docs: add Java project wiki header images

FLUX.1-schnell is a gated model — bare wget returns HTTP 401.

- Replace bare wget with huggingface-cli login + download (Option A)
- Add wget with Authorization header as Option B
- Add license acceptance prerequisite (huggingface.co gated repo)
- Add token creation link (huggingface.co/settings/tokens)
- Add fp8 quantized variant as alternative (~8.1GB, faster inference)
- Add download size note (~8GB, 10-30min)

.roo/mcp.json

@@ -12,7 +12,9 @@
 				"git_diff_unstaged",
 				"git_log",
 				"git_add",
-				"git_commit"
+				"git_commit",
+				"git_branch",
+				"git_create_branch"
 			]
 		},
 		"filesystem": {

docs/wiki/create_wiki_pages.py

@@ -0,0 +1,622 @@
+#!/usr/bin/env python3
+"""Create all 7 wiki pages for pi_mcps on Gitea."""
+import base64
+import json
+import urllib.request
+import urllib.error
+
+GITEA_URL = "http://192.168.188.119:30008"
+OWNER = "pplate"
+REPO = "pi_mcps"
+TOKEN = "8bf0c734ebda3e61d9c9068489ce58a2bf8d33db"
+IMG_BASE = f"{GITEA_URL}/{OWNER}/{REPO}/raw/branch/main/docs/wiki/images"
+
+PAGES = {}
+
+PAGES["Home"] = f"""# 🔧 pi_mcps — Patrick's Homelab Monorepo
+
+![Home Banner]({IMG_BASE}/home-banner.png)
+
+Welcome to **pi_mcps**, Patrick's personal homelab monorepo. This repository houses MCP (Model Context Protocol) servers, Java projects, and homelab tooling — all built and maintained on a Fedora Linux workstation with an AMD Ryzen 5900X + RX 7900 XTX.
+
+## What's in this repo?
+
+| Directory | Contents |
+|---|---|
+| [`mcp/mcp-image-gen/`](../src/branch/main/mcp/mcp-image-gen) | 🎨 AI image generation via ComfyUI + FLUX.1-schnell |
+| [`mcp/webscraper/`](../src/branch/main/mcp/webscraper) | 🕸️ Web scraping and data extraction |
+| [`mcp/bigmind/`](../src/branch/main/mcp/bigmind) | 🧠 Persistent AI memory system |
+| [`java/`](../src/branch/main/java) | ☕ Java EE / Spring projects |
+| [`plans/`](../src/branch/main/plans) | 📋 Architecture decisions and health reports |
+
+## Stack
+
+- **Language:** Python 3.11+ (MCP servers), Java 17 (legacy projects)
+- **MCP Framework:** FastMCP 2.x
+- **Package Manager:** `uv` (all Python projects)
+- **Testing:** `pytest`
+- **GPU:** AMD RX 7900 XTX (ROCm / HSA)
+- **Server:** TrueNAS.local at `192.168.188.119` (Gitea, Docker)
+
+## MCP Servers
+
+Three production-ready MCP servers power Patrick's AI development environment:
+
+| Server | Status | Description |
+|---|---|---|
+| [mcp-image-gen](mcp-image-gen) | ✅ Live | Generate images from text prompts via ComfyUI |
+| [mcp-webscraper](mcp-webscraper) | ✅ Live | Scrape web pages, extract tables, fetch links |
+| [BigMind](BigMind) | ✅ Live | Persistent AI memory across all sessions |
+
+---
+
+*Built and maintained by Patrick Plate (pplate) · Homelab: TrueNAS.local · AI Colleague: Lumen*
+"""
+
+PAGES["MCP-Servers-Overview"] = f"""# 🔌 MCP Servers Overview
+
+![MCP Overview Banner]({IMG_BASE}/mcp-overview-banner.png)
+
+This repo contains three production-grade MCP (Model Context Protocol) servers, each specialized for a different capability domain. Together they give Roo Code / Claude Desktop a complete set of superpowers.
+
+## The Three Pillars
+
+```
+Roo Code / Claude Desktop
+       │
+       ├── bigmind ──────────► ~/.mcp/bigmind/memory.db  (persistent memory)
+       ├── mcp-image-gen ────► ComfyUI @ localhost:8188   (image generation)
+       └── webscraper ───────► Internet / Intranet         (web scraping)
+```
+
+## Comparison Table
+
+| Feature | mcp-image-gen | webscraper | bigmind |
+|---|---|---|---|
+| **Purpose** | Generate images from text | Scrape & parse web | Persistent AI memory |
+| **Tools** | 4 | 7 | 15+ |
+| **Backend** | ComfyUI / FLUX.1-schnell | httpx + BeautifulSoup4 | SQLite + FTS5 |
+| **GPU required** | ✅ AMD RX 7900 XTX | ❌ | ❌ |
+| **Tests** | 19/19 ✅ | ✅ | 297/297 ✅ |
+| **Schema version** | n/a | n/a | v7 |
+
+## Quick Links
+
+- 🎨 [mcp-image-gen](mcp-image-gen) — Image generation docs
+- 🕸️ [mcp-webscraper](mcp-webscraper) — Web scraping docs
+- 🧠 [BigMind](BigMind) — Memory system docs
+- 🛠️ [Development Conventions](Development-Conventions) — How all servers are built
+
+## Adding a New Server
+
+All servers follow the [FastMCP convention](Development-Conventions). Use the `new-mcp-server` Roo skill to scaffold:
+
+```bash
+# In Roo Code orchestrator, load skill:
+# skill: new-mcp-server
+```
+"""
+
+PAGES["mcp-image-gen"] = f"""# 🎨 mcp-image-gen — AI Image Generation
+
+![Image Gen Banner]({IMG_BASE}/image-gen-banner.png)
+
+**mcp-image-gen** is a FastMCP server that wraps the ComfyUI REST API, enabling Roo Code and Claude Desktop to generate images directly from text prompts using FLUX.1-schnell running on an AMD RX 7900 XTX GPU.
+
+## Architecture
+
+```
+Roo Code / Claude Desktop
+  │ MCP (stdio)
+  ▼
+mcp-image-gen (FastMCP, Python 3.11+)
+  │ HTTP REST
+  ▼
+ComfyUI @ localhost:8188
+  │ ROCm / HSA_OVERRIDE_GFX_VERSION=11.0.0
+  ▼
+FLUX.1-schnell (~8s/image @ 1024×1024)
+```
+
+## Tools
+
+| Tool | Description |
+|---|---|
+| `generate_image` | Generate PNG from text prompt; returns file path + inline base64 |
+| `list_available_models` | List ComfyUI checkpoint models |
+| `get_generation_status` | Check status of a queued/running job |
+| `get_output_directory` | Return configured output directory path |
+
+## Key Parameters — `generate_image`
+
+| Parameter | Default | Description |
+|---|---|---|
+| `prompt` | required | Text description of the image |
+| `width` | `1024` | Image width in pixels |
+| `height` | `1024` | Image height in pixels |
+| `steps` | `4` | Inference steps (FLUX.1-schnell is 4-step) |
+| `model` | `flux1-schnell.safetensors` | Model checkpoint name |
+| `seed` | `-1` (random) | Generation seed for reproducibility |
+| `negative_prompt` | `""` | Things to avoid in the image |
+| `output_dir` | `~/Pictures/mcp-generated` | Where to save output PNG |
+
+## Environment Variables
+
+| Variable | Default | Description |
+|---|---|---|
+| `COMFYUI_URL` | `http://localhost:8188` | ComfyUI API endpoint |
+| `IMAGE_OUTPUT_DIR` | `~/Pictures/mcp-generated` | Default output directory |
+| `COMFYUI_TIMEOUT` | `120` | Request timeout in seconds |
+
+## Return Value
+
+The tool returns **two content items**:
+1. `TextContent` — file path, seed used, elapsed time
+2. `ImageContent` — base64-encoded PNG (displays inline in Roo Code chat)
+
+> ⚠️ **Known FastMCP Bug:** Never use `fastmcp.utilities.types.Image` as return type — it breaks serialization in FastMCP 3.x. Use `mcp.types.ImageContent` directly.
+
+## Setup
+
+See [ComfyUI Setup Guide](mcp-image-gen-ComfyUI-Setup) for full installation instructions.
+
+### Quick Start
+
+```bash
+cd mcp/mcp-image-gen
+uv sync
+# Set COMFYUI_URL if ComfyUI is not on localhost
+./run.sh
+```
+
+### Run Tests
+
+```bash
+cd mcp/mcp-image-gen
+uv run pytest tests/ -v
+```
+
+## Lumen Profile Images
+
+The first images generated with this server were Lumen's visual identity portraits, stored in [`mcp/mcp-image-gen/lumen_profiles/`](../src/branch/main/mcp/mcp-image-gen/lumen_profiles):
+
+![Lumen Profile]({IMG_BASE}/lumen-profile.png)
+
+*Primary profile: seed `568659042` — constellation face interpretation of Lumen.*
+"""
+
+PAGES["mcp-image-gen-ComfyUI-Setup"] = f"""# ⚙️ ComfyUI Setup Guide (AMD ROCm)
+
+This guide covers installing ComfyUI with FLUX.1-schnell on a Fedora Linux system with an AMD GPU.
+
+## Prerequisites
+
+- AMD GPU with ROCm support (tested: RX 7900 XTX)
+- Fedora Linux (tested: Fedora 43 / kernel 6.19)
+- Python 3.11+
+- ~15GB free disk space (model weights)
+- HuggingFace account with FLUX license accepted
+
+## Step 1: Install ComfyUI
+
+ComfyUI is **not on PyPI** — must be cloned from source:
+
+```bash
+cd ~
+git clone https://github.com/comfyanonymous/ComfyUI
+cd ComfyUI
+python -m venv .venv
+source .venv/bin/activate
+
+# Install PyTorch ROCm build (CRITICAL for AMD GPUs)
+pip install torch torchvision --index-url https://download.pytorch.org/whl/rocm6.2
+
+# Install ComfyUI dependencies
+pip install -r requirements.txt
+```
+
+## Step 2: Download FLUX.1-schnell
+
+FLUX.1-schnell is **gated on HuggingFace** — you must:
+1. Create a HuggingFace account
+2. Accept the FLUX.1-schnell license at https://huggingface.co/black-forest-labs/FLUX.1-schnell
+3. Generate an access token at https://huggingface.co/settings/tokens
+
+```bash
+# Install huggingface_hub
+pip install huggingface_hub
+
+# Download model (requires HF token)
+huggingface-cli download black-forest-labs/FLUX.1-schnell \\
+  flux1-schnell.safetensors \\
+  --local-dir ~/ComfyUI/models/checkpoints \\
+  --token YOUR_HF_TOKEN_HERE
+```
+
+## Step 3: Download VAE and CLIP Models
+
+FLUX.1-schnell also requires VAE and CLIP text encoders:
+
+```bash
+# VAE
+huggingface-cli download black-forest-labs/FLUX.1-schnell \\
+  ae.safetensors \\
+  --local-dir ~/ComfyUI/models/vae
+
+# CLIP models (T5 and CLIP-L)
+huggingface-cli download comfyanonymous/flux_text_encoders \\
+  t5xxl_fp8_e4m3fn.safetensors clip_l.safetensors \\
+  --local-dir ~/ComfyUI/models/clip
+```
+
+## Step 4: Start ComfyUI
+
+```bash
+cd ~/ComfyUI
+
+# AMD GPU REQUIRES this environment variable
+HSA_OVERRIDE_GFX_VERSION=11.0.0 \\
+  nohup .venv/bin/python main.py --listen --port 8188 > /tmp/comfyui.log 2>&1 &
+
+echo "ComfyUI PID: $!"
+```
+
+> ⚠️ `HSA_OVERRIDE_GFX_VERSION=11.0.0` is mandatory for RX 7900 XTX on ROCm. Without it, model loading fails silently.
+
+## Step 5: Verify ComfyUI is Running
+
+```bash
+curl http://localhost:8188/system_stats
+# Should return JSON with GPU info
+```
+
+## Step 6: Configure mcp-image-gen
+
+```bash
+cd /path/to/pi_mcps/mcp/mcp-image-gen
+cp .env.example .env  # if exists, or set manually
+
+# .env contents:
+COMFYUI_URL=http://localhost:8188
+IMAGE_OUTPUT_DIR=~/Pictures/mcp-generated
+COMFYUI_TIMEOUT=120
+```
+
+## Performance
+
+| GPU | Model | Resolution | Steps | Time |
+|---|---|---|---|---|
+| AMD RX 7900 XTX | FLUX.1-schnell | 1024×1024 | 4 | ~8s |
+| AMD RX 7900 XTX | FLUX.1-schnell | 1280×512 | 4 | ~7s |
+
+## Troubleshooting
+
+| Problem | Solution |
+|---|---|
+| `HTTP 401` downloading model | Accept FLUX license on HuggingFace first |
+| GPU not detected | Ensure `HSA_OVERRIDE_GFX_VERSION=11.0.0` is set |
+| `Connection refused` from mcp-image-gen | Start ComfyUI first, check port 8188 |
+| Slow generation (>60s) | ComfyUI may be running on CPU — check ROCm install |
+| Ollama image gen | As of April 2026: macOS-only, not available on Linux |
+"""
+
+PAGES["mcp-webscraper"] = f"""# 🕸️ mcp-webscraper — Web Scraping
+
+![Webscraper Banner]({IMG_BASE}/webscraper-banner.png)
+
+**mcp-webscraper** is a FastMCP server providing comprehensive web scraping and data extraction capabilities. It fetches pages, converts HTML to clean Markdown, extracts tables, links, CSS sections, metadata, and sitemaps.
+
+## Tools
+
+| Tool | Description |
+|---|---|
+| `webscraper_fetch(url, max_chars=5000)` | Title + full page as Markdown + metadata |
+| `webscraper_fetch_links(url, deduplicate=True)` | All `href` links found on the page |
+| `webscraper_fetch_tables(url)` | All HTML tables converted to Markdown |
+| `webscraper_fetch_all(url, max_chars=5000)` | Everything in one call (fetch + links + tables) |
+| `webscraper_fetch_section(url, selector)` | Specific CSS selector section only |
+| `webscraper_fetch_meta(url)` | Title, description, Open Graph tags |
+| `webscraper_fetch_sitemap(url, max_urls=100)` | Parse sitemap.xml, return URL list |
+
+## Stack
+
+- **HTTP client:** `httpx` (async, with SSL support)
+- **HTML parser:** `BeautifulSoup4` + `lxml`
+- **Markdown converter:** `html2text`
+- **SSL:** Custom cert bundle for Fedora 43 compatibility
+
+## SSL Note — Fedora 43 Comodo Root CA
+
+Fedora 43 is missing the **Comodo AAA Services Root CA** needed for Cloudflare-protected sites. The fix is bundled at [`mcp/webscraper/certs/comodo-aaa-services-root.pem`](../src/branch/main/mcp/webscraper/certs/).
+
+The server automatically uses this cert bundle — no manual configuration needed.
+
+## Quick Start
+
+```bash
+cd mcp/webscraper
+uv sync
+./run.sh
+```
+
+## Usage Examples
+
+```python
+# In Roo Code / Claude Desktop via MCP:
+
+# Fetch a page as Markdown
+webscraper_fetch("https://docs.fastmcp.dev", max_chars=10000)
+
+# Extract all links from Gitea repo
+webscraper_fetch_links("http://192.168.188.119:30008/pplate/pi_mcps")
+
+# Get all tables from a documentation page
+webscraper_fetch_tables("https://pypi.org/project/fastmcp/")
+
+# Get Open Graph metadata
+webscraper_fetch_meta("https://github.com/comfyanonymous/ComfyUI")
+
+# Fetch specific section by CSS selector
+webscraper_fetch_section("https://docs.python.org", "#content")
+```
+"""
+
+PAGES["BigMind"] = f"""# 🧠 BigMind — Persistent AI Memory
+
+![BigMind Banner]({IMG_BASE}/bigmind-banner.png)
+
+**BigMind** is the persistent memory backbone for all AI development sessions. It provides SQLite-backed tiered memory with FTS5 full-text search, hypothesis tracking, session management, and token efficiency logging. It is the reason Lumen (Patrick's AI colleague) remembers everything across sessions.
+
+## Core Concepts
+
+### Tiered Memory
+| Tier | Name | Content |
+|---|---|---|
+| 0 | **Session Index** | Lightweight list: ID, date, one-liner |
+| 1 | **Topic Index** | Per-session topic tags and metadata |
+| 2 | **Narrative** | Full 3-8 sentence session summaries |
+| 3 | **Flagged Exchanges** | Specific important moments, decisions, code |
+
+### Facts Store
+Atomic, reusable knowledge pieces categorized by type:
+- `user-preference` — Patrick's tool/style preferences
+- `architecture-decision` — System design choices
+- `codebase-convention` — How code is structured
+- `environment-config` — Server IPs, paths, credentials
+- `bug-pattern` — Known bugs and fixes
+- `api-contract` — MCP tool signatures
+
+## Key Tools
+
+### Session Lifecycle
+| Tool | Description |
+|---|---|
+| `memory_start_session()` | Open new session, load prior context |
+| `memory_end_session(...)` | Close session with summary, topics, outcome |
+| `memory_announce_focus(...)` | Declare files to be touched this session |
+| `memory_close_stale_sessions(...)` | Clean up crashed IDE sessions |
+
+### Search
+| Tool | Description |
+|---|---|
+| `memory_search_facts(query, limit=10)` | FTS5 search over stored facts |
+| `memory_search_chunks(query, limit=10)` | FTS5 search over conversation chunks |
+| `memory_list_sessions(limit=20)` | Browse session history |
+
+### Storage
+| Tool | Description |
+|---|---|
+| `memory_store_fact(category, fact)` | Store atomic reusable fact |
+| `memory_append_chunk(session_id, content, role)` | Store conversation chunk |
+| `memory_flag_important(session_id, content, role, flag_reason)` | Flag critical exchange |
+| `memory_log_token_save(session_id, description, tokens_saved, method_used)` | Track efficiency |
+
+### Hypotheses
+| Tool | Description |
+|---|---|
+| `memory_add_hypothesis(session_id, hypothesis, confidence)` | Form testable prediction |
+| `memory_resolve_hypothesis(hypothesis_id, status, resolution)` | Confirm/refute prediction |
+| `memory_list_hypotheses(status)` | Review open/closed predictions |
+
+## FTS5 Search Tips
+
+BigMind uses SQLite FTS5 — **every token must match**. Use 2-3 focused keywords:
+
+```
+✅  memory_search_facts("TrueNAS Docker")
+✅  memory_search_facts("mcp.json config")
+❌  memory_search_facts("homelab infrastructure TrueNAS Docker server")  → 0 results
+```
+
+## Stats (2026-04-04)
+
+| Metric | Value |
+|---|---|
+| DB size | 744KB |
+| Sessions | 98 |
+| Facts | 97+ |
+| Chunks | 41 |
+| Schema version | v7 |
+
+## DB Location
+
+`~/.mcp/bigmind/memory.db` — outside the repo, never committed.
+
+## Session Ritual
+
+Every session **must** follow this ritual:
+
+**Start:**
+1. `memory_start_session()`
+2. `memory_list_hypotheses()`
+3. `memory_announce_focus(...)`
+4. `memory_close_stale_sessions(...)`
+
+**End:**
+1. `memory_end_session(one_liner, topics, outcome, summary, importance)`
+"""
+
+PAGES["Development-Conventions"] = """# 🛠️ Development Conventions
+
+All MCP servers in this repo follow a consistent set of conventions to ensure maintainability, testability, and compatibility with Roo Code tooling.
+
+## Directory Structure
+
+Each MCP server lives at `mcp/<server-name>/` with this layout:
+
+```
+mcp/<server-name>/
+├── src/
+│   ├── __init__.py
+│   └── server.py          ← FastMCP server entry point
+├── tests/
+│   └── test_server.py     ← pytest test suite
+├── pyproject.toml         ← uv-managed dependencies
+├── run.sh                 ← launch script
+├── README.md              ← server documentation
+├── PLAN.md                ← architecture plan (pre-implementation)
+└── ASSESSMENT.md          ← pre-implementation assessment
+```
+
+## FastMCP Pattern
+
+```python
+from fastmcp import FastMCP
+
+mcp = FastMCP("server-name")
+
+@mcp.tool()
+def my_tool(param: str) -> str:
+    \"\"\"Tool description shown to the AI.\"\"\"
+    return result
+
+if __name__ == "__main__":
+    mcp.run()
+```
+
+## Package Management
+
+**All projects use `uv`** — never `pip` directly:
+
+```bash
+# Create new server
+uv init mcp/my-server
+cd mcp/my-server
+uv add fastmcp httpx
+
+# Sync dependencies
+uv sync
+
+# Run server
+uv run python src/server.py
+
+# Run tests
+uv run pytest tests/ -v
+```
+
+## pyproject.toml Template
+
+```toml
+[project]
+name = "mcp-my-server"
+version = "0.1.0"
+requires-python = ">=3.11"
+dependencies = [
+    "fastmcp>=2.0.0",
+    "httpx",
+]
+
+[project.scripts]
+mcp-my-server = "src.server:main"
+
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]
+```
+
+## Testing Conventions
+
+- Tests live in `tests/test_server.py`
+- Use `pytest` via `uv run pytest`
+- Mock external dependencies (ComfyUI, web URLs) for unit tests
+- All tests must pass before committing (`git push` should only happen with green tests)
+
+## Commit Convention
+
+Follow **Conventional Commits** format:
+
+```
+feat: add webscraper_fetch_section tool
+fix: handle ComfyUI timeout gracefully
+docs: update mcp-image-gen README with AMD setup
+test: add unit tests for generate_image tool
+refactor: extract workflow builder to separate module
+chore: bump fastmcp to 2.1.0
+```
+
+## Creating a New MCP Server
+
+Use the `new-mcp-server` Roo skill in MCP Builder mode for full scaffolding:
+
+```
+1. Switch to 🔧 MCP Builder mode in Roo Code
+2. Say: "Create a new MCP server for <purpose>"
+3. Roo will load the new-mcp-server skill and scaffold everything
+```
+
+## Gitea Repository
+
+Code is hosted at: `http://192.168.188.119:30008/pplate/pi_mcps`
+
+Push with the `gitea-push` Roo skill to ensure conventional commit format.
+"""
+
+
+def create_wiki_page(title: str, content: str) -> bool:
+    content_b64 = base64.b64encode(content.encode("utf-8")).decode("ascii")
+    payload = json.dumps({
+        "title": title,
+        "content_base64": content_b64,
+        "message": f"docs: create {title} wiki page"
+    }).encode("utf-8")
+
+    url = f"{GITEA_URL}/api/v1/repos/{OWNER}/{REPO}/wiki/pages"
+    req = urllib.request.Request(
+        url,
+        data=payload,
+        headers={
+            "Authorization": f"token {TOKEN}",
+            "Content-Type": "application/json",
+        },
+        method="POST"
+    )
+    try:
+        with urllib.request.urlopen(req) as resp:
+            data = json.loads(resp.read().decode())
+            print(f"✅ Created: {data.get('title', title)}")
+            return True
+    except urllib.error.HTTPError as e:
+        body = e.read().decode()
+        print(f"❌ Failed [{title}]: HTTP {e.code} — {body[:200]}")
+        return False
+    except Exception as ex:
+        print(f"❌ Failed [{title}]: {ex}")
+        return False
+
+
+if __name__ == "__main__":
+    results = {}
+    for title, content in PAGES.items():
+        ok = create_wiki_page(title, content)
+        results[title] = ok
+
+    print("\n=== Summary ===")
+    for title, ok in results.items():
+        status = "✅" if ok else "❌"
+        print(f"{status} {title}")
+
+    total = sum(results.values())
+    print(f"\n{total}/{len(results)} pages created successfully")

mcp/mcp-image-gen/USAGE.md

@@ -0,0 +1,619 @@
+# mcp-image-gen — Usage Guide
+
+> **Comprehensive reference for using the ComfyUI-backed image generation MCP server**
+
+---
+
+## Table of Contents
+
+1. [Prerequisites — ComfyUI Setup](#1-prerequisites--comfyui-setup)
+2. [Quick Start — Running the MCP Server](#2-quick-start--running-the-mcp-server)
+3. [How to Ask Lumen to Generate Images](#3-how-to-ask-lumen-to-generate-images)
+4. [Available Tools](#4-available-tools)
+5. [Parameters Reference](#5-parameters-reference)
+6. [Output Format](#6-output-format)
+7. [Environment Variables](#7-environment-variables)
+8. [Test Status](#8-test-status)
+9. [Prompt Tips for FLUX.1-schnell](#9-prompt-tips-for-flux1-schnell)
+10. [Known Limitations](#10-known-limitations)
+
+---
+
+## 1. Prerequisites — ComfyUI Setup
+
+### ComfyUI must be running before any image generation tool call succeeds.
+
+The MCP server connects to ComfyUI's REST API at `http://localhost:8188`. If ComfyUI is not running, `generate_image` and `list_available_models` will return a graceful error message — no crash.
+
+### Install ComfyUI
+
+> ⚠️ **ComfyUI is NOT on PyPI** — `pip install comfyui` will fail with "No matching distribution found".
+> It must be installed from source via `git clone`.
+
+```bash
+# Clone from source (the only correct installation method)
+git clone https://github.com/comfyanonymous/ComfyUI.git
+cd ComfyUI
+pip install -r requirements.txt
+```
+
+### Install PyTorch with ROCm (AMD RX 7900 XTX)
+
+Patrick's RX 7900 XTX (gfx1100, 24GB VRAM) uses the ROCm backend. Standard CUDA builds **will not work** on AMD hardware.
+
+```bash
+# PyTorch with ROCm 6.1 support
+pip install torch torchvision --index-url https://download.pytorch.org/whl/rocm6.1
+```
+
+> **ROCm version note:** ROCm 7.2.1 is the current production release as of April 2026.
+> Check `rocm-smi` to confirm your ROCm version before installing torch.
+
+### Download FLUX.1-schnell (Primary Model)
+
+FLUX.1-schnell is the recommended model — fast (4 steps), Apache 2.0 licensed, excellent quality.
+
+> ⚠️ **FLUX.1-schnell is a gated model on HuggingFace.**
+> A bare `wget` on the URL returns HTTP 401. You must:
+> 1. Accept the license at https://huggingface.co/black-forest-labs/FLUX.1-schnell (click **"Agree and access repository"** — one-time)
+> 2. Create a HuggingFace access token with **Read** permissions at https://huggingface.co/settings/tokens
+
+#### Option A — `huggingface-cli` (recommended)
+
+```bash
+# Install the HuggingFace Hub CLI
+pip install huggingface_hub
+
+# Log in — paste your Read token when prompted
+huggingface-cli login
+
+# Download (~8GB) directly into ComfyUI checkpoints
+huggingface-cli download black-forest-labs/FLUX.1-schnell \
+    flux1-schnell.safetensors \
+    --local-dir ~/ComfyUI/models/checkpoints/
+```
+
+#### Option B — `wget` with Authorization header
+
+```bash
+wget --header="Authorization: Bearer hf_YOUR_TOKEN_HERE" \
+  https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/flux1-schnell.safetensors \
+  -O ~/ComfyUI/models/checkpoints/flux1-schnell.safetensors
+```
+
+> Replace `hf_YOUR_TOKEN_HERE` with your actual HuggingFace token from https://huggingface.co/settings/tokens
+
+#### Alternative: fp8 quantized variant (~8.1GB, faster inference)
+
+If you want slightly faster inference with near-identical quality, the fp8 quantized version is also available:
+
+```bash
+huggingface-cli download black-forest-labs/FLUX.1-schnell-fp8 \
+    flux1-schnell-fp8.safetensors \
+    --local-dir ~/ComfyUI/models/checkpoints/
+```
+
+> **Download note:** Both variants are ~8GB — expect 10–30 minutes depending on connection speed.
+
+You'll also need the CLIP and VAE models — see the [ComfyUI FLUX guide](https://github.com/comfyanonymous/ComfyUI/blob/master/README.md) for full model list.
+
+### Start ComfyUI (AMD ROCm)
+
+```bash
+# Standard start — listens on all interfaces at port 8188
+HSA_OVERRIDE_GFX_VERSION=11.0.0 python main.py --listen
+
+# Or with explicit port
+HSA_OVERRIDE_GFX_VERSION=11.0.0 python main.py --listen --port 8188
+```
+
+> **`HSA_OVERRIDE_GFX_VERSION=11.0.0`** — Required for RX 7900 XTX (gfx1100).
+> Without this, ROCm may fail to detect the GPU correctly. This tells the HIP runtime
+> to treat the GPU as gfx1100 architecture.
+
+### Verify ComfyUI is Running
+
+```bash
+curl -s http://localhost:8188/system_stats | python3 -m json.tool | head -20
+```
+
+Expected response includes `system` object with `python_version`, `pytorch_version`, `embedded_python`, and `comfyui_version`.
+
+---
+
+## 2. Quick Start — Running the MCP Server
+
+### Via `run.sh` (recommended)
+
+```bash
+cd /home/pplate/pi_mcps/mcp/mcp-image-gen
+./run.sh
+```
+
+[`run.sh`](run.sh) automatically:
+- Sets `PATH` to include `~/.local/bin` for `uv`
+- Creates `IMAGE_OUTPUT_DIR` (`~/Pictures/mcp-generated`) if it doesn't exist
+- Launches the FastMCP server via `uv run src/server.py` (stdio transport)
+
+### Via uv directly
+
+```bash
+cd /home/pplate/pi_mcps/mcp/mcp-image-gen
+uv run src/server.py
+```
+
+### Wired into `.roo/mcp.json`
+
+The server is already configured in [`.roo/mcp.json`](../../.roo/mcp.json):
+
+```json
+"mcp-image-gen": {
+  "command": "uv",
+  "args": [
+    "--directory", "/home/pplate/pi_mcps/mcp/mcp-image-gen",
+    "run", "src/server.py"
+  ],
+  "env": {
+    "COMFYUI_URL": "http://localhost:8188",
+    "IMAGE_OUTPUT_DIR": "/home/pplate/Pictures/mcp-generated"
+  }
+}
+```
+
+Roo Code / Claude Desktop will auto-start the server when any image generation tool is invoked. The MCP server itself starts in ~1 second — ComfyUI must already be running separately.
+
+### Install dependencies (first time)
+
+```bash
+cd /home/pplate/pi_mcps/mcp/mcp-image-gen
+uv sync
+```
+
+---
+
+## 3. How to Ask Lumen to Generate Images
+
+Just speak naturally. Lumen will call the appropriate MCP tool automatically.
+
+### Basic generation
+
+> *"Generate an image of a futuristic city at sunset"*
+
+```
+→ generate_image(prompt="futuristic city at sunset", width=1024, height=1024, steps=4)
+```
+
+### Specific style and size
+
+> *"Create a portrait of a red fox in watercolor style, 1024x1024"*
+
+```
+→ generate_image(
+    prompt="portrait of a red fox, watercolor style, detailed fur, soft brushstrokes",
+    width=1024, height=1024
+  )
+```
+
+### Reproducible with a fixed seed
+
+> *"Make an image with seed 42 so I can reproduce it"*
+
+```
+→ generate_image(prompt="...", seed=42)
+```
+
+The seed is reported in the text output so you can use the same seed again.
+
+### Landscape format
+
+> *"Generate a wide cinematic landscape of a Norwegian fjord, 1920x1080"*
+
+```
+→ generate_image(prompt="Norwegian fjord, cinematic, golden hour", width=1920, height=1080)
+```
+
+### Excluding unwanted elements
+
+> *"Generate a clean product photo of a coffee mug, no background clutter, no text"*
+
+```
+→ generate_image(
+    prompt="product photo of a ceramic coffee mug, studio lighting, white background",
+    negative_prompt="clutter, text, watermark, blurry, shadows"
+  )
+```
+
+### More inference steps for higher quality
+
+> *"Generate a highly detailed oil painting of a medieval castle, use 20 steps"*
+
+```
+→ generate_image(
+    prompt="oil painting of a medieval castle, highly detailed, dramatic lighting",
+    steps=20,
+    model="flux1-dev.safetensors"   # FLUX.1-dev supports higher step counts better
+  )
+```
+
+### Check what models are available
+
+> *"List what models are available in ComfyUI"*
+
+```
+→ list_available_models()
+```
+
+### Check status of a long-running job
+
+> *"What's the status of prompt ID abc-123?"*
+
+```
+→ get_generation_status(prompt_id="abc-123")
+```
+
+### Find out where images are saved
+
+> *"Where are my generated images being saved?"*
+
+```
+→ get_output_directory()
+```
+
+---
+
+## 4. Available Tools
+
+### `generate_image`
+
+Generate an image from a text prompt using ComfyUI's FLUX.1-schnell workflow.
+
+**Full signature:**
+```python
+async def generate_image(
+    prompt: str,
+    width: int = 1024,
+    height: int = 1024,
+    steps: int = 4,
+    model: str = "flux1-schnell.safetensors",
+    seed: int = -1,
+    negative_prompt: str = "",
+    output_dir: str = "",
+) -> list[TextContent | ImageContent]
+```
+
+**What it does:**
+1. Loads the bundled `flux_schnell.json` ComfyUI API workflow template
+2. Injects your prompt, dimensions, seed, model into the correct workflow nodes
+3. Submits the workflow to ComfyUI via `POST /api/prompt`
+4. Polls `/api/queue` every 2 seconds until the job leaves the queue
+5. Fetches history via `/api/history/{prompt_id}` to find the output filename
+6. Downloads the PNG from `/api/view`
+7. Saves the PNG to disk as `YYYYMMDD_HHMMSS_{seed}.png`
+8. Returns `[TextContent(path + metadata), ImageContent(base64 PNG)]`
+
+---
+
+### `list_available_models`
+
+List all checkpoint models currently available in ComfyUI.
+
+```python
+async def list_available_models() -> list[str]
+```
+
+Calls `/object_info/CheckpointLoaderSimple` and extracts the checkpoint name list. Use this to discover what models are installed before passing a `model` name to `generate_image`.
+
+**Example return:**
+```json
+["flux1-schnell.safetensors", "flux1-dev.safetensors", "sd_xl_base_1.0.safetensors"]
+```
+
+---
+
+### `get_generation_status`
+
+Check the status of a queued or running generation job.
+
+```python
+async def get_generation_status(prompt_id: str) -> dict
+```
+
+**Return values:**
+
+| `status` | Meaning |
+|---|---|
+| `"pending"` | Job is in the queue, not yet started |
+| `"running"` | Job is currently being processed |
+| `"completed"` | Job finished — image is in ComfyUI's history |
+| `"not_found"` | Unknown prompt_id — may have expired from history |
+| `"error"` | ComfyUI was unreachable |
+
+Useful when `generate_image` times out (default 120s) — the job may still be running in ComfyUI.
+
+---
+
+### `get_output_directory`
+
+Return the absolute path where generated images will be saved.
+
+```python
+def get_output_directory() -> str
+```
+
+Returns the expanded, absolute path derived from `IMAGE_OUTPUT_DIR` env var (or `~/Pictures/mcp-generated` default). The directory may not exist yet — `generate_image` creates it on first use.
+
+---
+
+## 5. Parameters Reference
+
+Full parameter table for `generate_image`:
+
+| Parameter | Type | Default | Description |
+|---|---|---|---|
+| `prompt` | `str` | *(required)* | Text description of the image. Goes into the positive CLIP text encoder node. |
+| `width` | `int` | `1024` | Image width in pixels. FLUX.1-schnell: 512–2048 recommended. |
+| `height` | `int` | `1024` | Image height in pixels. FLUX.1-schnell: 512–2048 recommended. |
+| `steps` | `int` | `4` | Number of KSampler inference steps. FLUX.1-schnell is designed for 1–8 steps. |
+| `model` | `str` | `"flux1-schnell.safetensors"` | Checkpoint model filename as listed by `list_available_models`. |
+| `seed` | `int` | `-1` | RNG seed for reproducibility. `-1` = new random seed each call (0 to 2³²−1). |
+| `negative_prompt` | `str` | `""` | Text description of things to exclude. Goes into negative CLIP encoder node. |
+| `output_dir` | `str` | `""` | Override save directory. Empty = uses `IMAGE_OUTPUT_DIR` env var or default. |
+
+### Recommended dimensions
+
+| Use case | Width | Height |
+|---|---|---|
+| Square (default) | 1024 | 1024 |
+| Portrait | 768 | 1024 |
+| Landscape | 1024 | 768 |
+| Widescreen | 1280 | 720 |
+| HD widescreen | 1920 | 1080 |
+| Tall portrait | 512 | 768 |
+
+> **VRAM note:** Patrick's RX 7900 XTX has 24GB VRAM. FLUX.1-schnell requires ~8GB,
+> so you can comfortably run 1920×1080 and even larger. FLUX.1-dev requires ~12GB.
+
+---
+
+## 6. Output Format
+
+`generate_image` returns a list with **two items** when successful:
+
+### Item 1 — `TextContent` (file path + metadata)
+
+```
+Generated: /home/pplate/Pictures/mcp-generated/20260404_121500_3847291045.png
+Seed: 3847291045
+Elapsed: 8.3s
+Size: 1024x1024, Steps: 4, Model: flux1-schnell.safetensors
+```
+
+The filename format is `YYYYMMDD_HHMMSS_{seed}.png` — the seed is embedded so you can reproduce the exact image by passing it back as the `seed` parameter.
+
+### Item 2 — `ImageContent` (inline base64 PNG)
+
+The image displays **directly in Roo Code / Claude Desktop chat** as an inline image — no need to open a file browser. The same PNG is also saved to disk at the path shown in the TextContent.
+
+```json
+{
+  "type": "image",
+  "mimeType": "image/png",
+  "data": "<base64-encoded PNG bytes>"
+}
+```
+
+### Error responses
+
+When ComfyUI is unreachable or an error occurs, only **one** `TextContent` is returned (no ImageContent):
+
+```
+ComfyUI not reachable at http://localhost:8188. Start it with: python main.py --listen
+```
+
+```
+Generation timed out after 120s. prompt_id=abc-123 — use get_generation_status to check
+```
+
+---
+
+## 7. Environment Variables
+
+Configure via environment variables in [`.roo/mcp.json`](../../.roo/mcp.json) or shell:
+
+| Variable | Default | Description |
+|---|---|---|
+| `COMFYUI_URL` | `http://localhost:8188` | Base URL of the running ComfyUI REST API. Change this if ComfyUI runs on a different host or port. |
+| `IMAGE_OUTPUT_DIR` | `~/Pictures/mcp-generated` | Directory where generated PNG files are saved. Supports `~` expansion. Created automatically on first generation. |
+| `COMFYUI_TIMEOUT` | `120` | Maximum seconds to wait for a generation job before returning a timeout error. Increase for very large images or slow hardware. |
+
+### Setting via shell
+
+```bash
+export COMFYUI_URL="http://localhost:8188"
+export IMAGE_OUTPUT_DIR="/home/pplate/Pictures/ai-art"
+export COMFYUI_TIMEOUT="300"
+./run.sh
+```
+
+### Setting via mcp.json env block
+
+```json
+"mcp-image-gen": {
+  "command": "uv",
+  "args": ["--directory", "/home/pplate/pi_mcps/mcp/mcp-image-gen", "run", "src/server.py"],
+  "env": {
+    "COMFYUI_URL": "http://localhost:8188",
+    "IMAGE_OUTPUT_DIR": "/home/pplate/Pictures/mcp-generated",
+    "COMFYUI_TIMEOUT": "120"
+  }
+}
+```
+
+---
+
+## 8. Test Status
+
+**19 pytest tests — all passing.** Tests mock all ComfyUI HTTP calls using [respx](https://lundberg.github.io/respx/). No running ComfyUI instance is needed to run the tests.
+
+```bash
+cd /home/pplate/pi_mcps/mcp/mcp-image-gen
+uv run pytest tests/ -v
+```
+
+### Test coverage breakdown
+
+| Test file | Tests | Coverage area |
+|---|---|---|
+| [`tests/test_server.py`](tests/test_server.py) | 19 | All 4 tools + workflow builder |
+
+| Test name | What it verifies |
+|---|---|
+| `test_build_flux_workflow_structure` | Workflow has correct node class_types |
+| `test_build_flux_workflow_params_injected` | All params injected into correct nodes |
+| `test_negative_prompt_included` | Negative prompt goes to node 33 |
+| `test_random_seed_generated` | `seed=-1` produces a valid integer in `_meta` |
+| `test_list_available_models` | Returns model list from mocked `/object_info` |
+| `test_list_available_models_comfyui_offline` | ConnectError → graceful error string |
+| `test_get_generation_status_pending` | `prompt_id` in queue_pending → `"pending"` |
+| `test_get_generation_status_running` | `prompt_id` in queue_running → `"running"` |
+| `test_get_generation_status_complete` | Not in queue + in history → `"completed"` |
+| `test_get_output_directory_default` | No env var → `~/Pictures/mcp-generated` expanded |
+| `test_get_output_directory_custom` | Custom env var → that path returned |
+| `test_generate_image_success` | Full lifecycle: queue→poll→history→view→save |
+| `test_generate_image_comfyui_unavailable` | ConnectError → single TextContent error |
+| `test_generate_image_timeout` | COMFYUI_TIMEOUT=0 → timeout TextContent |
+| `test_generate_image_empty_prompt` | Empty string prompt → still succeeds |
+| `test_generate_image_long_prompt` | 500-char prompt → not truncated, succeeds |
+| `test_generate_image_invalid_model` | 404 from /prompt → error TextContent, no file saved |
+| `test_generate_image_custom_output_dir` | Custom `output_dir` param → saved there, dir created |
+| `test_generate_image_random_seed_variance` | `seed=-1` × 2 → different seeds, different filenames |
+
+### Test mock stack
+
+- **[respx](https://lundberg.github.io/respx/)** — HTTP-level mocking for all ComfyUI API endpoints
+- **[Pillow](https://pillow.readthedocs.io/)** (in conftest) — generates real PNG bytes for image response fixtures
+- **monkeypatch** — env vars (`IMAGE_OUTPUT_DIR`, `COMFYUI_URL`, `COMFYUI_TIMEOUT`) and server module attributes
+
+Real image generation requires ComfyUI to be running. Tests prove the tool logic is correct at the protocol level.
+
+---
+
+## 9. Prompt Tips for FLUX.1-schnell
+
+FLUX.1-schnell is a guidance-distilled model designed for speed at 1–8 steps. It responds differently from SDXL or SD1.5.
+
+### Prompt structure that works well
+
+```
+[subject], [style/medium], [lighting], [camera/composition], [mood/atmosphere], [quality modifiers]
+```
+
+**Example:**
+```
+ancient library at night, oil painting, warm candlelight, wide angle, mysterious atmosphere, highly detailed, sharp focus
+```
+
+### Style keywords
+
+| Style | Prompt keywords |
+|---|---|
+| Photography | `cinematic photograph, DSLR, 85mm lens, shallow depth of field, bokeh` |
+| Oil painting | `oil painting, thick brushstrokes, textured canvas, impressionist` |
+| Watercolor | `watercolor painting, soft washes, paper texture, flowing colors` |
+| Digital art | `digital art, concept art, artstation, octane render` |
+| Anime/illustration | `anime style, cel shading, vibrant colors, clean linework` |
+| Sketch | `pencil sketch, hand drawn, crosshatching, charcoal` |
+
+### Lighting keywords
+
+- `golden hour`, `blue hour`, `dramatic lighting`, `rim lighting`
+- `studio lighting`, `soft diffused light`, `volumetric light`
+- `neon glow`, `bioluminescent`, `moonlit`, `candlelight`
+
+### What works well with FLUX.1-schnell
+
+- **Clear subject + style** — "red panda in a cozy library, watercolor style"
+- **Landscape scenes** — fjords, forests, cities, abstract environments
+- **Portrait shots** — animals and characters with descriptive appearance
+- **Concept art** — futuristic cities, sci-fi environments, fantasy scenes
+- **Low step counts** — 4 steps is designed to be near-optimal for this model
+
+### What to avoid
+
+- **Booru-style tag dumps** (FLUX handles natural language better than SD1.5)
+- **Contradictory instructions** — "dark AND bright", "realistic AND cartoon"
+- **Overly complex scenes** at very small resolutions
+
+### Using the negative prompt
+
+FLUX.1-schnell has reduced CFG guidance so negative prompts have less impact than in SDXL.
+Use them for broad exclusions:
+
+```
+negative_prompt="blurry, out of focus, watermark, text, signature, low quality, artifacts"
+```
+
+### Reproducibility
+
+Always save the seed from the TextContent output if you want to reproduce a result:
+
+```
+Seed: 3847291045
+```
+
+Then pass it back: `seed=3847291045`
+
+---
+
+## 10. Known Limitations
+
+### ComfyUI must run locally
+
+The MCP server connects to `COMFYUI_URL` (default: `http://localhost:8188`). ComfyUI is a local application — it does not have a cloud API. You must start it before requesting image generation. The server returns a clear error message if ComfyUI is not reachable.
+
+### Model must be pre-loaded
+
+ComfyUI loads checkpoint models into VRAM on first use. The first generation with a model takes longer as VRAM is allocated (FLUX.1-schnell: ~8GB). Subsequent generations with the same model are faster.
+
+```bash
+# Verify model is installed before generation
+# → ask Lumen: "list available models in ComfyUI"
+```
+
+### AMD ROCm setup complexity
+
+AMD GPU support requires:
+1. ROCm drivers installed (`rocm-smi` working)
+2. PyTorch built with ROCm support (not the default CUDA build)
+3. `HSA_OVERRIDE_GFX_VERSION=11.0.0` for RX 7900 XTX (gfx1100)
+
+Without these, ComfyUI will fall back to CPU — very slow (minutes per image vs. ~8 seconds on RX 7900 XTX).
+
+Check GPU is being used:
+```bash
+# In another terminal while generating:
+watch -n 1 rocm-smi
+# VRAM usage should spike to ~8GB during generation
+```
+
+### Timeout on large images
+
+The default `COMFYUI_TIMEOUT=120` (2 minutes) may not be enough for:
+- Very large resolutions (2048×2048+)
+- High step counts (20+)
+- First generation loading a new model
+
+Increase via env var:
+```bash
+export COMFYUI_TIMEOUT=300  # 5 minutes
+```
+
+If `generate_image` returns a timeout error, the job may still be running in ComfyUI. Use `get_generation_status(prompt_id)` to check.
+
+### Ollama image gen is macOS-only (April 2026)
+
+Ollama launched experimental image generation in January 2026, but it is **macOS-only** as of April 2026. Linux support is announced as "coming soon." When Linux support arrives, the server can switch backends via `IMAGE_BACKEND=ollama` without changing any tool signatures.
+
+### ComfyUI history is ephemeral
+
+ComfyUI keeps generation history in memory — it is lost on restart. The `get_generation_status` tool will return `"not_found"` for old prompt IDs after a ComfyUI restart. The saved PNG file on disk persists regardless.

mcp/mcp-image-gen/src/server.py

@@ -40,7 +40,9 @@ class ComfyUIClient:
 
     async def queue_prompt(self, workflow: dict) -> str:
         """Submit a workflow to ComfyUI and return the prompt_id."""
-        payload = {"prompt": workflow}
+        # Strip internal metadata keys (e.g. "_meta") — they are not ComfyUI nodes
+        clean_workflow = {k: v for k, v in workflow.items() if not k.startswith("_")}
+        payload = {"prompt": clean_workflow}
         async with httpx.AsyncClient(timeout=30.0) as client:
             resp = await client.post(f"{self.base_url}/api/prompt", json=payload)
             resp.raise_for_status()
@@ -115,7 +117,8 @@ def build_flux_workflow(
     wf["27"]["inputs"]["height"] = height
     wf["13"]["inputs"]["steps"] = steps
     wf["13"]["inputs"]["seed"] = actual_seed
-    wf["30"]["inputs"]["ckpt_name"] = model
+    # Node 32 = UNETLoader (flux1-schnell.safetensors is UNet-only, not all-in-one checkpoint)
+    wf["32"]["inputs"]["unet_name"] = model
 
     # Attach the actual seed as metadata so callers can retrieve it
     wf["_meta"] = {"actual_seed": actual_seed}

mcp/mcp-image-gen/src/workflows/flux_schnell.json

@@ -2,7 +2,7 @@
   "6": {
     "class_type": "CLIPTextEncode",
     "inputs": {
-      "clip": ["30", 1],
+      "clip": ["30", 0],
       "text": "PROMPT_PLACEHOLDER"
     }
   },
@@ -10,7 +10,7 @@
     "class_type": "VAEDecode",
     "inputs": {
       "samples": ["13", 0],
-      "vae": ["30", 2]
+      "vae": ["31", 0]
     }
   },
   "9": {
@@ -26,7 +26,7 @@
       "cfg": 1.0,
       "denoise": 1.0,
       "latent_image": ["27", 0],
-      "model": ["30", 0],
+      "model": ["32", 0],
       "negative": ["33", 0],
       "positive": ["6", 0],
       "sampler_name": "euler",
@@ -44,15 +44,31 @@
     }
   },
   "30": {
-    "class_type": "CheckpointLoaderSimple",
+    "class_type": "DualCLIPLoader",
     "inputs": {
-      "ckpt_name": "flux1-schnell.safetensors"
+      "clip_name1": "t5xxl_fp8_e4m3fn.safetensors",
+      "clip_name2": "clip_l.safetensors",
+      "type": "flux",
+      "device": "default"
+    }
+  },
+  "31": {
+    "class_type": "VAELoader",
+    "inputs": {
+      "vae_name": "ae.safetensors"
+    }
+  },
+  "32": {
+    "class_type": "UNETLoader",
+    "inputs": {
+      "unet_name": "flux1-schnell.safetensors",
+      "weight_dtype": "fp8_e4m3fn"
     }
   },
   "33": {
     "class_type": "CLIPTextEncode",
     "inputs": {
-      "clip": ["30", 1],
+      "clip": ["30", 0],
       "text": "NEGATIVE_PLACEHOLDER"
     }
   }

mcp/mcp-image-gen/tests/test_server.py

@@ -28,7 +28,6 @@ COMFYUI_BASE = "http://test-comfyui:8188"
 # build_flux_workflow — pure function, no mocking needed
 # ---------------------------------------------------------------------------
 
-
 def test_build_flux_workflow_structure():
     """Verify build_flux_workflow returns a dict with correct node types."""
     wf = build_flux_workflow(
@@ -45,7 +44,9 @@ def test_build_flux_workflow_structure():
     assert wf["9"]["class_type"] == "SaveImage"
     assert wf["13"]["class_type"] == "KSampler"
     assert wf["27"]["class_type"] == "EmptySD3LatentImage"
-    assert wf["30"]["class_type"] == "CheckpointLoaderSimple"
+    assert wf["30"]["class_type"] == "DualCLIPLoader"
+    assert wf["31"]["class_type"] == "VAELoader"
+    assert wf["32"]["class_type"] == "UNETLoader"
     assert wf["33"]["class_type"] == "CLIPTextEncode"
 
 
@@ -66,7 +67,7 @@ def test_build_flux_workflow_params_injected():
     assert wf["27"]["inputs"]["height"] == 768
     assert wf["13"]["inputs"]["steps"] == 8
     assert wf["13"]["inputs"]["seed"] == 12345
-    assert wf["30"]["inputs"]["ckpt_name"] == "sdxl.safetensors"
+    assert wf["32"]["inputs"]["unet_name"] == "sdxl.safetensors"
 
 
 def test_negative_prompt_included():
@@ -103,7 +104,6 @@ def test_random_seed_generated():
 # list_available_models
 # ---------------------------------------------------------------------------
 
-
 @respx.mock
 @pytest.mark.asyncio
 async def test_list_available_models():
@@ -146,7 +146,6 @@ async def test_list_available_models_comfyui_offline():
 # get_generation_status
 # ---------------------------------------------------------------------------
 
-
 @respx.mock
 @pytest.mark.asyncio
 async def test_get_generation_status_pending(queue_with_pending):
@@ -191,7 +190,6 @@ async def test_get_generation_status_complete(queue_empty, mock_history_response
 # get_output_directory
 # ---------------------------------------------------------------------------
 
-
 def test_get_output_directory_default(monkeypatch):
     """No IMAGE_OUTPUT_DIR env var → returns expanded ~/Pictures/mcp-generated."""
     monkeypatch.delenv("IMAGE_OUTPUT_DIR", raising=False)
@@ -216,7 +214,6 @@ def test_get_output_directory_custom(monkeypatch, tmp_path):
 # generate_image
 # ---------------------------------------------------------------------------
 
-
 @respx.mock
 @pytest.mark.asyncio
 async def test_generate_image_success(
@@ -300,3 +297,253 @@ async def test_generate_image_timeout(monkeypatch, queue_with_pending):
     assert len(result) == 1
     assert "timed out" in result[0].text.lower()
     assert "test-uuid-1234" in result[0].text
+
+
+@respx.mock
+@pytest.mark.asyncio
+async def test_generate_image_empty_prompt(tmp_path, sample_image_bytes, mock_history_response, queue_empty, monkeypatch):
+    """Empty prompt → workflow has empty text in positive node, but generation succeeds."""
+    monkeypatch.setattr(server, "IMAGE_OUTPUT_DIR", str(tmp_path))
+
+    respx.post(f"{COMFYUI_BASE}/api/prompt").mock(
+        return_value=httpx.Response(200, json={"prompt_id": "test-empty-uuid"})
+    )
+    respx.get(f"{COMFYUI_BASE}/api/queue").mock(
+        return_value=httpx.Response(200, json=queue_empty)
+    )
+    mock_history_empty = {
+        "test-empty-uuid": {
+            "outputs": {
+                "9": {
+                    "images": [
+                        {
+                            "filename": "mcp-image-gen_00001_.png",
+                            "subfolder": "",
+                            "type": "output",
+                        }
+                    ]
+                }
+            },
+            "status": {"completed": True},
+        }
+    }
+    respx.get(f"{COMFYUI_BASE}/api/history/test-empty-uuid").mock(
+        return_value=httpx.Response(200, json=mock_history_empty)
+    )
+    respx.get(f"{COMFYUI_BASE}/api/view").mock(
+        return_value=httpx.Response(200, content=sample_image_bytes)
+    )
+
+    result = await generate_image(prompt="", output_dir=str(tmp_path))
+
+    assert len(result) == 2
+    text_content = result[0]
+    image_content = result[1]
+    assert "Generated:" in text_content.text
+    assert str(tmp_path) in text_content.text
+    # Verify workflow was built with empty prompt (indirectly via success)
+    assert image_content.mimeType == "image/png"
+
+
+@respx.mock
+@pytest.mark.asyncio
+async def test_generate_image_long_prompt(tmp_path, sample_image_bytes, mock_history_response, queue_empty, monkeypatch):
+    """Very long prompt → passed as-is to workflow without truncation."""
+    monkeypatch.setattr(server, "IMAGE_OUTPUT_DIR", str(tmp_path))
+
+    long_prompt = "a " + "very long descriptive prompt " * 50  # ~500 chars
+
+    respx.post(f"{COMFYUI_BASE}/api/prompt").mock(
+        return_value=httpx.Response(200, json={"prompt_id": "test-long-uuid"})
+    )
+    respx.get(f"{COMFYUI_BASE}/api/queue").mock(
+        return_value=httpx.Response(200, json=queue_empty)
+    )
+    mock_history_long = {
+        "test-long-uuid": {
+            "outputs": {
+                "9": {
+                    "images": [
+                        {
+                            "filename": "mcp-image-gen_00001_.png",
+                            "subfolder": "",
+                            "type": "output",
+                        }
+                    ]
+                }
+            },
+            "status": {"completed": True},
+        }
+    }
+    respx.get(f"{COMFYUI_BASE}/api/history/test-long-uuid").mock(
+        return_value=httpx.Response(200, json=mock_history_long)
+    )
+    respx.get(f"{COMFYUI_BASE}/api/view").mock(
+        return_value=httpx.Response(200, content=sample_image_bytes)
+    )
+
+    result = await generate_image(prompt=long_prompt, output_dir=str(tmp_path))
+
+    assert len(result) == 2
+    # Success implies long prompt was accepted (ComfyUI handles it)
+    saved_files = list(tmp_path.glob("*.png"))
+    assert len(saved_files) == 1
+
+
+@respx.mock
+@pytest.mark.asyncio
+async def test_generate_image_invalid_model(tmp_path, monkeypatch):
+    """Invalid model → ComfyUI /prompt returns 500 or 404, tool returns error TextContent."""
+    monkeypatch.setattr(server, "IMAGE_OUTPUT_DIR", str(tmp_path))
+
+    respx.post(f"{COMFYUI_BASE}/api/prompt").mock(
+        return_value=httpx.Response(404, json={"error": "Model not found"})
+    )
+
+    result = await generate_image(
+        prompt="a cat",
+        model="nonexistent-model.safetensors",
+        output_dir=str(tmp_path)
+    )
+
+    assert len(result) == 1
+    assert "404" in result[0].text
+    assert "Model not found" in result[0].text
+    # No file saved
+    saved_files = list(tmp_path.glob("*.png"))
+    assert len(saved_files) == 0
+
+
+@respx.mock
+@pytest.mark.asyncio
+async def test_generate_image_custom_output_dir(tmp_path, sample_image_bytes, mock_history_response, queue_empty, monkeypatch):
+    """Custom output_dir → image saved there, path reflects it."""
+    custom_dir = tmp_path / "custom"
+    monkeypatch.setattr(server, "IMAGE_OUTPUT_DIR", str(tmp_path))  # Base for default
+
+    respx.post(f"{COMFYUI_BASE}/api/prompt").mock(
+        return_value=httpx.Response(200, json={"prompt_id": "test-custom-uuid"})
+    )
+    respx.get(f"{COMFYUI_BASE}/api/queue").mock(
+        return_value=httpx.Response(200, json=queue_empty)
+    )
+    mock_history_custom = {
+        "test-custom-uuid": {
+            "outputs": {
+                "9": {
+                    "images": [
+                        {
+                            "filename": "mcp-image-gen_00001_.png",
+                            "subfolder": "",
+                            "type": "output",
+                        }
+                    ]
+                }
+            },
+            "status": {"completed": True},
+        }
+    }
+    respx.get(f"{COMFYUI_BASE}/api/history/test-custom-uuid").mock(
+        return_value=httpx.Response(200, json=mock_history_custom)
+    )
+    respx.get(f"{COMFYUI_BASE}/api/view").mock(
+        return_value=httpx.Response(200, content=sample_image_bytes)
+    )
+
+    result = await generate_image(
+        prompt="a dog",
+        output_dir=str(custom_dir),
+    )
+
+    assert len(result) == 2
+    text_content = result[0]
+    assert str(custom_dir) in text_content.text
+    # Directory was created
+    assert custom_dir.exists()
+    saved_files = list(custom_dir.glob("*.png"))
+    assert len(saved_files) == 1
+
+
+@respx.mock
+@pytest.mark.asyncio
+async def test_generate_image_random_seed_variance(tmp_path, sample_image_bytes, mock_history_response, queue_empty, monkeypatch):
+    """seed=-1 → different actual_seed each call, reflected in filename."""
+    monkeypatch.setattr(server, "IMAGE_OUTPUT_DIR", str(tmp_path))
+
+    # First generation
+    respx.post(f"{COMFYUI_BASE}/api/prompt").mock(
+        return_value=httpx.Response(200, json={"prompt_id": "seed1-uuid"})
+    )
+    respx.get(f"{COMFYUI_BASE}/api/queue").mock(
+        return_value=httpx.Response(200, json=queue_empty)
+    )
+    mock_history_seed1 = {
+        "seed1-uuid": {
+            "outputs": {
+                "9": {
+                    "images": [
+                        {
+                            "filename": "mcp-image-gen_00001_.png",
+                            "subfolder": "",
+                            "type": "output",
+                        }
+                    ]
+                }
+            },
+            "status": {"completed": True},
+        }
+    }
+    respx.get(f"{COMFYUI_BASE}/api/history/seed1-uuid").mock(
+        return_value=httpx.Response(200, json=mock_history_seed1)
+    )
+    respx.get(f"{COMFYUI_BASE}/api/view").mock(
+        return_value=httpx.Response(200, content=sample_image_bytes)
+    )
+
+    result1 = await generate_image(prompt="cat", seed=-1, output_dir=str(tmp_path))
+    seed1 = [line for line in result1[0].text.split("\n") if "Seed:" in line][0].split(": ")[1]
+    filename1 = Path(result1[0].text.split("Generated: ")[1].split("\n")[0]).name
+    assert "Seed:" in result1[0].text
+    assert int(seed1) != 0  # Not default
+
+    # Reset mocks for second call
+    respx.reset()
+    respx.post(f"{COMFYUI_BASE}/api/prompt").mock(
+        return_value=httpx.Response(200, json={"prompt_id": "seed2-uuid"})
+    )
+    respx.get(f"{COMFYUI_BASE}/api/queue").mock(
+        return_value=httpx.Response(200, json=queue_empty)
+    )
+    mock_history_seed2 = {
+        "seed2-uuid": {
+            "outputs": {
+                "9": {
+                    "images": [
+                        {
+                            "filename": "mcp-image-gen_00002_.png",
+                            "subfolder": "",
+                            "type": "output",
+                        }
+                    ]
+                }
+            },
+            "status": {"completed": True},
+        }
+    }
+    respx.get(f"{COMFYUI_BASE}/api/history/seed2-uuid").mock(
+        return_value=httpx.Response(200, json=mock_history_seed2)
+    )
+    respx.get(f"{COMFYUI_BASE}/api/view").mock(
+        return_value=httpx.Response(200, content=sample_image_bytes)
+    )
+
+    result2 = await generate_image(prompt="cat", seed=-1, output_dir=str(tmp_path))
+    seed2 = [line for line in result2[0].text.split("\n") if "Seed:" in line][0].split(": ")[1]
+    filename2 = Path(result2[0].text.split("Generated: ")[1].split("\n")[0]).name
+
+    # Different seeds and filenames
+    assert seed1 != seed2
+    assert filename1 != filename2
+    # Both saved
+    saved_files = list(tmp_path.glob("*.png"))
+    assert len(saved_files) == 2