Add inference-image MCP server (accessible image generation) + proper favicon sizes

- mcp/image_server.py: generate_image tool via the co-op gateway; per-member
  key injected through customUserVars (IC_API_KEY) -> env; verified standalone
  (handshake, tools/list, real 1024x1024 + 1536x1024 images, error paths)
- librechat.yaml: register inference-image with startup:false (per-user
  credentials), 120s timeout
- assets/: real favicon sizes (16/32/192), opaque apple-touch 180, padded
  maskable 512

README.md

@@ -17,7 +17,7 @@ from this directory.
 | App config | `/app/data/librechat.yaml` (volume) | yes | `librechat.yaml` |
 | Env vars (env-file wins!) | `/app/data/env` (volume, sourced LAST by start.sh) | yes | changed lines documented below |
 | Env vars (Cloudron-managed) | `cloudron env` (process env) | yes | documented below |
-| MCP docs server | `/app/data/mcp/server.py` (volume) | yes | `mcp/server.py` |
+| MCP servers | `/app/data/mcp/server.py` + `/app/data/mcp/image_server.py` (volume) | yes | `mcp/` |
 | Brand assets (logo, favicons) | baked into image overlay | no — rebuilt per image | `assets/` |
 | CloudronManifest.json | repo only (used by `cloudron update`) | n/a | byte-identical copy of the OFFICIAL appstore manifest |
 
@@ -106,13 +106,21 @@ Set via `cloudron env set` (process env):
   keys, 8 chat modelSpecs (emoji convention shared with models.md).
   `regolo/qwen-image` is deliberately NOT in the chat picker (image models
   are not chat models; chat-calling it was the source of the old 500s).
-- **Image generation**: LibreChat exposes image gen through an **Agent with
+- **Image generation — two paths**:
-  the "OpenAI Image Tools" tool**, not a chat-picker button. Members create
+  1. **MCP tool (easiest)**: `inference-image` server (tool `generate_image`)
-  an agent, add the tool, and paste their co-op key on first use.
+     — members select it in the composer's **+** palette and just ask. No
-- **MCP**: `inference-docs` (stdio, `/app/data/mcp/server.py`, tool
+     agent setup. First use: save your co-op key in the MCP panel
-  `search_docs`) is admin-configured; members may add their own via the MCP
+     (customUserVars `IC_API_KEY`, injected per-user into the server process
-  Settings panel (the "bring your own tools" pattern). Permissions seeded via
+     env; `startup: false` because the key only exists per-user after they
-  `interface.mcpServers` (use/create true, share false).
+     configure it). Bills to the member's own key (~1–2¢/image at 1024²).
+  2. **Agent tool**: LibreChat's built-in "OpenAI Image Tools" (env
+     `IMAGE_GEN_OAI_BASEURL`/`IMAGE_GEN_OAI_MODEL` → gateway; no admin key,
+     members paste their own). Note LibreChat v0.8.8 renders the generated
+     image only inside the agent conversation.
+- **MCP**: `inference-docs` (tool `search_docs`, starts at boot) and
+  `inference-image` (tool `generate_image`, per-user startup). Members may
+  add their own via the MCP Settings panel ("bring your own tools").
+  Permissions seeded via `interface.mcpServers` (use/create true, share false).
 - **Speech**: TTS `tinfoil/voxtral-tts` (voice `casual_female`), STT
   `tinfoil/whisper-large-v3-turbo`, both via the gateway with the capped key.
 - **Web search**: SearXNG (self-hosted) + keyless Keenable scraper, no reranker.
@@ -124,10 +132,10 @@ Set via `cloudron env set` (process env):
 
 ## Known state / limitations
 
-- `patches/`, `logo.png`, `logo-coop.png`, `logo.svg` (repo root) are legacy
-  leftovers from the abandoned 2026-10-18 hand-built-image attempt — unused,
-  pending cleanup.
 - The `[schedules]` Redis guard error is resolved by
   `SCHEDULES_SINGLE_PROCESS=true` (above).
 - The `RAG API not reachable` warning is expected (no rag_api addon); it only
   affects LibreChat's own file-RAG, which we do not use.
+- The `inference-image` MCP server logs `Tools: undefined` at boot — expected:
+  with `startup: false` it registers but only spawns per-user after a member
+  saves their key and reinitializes in the MCP panel.

librechat.yaml

@@ -160,11 +160,28 @@ webSearch:
   scraperProvider: "keenable"
   rerankerType: "none"
 
-# MCP servers (admin-configured). The docs server is the same stdlib-only
+# MCP servers (admin-configured, on the persistent volume so they survive
-# Python server Open WebUI uses; it lives on the persistent volume at
+# image updates). Members opt in per conversation via the composer's +
-# /app/data/mcp/server.py so it survives image updates.
+# palette; each user gets their own isolated server connection.
 mcpServers:
   inference-docs:
     command: /app/code/venv/bin/python3
     args: ["/app/data/mcp/server.py"]
     timeout: 60000
+  # Accessible image generation: one click in the + palette, no agent setup.
+  # Each member's own co-op key is injected per-user via customUserVars →
+  # env, so images bill to the member's own balance (chat-endpoint model).
+  inference-image:
+    command: /app/code/venv/bin/python3
+    args: ["/app/data/mcp/image_server.py"]
+    timeout: 120000
+    # Per-user credentials aren't available at boot; each member initializes
+    # their own connection after saving their key in the MCP panel.
+    startup: false
+    customUserVars:
+      IC_API_KEY:
+        title: "Inference Co-op API key"
+        description: "Your co-op API key — create one in the <a href='https://dashboard.inference.coop' target='_blank'>Member Dashboard</a>. Used for image generation; bills to your own balance."
+        sensitive: true
+    env:
+      IC_API_KEY: "{{IC_API_KEY}}"

mcp/image_server.py

@@ -0,0 +1,210 @@
+#!/usr/bin/env python3
+"""Inference Cooperative image-generation MCP server (stdio, stdlib-only).
+
+Exposes a `generate_image` tool that calls the co-op gateway
+(`regolo/qwen-image` via POST /v1/images/generations). Each member's own
+co-op API key is injected per-user by LibreChat as the IC_API_KEY env var
+(`customUserVars` + `env` in librechat.yaml), so usage bills to the member's
+own LiteLLM team budget — the same model as the chat endpoint.
+
+Design notes (mirrors mcp/server.py, the docs server):
+- stdio transport: newline-delimited JSON-RPC 2.0; logging to stderr only.
+- No third-party dependencies. The image is returned inline as MCP image
+  content (base64 PNG), which LibreChat renders directly in the chat.
+"""
+
+import base64
+import json
+import os
+import sys
+import urllib.error
+import urllib.request
+
+GATEWAY_IMAGES_URL = "https://gateway.inference.coop/v1/images/generations"
+MODEL = "regolo/qwen-image"
+REQUEST_TIMEOUT = 90
+ALLOWED_SIZES = ("1024x1024", "1536x1024", "1024x1536")
+DEFAULT_SIZE = "1024x1024"
+DASHBOARD_URL = "https://dashboard.inference.coop"
+
+SERVER_NAME = "inference-cooperative-image"
+SERVER_VERSION = "1.0.0"
+
+
+def log(msg):
+    sys.stderr.write(f"[inference-image-mcp] {msg}\n")
+    sys.stderr.flush()
+
+
+def tool_generate_image(args):
+    key = (os.environ.get("IC_API_KEY") or "").strip()
+    if not key:
+        return {
+            "content": [{
+                "type": "text",
+                "text": (
+                    "No Inference Co-op API key configured yet. Open the MCP panel "
+                    f"(inference-image) and save your key — create one at {DASHBOARD_URL}. "
+                    "Your key is only visible to you, and image generation bills to your "
+                    "own co-op balance."
+                ),
+            }],
+            "isError": True,
+        }
+
+    prompt = str(args.get("prompt") or "").strip()
+    if not prompt:
+        return {"content": [{"type": "text", "text": "Provide a prompt for the image."}]}
+
+    size = str(args.get("size") or DEFAULT_SIZE).strip()
+    if size not in ALLOWED_SIZES:
+        return {
+            "content": [{
+                "type": "text",
+                "text": f"Unsupported size '{size}'. Use one of: {', '.join(ALLOWED_SIZES)}.",
+            }],
+            "isError": True,
+        }
+
+    body = json.dumps({"model": MODEL, "prompt": prompt, "n": 1, "size": size}).encode()
+    req = urllib.request.Request(
+        GATEWAY_IMAGES_URL,
+        data=body,
+        headers={"Authorization": f"Bearer {key}", "Content-Type": "application/json"},
+    )
+    try:
+        with urllib.request.urlopen(req, timeout=REQUEST_TIMEOUT) as resp:
+            data = json.loads(resp.read().decode("utf-8"))
+    except urllib.error.HTTPError as exc:
+        detail = exc.read().decode("utf-8", "replace")[:400]
+        log(f"gateway HTTP {exc.code}: {detail}")
+        hint = "Check that your co-op API key is valid and has balance."
+        if exc.code == 401:
+            hint = f"Your key was rejected — re-save it in the MCP panel (create one at {DASHBOARD_URL})."
+        elif exc.code == 429:
+            hint = "Budget or rate limit reached on your key."
+        return {
+            "content": [{"type": "text", "text": f"Image generation failed (HTTP {exc.code}). {hint}"}],
+            "isError": True,
+        }
+    except Exception as exc:  # noqa: BLE001 - report transport errors in-band
+        log(f"gateway error: {exc}")
+        return {
+            "content": [{"type": "text", "text": f"Image generation failed: {exc}"}],
+            "isError": True,
+        }
+
+    items = (data or {}).get("data") or []
+    if not items:
+        return {"content": [{"type": "text", "text": "The gateway returned no image."}], "isError": True}
+
+    item = items[0]
+    content = []
+    if item.get("b64_json"):
+        content.append({"type": "image", "data": item["b64_json"], "mimeType": "image/png"})
+    notes = [f"Generated with {MODEL} ({size}) — billed to your co-op key."]
+    if item.get("revised_prompt"):
+        notes.append(f"Prompt: {item['revised_prompt']}")
+    if not item.get("b64_json") and item.get("url"):
+        notes.append(f"Image URL: {item['url']}")
+    content.append({"type": "text", "text": " ".join(notes)})
+    return {"content": content}
+
+
+TOOLS = [
+    {
+        "name": "generate_image",
+        "description": (
+            "Generate an image from a text prompt (regolo/qwen-image via the Inference "
+            "Co-op gateway, EU renewable-powered). Costs roughly 1-2 cents per image, "
+            "billed to your own co-op API key."
+        ),
+        "inputSchema": {
+            "type": "object",
+            "properties": {
+                "prompt": {"type": "string", "description": "What to draw"},
+                "size": {
+                    "type": "string",
+                    "enum": list(ALLOWED_SIZES),
+                    "description": "Image size (default 1024x1024)",
+                },
+            },
+            "required": ["prompt"],
+        },
+    }
+]
+
+TOOLS_HANDLERS = {"generate_image": tool_generate_image}
+
+
+def handle(msg):
+    method = msg.get("method")
+    msg_id = msg.get("id")
+
+    if method == "initialize":
+        return {
+            "jsonrpc": "2.0",
+            "id": msg_id,
+            "result": {
+                "protocolVersion": "2024-11-05",
+                "capabilities": {"tools": {}},
+                "serverInfo": {"name": SERVER_NAME, "version": SERVER_VERSION},
+            },
+        }
+    if method == "ping":
+        return {"jsonrpc": "2.0", "id": msg_id, "result": {}}
+    if method == "tools/list":
+        return {"jsonrpc": "2.0", "id": msg_id, "result": {"tools": TOOLS}}
+    if method == "tools/call":
+        params = msg.get("params") or {}
+        name = params.get("name")
+        handler = TOOLS_HANDLERS.get(name)
+        if not handler:
+            return {
+                "jsonrpc": "2.0",
+                "id": msg_id,
+                "error": {"code": -32602, "message": f"Unknown tool: {name}"},
+            }
+        try:
+            result = handler(params.get("arguments") or {})
+            return {"jsonrpc": "2.0", "id": msg_id, "result": result}
+        except Exception as exc:  # noqa: BLE001 - report tool errors in-band
+            log(f"tool error: {exc}")
+            return {
+                "jsonrpc": "2.0",
+                "id": msg_id,
+                "result": {"content": [{"type": "text", "text": f"Tool failed: {exc}"}], "isError": True},
+            }
+    if method and method.startswith("notifications/"):
+        return None
+    if msg_id is not None:
+        return {"jsonrpc": "2.0", "id": msg_id, "error": {"code": -32601, "message": "Method not found"}}
+    return None
+
+
+def main():
+    log("starting (stdio)")
+    for line in sys.stdin:
+        line = line.strip()
+        if not line:
+            continue
+        try:
+            msg = json.loads(line)
+        except json.JSONDecodeError as exc:
+            log(f"bad json: {exc}")
+            continue
+        try:
+            resp = handle(msg)
+        except Exception as exc:  # noqa: BLE001 - never crash the protocol loop
+            log(f"handler crash: {exc}")
+            if msg.get("id") is not None:
+                resp = {"jsonrpc": "2.0", "id": msg["id"], "error": {"code": -32603, "message": "Internal error"}}
+            else:
+                resp = None
+        if resp is not None:
+            sys.stdout.write(json.dumps(resp) + "\n")
+            sys.stdout.flush()
+
+
+if __name__ == "__main__":
+    main()