `data: {json}\n\n`:
 | test_conversations.py | Full CRUD, guest admin, attachment_count |
 | test_error_envelopes.py | Global exception handler + stream errors |
 | test_gpu.py | GPU stats — rocm-smi (Linux), system_profiler (Darwin/Apple Silicon) |
+| test_hardware.py | Hardware assessment, service reachability |
+| test_image.py | Image generation — cluster handlers, router proxy, node agent ComfyUI integration, hardware probe, capability detection |
 | test_ingest.py | Bearer auth, chunk/embed/upsert, validation |
 | test_ip_allowlist.py | IP allowlist helper + middleware |
 | test_memories.py | Edit, search, stats |
@@ -311,5 +327,6 @@ On startup, `assess_hardware()` probes:
 - llama-server reachability + model list
 - Qdrant reachability + collection list
 - SearXNG reachability
+- ComfyUI reachability + checkpoint model list
 
 Writes `hardware_state.json` to working directory.