feat(plugins): pluggable image_gen backends + OpenAI provider (#13799)

* feat(plugins): pluggable image_gen backends + OpenAI provider

Adds a ImageGenProvider ABC so image generation backends register as
bundled plugins under `plugins/image_gen/<name>/`. The plugin scanner
gains three primitives to make this work generically:

- `kind:` manifest field (`standalone` | `backend` | `exclusive`).
  Bundled `kind: backend` plugins auto-load — no `plugins.enabled`
  incantation. User-installed backends stay opt-in.
- Path-derived keys: `plugins/image_gen/openai/` gets key
  `image_gen/openai`, so a future `tts/openai` cannot collide.
- Depth-2 recursion into category namespaces (parent dirs without a
  `plugin.yaml` of their own).

Includes `OpenAIImageGenProvider` as the first consumer (gpt-image-1.5
default, plus gpt-image-1, gpt-image-1-mini, DALL-E 3/2). Base64
responses save to `$HERMES_HOME/cache/images/`; URL responses pass
through.

FAL stays in-tree for this PR — a follow-up ports it into
`plugins/image_gen/fal/` so the in-tree `image_generation_tool.py`
slims down. The dispatch shim in `_handle_image_generate` only fires
when `image_gen.provider` is explicitly set to a non-FAL value, so
existing FAL setups are untouched.

- 41 unit tests (scanner recursion, kind parsing, gate logic,
  registry, OpenAI payload shapes)
- E2E smoke verified: bundled plugin autoloads, registers, and
  `_handle_image_generate` routes to OpenAI when configured

* fix(image_gen/openai): don't send response_format to gpt-image-*

The live API rejects it: 'Unknown parameter: response_format'
(verified 2026-04-21 with gpt-image-1.5). gpt-image-* models return
b64_json unconditionally, so the parameter was both unnecessary and
actively broken.

* feat(image_gen/openai): gpt-image-2 only, drop legacy catalog

gpt-image-2 is the latest/best OpenAI image model (released 2026-04-21)
and there's no reason to expose the older gpt-image-1.5 / gpt-image-1 /
dall-e-3 / dall-e-2 alongside it — slower, lower quality, or awkward
(dall-e-2 squares only). Trim the catalog down to a single model.

Live-verified end-to-end: landscape 1536x1024 render of a Moog-style
synth matches prompt exactly, 2.4MB PNG saved to cache.

* feat(image_gen/openai): expose gpt-image-2 as three quality tiers

Users pick speed/fidelity via the normal model picker instead of a
hidden quality knob. All three tier IDs resolve to the single underlying
gpt-image-2 API model with a different quality parameter:

  gpt-image-2-low     ~15s   fast iteration
  gpt-image-2-medium  ~40s   default
  gpt-image-2-high    ~2min  highest fidelity

Live-measured on OpenAI's API today: 15.4s / 40.8s / 116.9s for the
same 1024x1024 prompt.

Config:
  image_gen.openai.model: gpt-image-2-high
  # or
  image_gen.model: gpt-image-2-low
  # or env var for scripts/tests
  OPENAI_IMAGE_MODEL=gpt-image-2-medium

Live-verified end-to-end with the low tier: 18.8s landscape render of a
golden retriever in wildflowers, vision-confirmed exact match.

* feat(tools_config): plugin image_gen providers inject themselves into picker

'hermes tools' → Image Generation now shows plugin-registered backends
alongside Nous Subscription and FAL.ai without tools_config.py needing
to know about them. OpenAI appears as a third option today; future
backends appear automatically as they're added.

Mechanism:
- ImageGenProvider gains an optional get_setup_schema() hook
  (name, badge, tag, env_vars). Default derived from display_name.
- tools_config._plugin_image_gen_providers() pulls the schemas from
  every registered non-FAL plugin provider.
- _visible_providers() appends those rows when rendering the Image
  Generation category.
- _configure_provider() handles the new image_gen_plugin_name marker:
  writes image_gen.provider and routes to the plugin's list_models()
  catalog for the model picker.
- _toolset_needs_configuration_prompt('image_gen') stops demanding a
  FAL key when any plugin provider reports is_available().

FAL is skipped in the plugin path because it already has hardcoded
TOOL_CATEGORIES rows — when it gets ported to a plugin in a follow-up
PR the hardcoded rows go away and it surfaces through the same path
as OpenAI.

Verified live: picker shows Nous Subscription / FAL.ai / OpenAI.
Picking OpenAI prompts for OPENAI_API_KEY, then shows the
gpt-image-2-low/medium/high model picker sourced from the plugin.

397 tests pass across plugins/, tools_config, registry, and picker.

* fix(image_gen): close final gaps for plugin-backend parity with FAL

Two small places that still hardcoded FAL:

- hermes_cli/setup.py status line: an OpenAI-only setup showed
  'Image Generation: missing FAL_KEY'. Now probes plugin providers
  and reports '(OpenAI)' when one is_available() — or falls back to
  'missing FAL_KEY or OPENAI_API_KEY' if nothing is configured.

- image_generate tool schema description: said 'using FAL.ai, default
  FLUX 2 Klein 9B'. Rewrote provider-neutral — 'backend and model are
  user-configured' — and notes the 'image' field can be a URL or an
  absolute path, which the gateway delivers either way via
  extract_local_files().

agent/image_gen_provider.py

@@ -0,0 +1,242 @@
+"""
+Image Generation Provider ABC
+=============================
+
+Defines the pluggable-backend interface for image generation. Providers register
+instances via ``PluginContext.register_image_gen_provider()``; the active one
+(selected via ``image_gen.provider`` in ``config.yaml``) services every
+``image_generate`` tool call.
+
+Providers live in ``<repo>/plugins/image_gen/<name>/`` (built-in, auto-loaded
+as ``kind: backend``) or ``~/.hermes/plugins/image_gen/<name>/`` (user, opt-in
+via ``plugins.enabled``).
+
+Response shape
+--------------
+All providers return a dict that :func:`success_response` / :func:`error_response`
+produce. The tool wrapper JSON-serializes it. Keys:
+
+    success        bool
+    image          str | None       URL or absolute file path
+    model          str              provider-specific model identifier
+    prompt         str              echoed prompt
+    aspect_ratio   str              "landscape" | "square" | "portrait"
+    provider       str              provider name (for diagnostics)
+    error          str              only when success=False
+    error_type     str              only when success=False
+"""
+
+from __future__ import annotations
+
+import abc
+import base64
+import datetime
+import logging
+import uuid
+from pathlib import Path
+from typing import Any, Dict, List, Optional, Tuple
+
+logger = logging.getLogger(__name__)
+
+
+VALID_ASPECT_RATIOS: Tuple[str, ...] = ("landscape", "square", "portrait")
+DEFAULT_ASPECT_RATIO = "landscape"
+
+
+# ---------------------------------------------------------------------------
+# ABC
+# ---------------------------------------------------------------------------
+
+
+class ImageGenProvider(abc.ABC):
+    """Abstract base class for an image generation backend.
+
+    Subclasses must implement :meth:`generate`. Everything else has sane
+    defaults — override only what your provider needs.
+    """
+
+    @property
+    @abc.abstractmethod
+    def name(self) -> str:
+        """Stable short identifier used in ``image_gen.provider`` config.
+
+        Lowercase, no spaces. Examples: ``fal``, ``openai``, ``replicate``.
+        """
+
+    @property
+    def display_name(self) -> str:
+        """Human-readable label shown in ``hermes tools``. Defaults to ``name.title()``."""
+        return self.name.title()
+
+    def is_available(self) -> bool:
+        """Return True when this provider can service calls.
+
+        Typically checks for a required API key. Default: True
+        (providers with no external dependencies are always available).
+        """
+        return True
+
+    def list_models(self) -> List[Dict[str, Any]]:
+        """Return catalog entries for ``hermes tools`` model picker.
+
+        Each entry::
+
+            {
+                "id": "gpt-image-1.5",               # required
+                "display": "GPT Image 1.5",          # optional; defaults to id
+                "speed": "~10s",                     # optional
+                "strengths": "...",                  # optional
+                "price": "$...",                     # optional
+            }
+
+        Default: empty list (provider has no user-selectable models).
+        """
+        return []
+
+    def get_setup_schema(self) -> Dict[str, Any]:
+        """Return provider metadata for the ``hermes tools`` picker.
+
+        Used by ``tools_config.py`` to inject this provider as a row in
+        the Image Generation provider list. Shape::
+
+            {
+                "name": "OpenAI",                     # picker label
+                "badge": "paid",                      # optional short tag
+                "tag": "One-line description...",     # optional subtitle
+                "env_vars": [                         # keys to prompt for
+                    {"key": "OPENAI_API_KEY",
+                     "prompt": "OpenAI API key",
+                     "url": "https://platform.openai.com/api-keys"},
+                ],
+            }
+
+        Default: minimal entry derived from ``display_name``. Override to
+        expose API key prompts and custom badges.
+        """
+        return {
+            "name": self.display_name,
+            "badge": "",
+            "tag": "",
+            "env_vars": [],
+        }
+
+    def default_model(self) -> Optional[str]:
+        """Return the default model id, or None if not applicable."""
+        models = self.list_models()
+        if models:
+            return models[0].get("id")
+        return None
+
+    @abc.abstractmethod
+    def generate(
+        self,
+        prompt: str,
+        aspect_ratio: str = DEFAULT_ASPECT_RATIO,
+        **kwargs: Any,
+    ) -> Dict[str, Any]:
+        """Generate an image.
+
+        Implementations should return the dict from :func:`success_response`
+        or :func:`error_response`. ``kwargs`` may contain forward-compat
+        parameters future versions of the schema will expose — implementations
+        should ignore unknown keys.
+        """
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+
+def resolve_aspect_ratio(value: Optional[str]) -> str:
+    """Clamp an aspect_ratio value to the valid set, defaulting to landscape.
+
+    Invalid values are coerced rather than rejected so the tool surface is
+    forgiving of agent mistakes.
+    """
+    if not isinstance(value, str):
+        return DEFAULT_ASPECT_RATIO
+    v = value.strip().lower()
+    if v in VALID_ASPECT_RATIOS:
+        return v
+    return DEFAULT_ASPECT_RATIO
+
+
+def _images_cache_dir() -> Path:
+    """Return ``$HERMES_HOME/cache/images/``, creating parents as needed."""
+    from hermes_constants import get_hermes_home
+
+    path = get_hermes_home() / "cache" / "images"
+    path.mkdir(parents=True, exist_ok=True)
+    return path
+
+
+def save_b64_image(
+    b64_data: str,
+    *,
+    prefix: str = "image",
+    extension: str = "png",
+) -> Path:
+    """Decode base64 image data and write it under ``$HERMES_HOME/cache/images/``.
+
+    Returns the absolute :class:`Path` to the saved file.
+
+    Filename format: ``<prefix>_<YYYYMMDD_HHMMSS>_<short-uuid>.<ext>``.
+    """
+    raw = base64.b64decode(b64_data)
+    ts = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
+    short = uuid.uuid4().hex[:8]
+    path = _images_cache_dir() / f"{prefix}_{ts}_{short}.{extension}"
+    path.write_bytes(raw)
+    return path
+
+
+def success_response(
+    *,
+    image: str,
+    model: str,
+    prompt: str,
+    aspect_ratio: str,
+    provider: str,
+    extra: Optional[Dict[str, Any]] = None,
+) -> Dict[str, Any]:
+    """Build a uniform success response dict.
+
+    ``image`` may be an HTTP URL or an absolute filesystem path (for b64
+    providers like OpenAI). Callers that need to pass through additional
+    backend-specific fields can supply ``extra``.
+    """
+    payload: Dict[str, Any] = {
+        "success": True,
+        "image": image,
+        "model": model,
+        "prompt": prompt,
+        "aspect_ratio": aspect_ratio,
+        "provider": provider,
+    }
+    if extra:
+        for k, v in extra.items():
+            payload.setdefault(k, v)
+    return payload
+
+
+def error_response(
+    *,
+    error: str,
+    error_type: str = "provider_error",
+    provider: str = "",
+    model: str = "",
+    prompt: str = "",
+    aspect_ratio: str = DEFAULT_ASPECT_RATIO,
+) -> Dict[str, Any]:
+    """Build a uniform error response dict."""
+    return {
+        "success": False,
+        "image": None,
+        "error": error,
+        "error_type": error_type,
+        "model": model,
+        "prompt": prompt,
+        "aspect_ratio": aspect_ratio,
+        "provider": provider,
+    }

agent/image_gen_registry.py

@@ -0,0 +1,120 @@
+"""
+Image Generation Provider Registry
+==================================
+
+Central map of registered providers. Populated by plugins at import-time via
+``PluginContext.register_image_gen_provider()``; consumed by the
+``image_generate`` tool to dispatch each call to the active backend.
+
+Active selection
+----------------
+The active provider is chosen by ``image_gen.provider`` in ``config.yaml``.
+If unset, :func:`get_active_provider` applies fallback logic:
+
+1. If exactly one provider is registered, use it.
+2. Otherwise if a provider named ``fal`` is registered, use it (legacy
+   default — matches pre-plugin behavior).
+3. Otherwise return ``None`` (the tool surfaces a helpful error pointing
+   the user at ``hermes tools``).
+"""
+
+from __future__ import annotations
+
+import logging
+import threading
+from typing import Dict, List, Optional
+
+from agent.image_gen_provider import ImageGenProvider
+
+logger = logging.getLogger(__name__)
+
+
+_providers: Dict[str, ImageGenProvider] = {}
+_lock = threading.Lock()
+
+
+def register_provider(provider: ImageGenProvider) -> None:
+    """Register an image generation provider.
+
+    Re-registration (same ``name``) overwrites the previous entry and logs
+    a debug message — this makes hot-reload scenarios (tests, dev loops)
+    behave predictably.
+    """
+    if not isinstance(provider, ImageGenProvider):
+        raise TypeError(
+            f"register_provider() expects an ImageGenProvider instance, "
+            f"got {type(provider).__name__}"
+        )
+    name = provider.name
+    if not isinstance(name, str) or not name.strip():
+        raise ValueError("Image gen provider .name must be a non-empty string")
+    with _lock:
+        existing = _providers.get(name)
+        _providers[name] = provider
+    if existing is not None:
+        logger.debug("Image gen provider '%s' re-registered (was %r)", name, type(existing).__name__)
+    else:
+        logger.debug("Registered image gen provider '%s' (%s)", name, type(provider).__name__)
+
+
+def list_providers() -> List[ImageGenProvider]:
+    """Return all registered providers, sorted by name."""
+    with _lock:
+        items = list(_providers.values())
+    return sorted(items, key=lambda p: p.name)
+
+
+def get_provider(name: str) -> Optional[ImageGenProvider]:
+    """Return the provider registered under *name*, or None."""
+    if not isinstance(name, str):
+        return None
+    with _lock:
+        return _providers.get(name.strip())
+
+
+def get_active_provider() -> Optional[ImageGenProvider]:
+    """Resolve the currently-active provider.
+
+    Reads ``image_gen.provider`` from config.yaml; falls back per the
+    module docstring.
+    """
+    configured: Optional[str] = None
+    try:
+        from hermes_cli.config import load_config
+
+        cfg = load_config()
+        section = cfg.get("image_gen") if isinstance(cfg, dict) else None
+        if isinstance(section, dict):
+            raw = section.get("provider")
+            if isinstance(raw, str) and raw.strip():
+                configured = raw.strip()
+    except Exception as exc:
+        logger.debug("Could not read image_gen.provider from config: %s", exc)
+
+    with _lock:
+        snapshot = dict(_providers)
+
+    if configured:
+        provider = snapshot.get(configured)
+        if provider is not None:
+            return provider
+        logger.debug(
+            "image_gen.provider='%s' configured but not registered; falling back",
+            configured,
+        )
+
+    # Fallback: single-provider case
+    if len(snapshot) == 1:
+        return next(iter(snapshot.values()))
+
+    # Fallback: prefer legacy FAL for backward compat
+    if "fal" in snapshot:
+        return snapshot["fal"]
+
+    return None
+
+
+def _reset_for_tests() -> None:
+    """Clear the registry. **Test-only.**"""
+    with _lock:
+        _providers.clear()

hermes_cli/plugins.py

@@ -133,6 +133,9 @@ def _get_enabled_plugins() -> Optional[set]:
 # Data classes
 # ---------------------------------------------------------------------------
 
+_VALID_PLUGIN_KINDS: Set[str] = {"standalone", "backend", "exclusive"}
+
+
 @dataclass
 class PluginManifest:
     """Parsed representation of a plugin.yaml manifest."""
@@ -146,6 +149,23 @@ class PluginManifest:
     provides_hooks: List[str] = field(default_factory=list)
     source: str = ""        # "user", "project", or "entrypoint"
     path: Optional[str] = None
+    # Plugin kind — see plugins.py module docstring for semantics.
+    # ``standalone`` (default): hooks/tools of its own; opt-in via
+    #                           ``plugins.enabled``.
+    # ``backend``: pluggable backend for an existing core tool (e.g.
+    #              image_gen). Built-in (bundled) backends auto-load;
+    #              user-installed still gated by ``plugins.enabled``.
+    # ``exclusive``: category with exactly one active provider (memory).
+    #              Selection via ``<category>.provider`` config key; the
+    #              category's own discovery system handles loading and the
+    #              general scanner skips these.
+    kind: str = "standalone"
+    # Registry key — path-derived, used by ``plugins.enabled``/``disabled``
+    # lookups and by ``hermes plugins list``. For a flat plugin at
+    # ``plugins/disk-cleanup/`` the key is ``disk-cleanup``; for a nested
+    # category plugin at ``plugins/image_gen/openai/`` the key is
+    # ``image_gen/openai``. When empty, falls back to ``name``.
+    key: str = ""
 
 
 @dataclass
@@ -366,6 +386,33 @@ class PluginContext:
             self.manifest.name, engine.name,
         )
 
+    # -- image gen provider registration ------------------------------------
+
+    def register_image_gen_provider(self, provider) -> None:
+        """Register an image generation backend.
+
+        ``provider`` must be an instance of
+        :class:`agent.image_gen_provider.ImageGenProvider`. The
+        ``provider.name`` attribute is what ``image_gen.provider`` in
+        ``config.yaml`` matches against when routing ``image_generate``
+        tool calls.
+        """
+        from agent.image_gen_provider import ImageGenProvider
+        from agent.image_gen_registry import register_provider
+
+        if not isinstance(provider, ImageGenProvider):
+            logger.warning(
+                "Plugin '%s' tried to register an image_gen provider that does "
+                "not inherit from ImageGenProvider. Ignoring.",
+                self.manifest.name,
+            )
+            return
+        register_provider(provider)
+        logger.info(
+            "Plugin '%s' registered image_gen provider: %s",
+            self.manifest.name, provider.name,
+        )
+
     # -- hook registration --------------------------------------------------
 
     def register_hook(self, hook_name: str, callback: Callable) -> None:
@@ -465,11 +512,16 @@ class PluginManager:
         manifests: List[PluginManifest] = []
 
         # 1. Bundled plugins (<repo>/plugins/<name>/)
-        # Repo-shipped generic plugins live next to hermes_cli/.  Memory and
+        #
-        # context_engine subdirs are handled by their own discovery paths, so
+        # Repo-shipped plugins live next to hermes_cli/. Two layouts are
-        # skip those names here.  Bundled plugins are discovered (so they
+        # supported (see ``_scan_directory`` for details):
-        # show up in `hermes plugins`) but only loaded when added to
+        #
-        # `plugins.enabled` in config.yaml — opt-in like any other plugin.
+        #   - flat: ``plugins/disk-cleanup/plugin.yaml`` (standalone)
+        #   - category: ``plugins/image_gen/openai/plugin.yaml`` (backend)
+        #
+        # ``memory/`` and ``context_engine/`` are skipped at the top level —
+        # they have their own discovery systems. Porting those to the
+        # category-namespace ``kind: exclusive`` model is a future PR.
         repo_plugins = Path(__file__).resolve().parent.parent / "plugins"
         manifests.extend(
             self._scan_directory(
@@ -492,36 +544,69 @@ class PluginManager:
         manifests.extend(self._scan_entry_points())
 
         # Load each manifest (skip user-disabled plugins).
-        # Later sources override earlier ones on name collision — user plugins
+        # Later sources override earlier ones on key collision — user
-        # take precedence over bundled, project plugins take precedence over
+        # plugins take precedence over bundled, project plugins take
-        # user.  Dedup here so we only load the final winner.
+        # precedence over user. Dedup here so we only load the final
+        # winner. Keys are path-derived (``image_gen/openai``,
+        # ``disk-cleanup``) so ``tts/openai`` and ``image_gen/openai``
+        # don't collide even when both manifests say ``name: openai``.
         disabled = _get_disabled_plugins()
         enabled = _get_enabled_plugins()  # None = opt-in default (nothing enabled)
         winners: Dict[str, PluginManifest] = {}
         for manifest in manifests:
-            winners[manifest.name] = manifest
+            winners[manifest.key or manifest.name] = manifest
         for manifest in winners.values():
-            # Explicit disable always wins.
+            lookup_key = manifest.key or manifest.name
-            if manifest.name in disabled:
+
+            # Explicit disable always wins (matches on key or on legacy
+            # bare name for back-compat with existing user configs).
+            if lookup_key in disabled or manifest.name in disabled:
                 loaded = LoadedPlugin(manifest=manifest, enabled=False)
                 loaded.error = "disabled via config"
-                self._plugins[manifest.name] = loaded
+                self._plugins[lookup_key] = loaded
-                logger.debug("Skipping disabled plugin '%s'", manifest.name)
+                logger.debug("Skipping disabled plugin '%s'", lookup_key)
                 continue
-            # Opt-in gate: plugins must be in the enabled allow-list.
+
-            # If the allow-list is missing (None), treat as "nothing enabled"
+            # Exclusive plugins (memory providers) have their own
-            # — users have to explicitly enable plugins to load them.
+            # discovery/activation path. The general loader records the
-            # Memory and context_engine providers are excluded from this gate
+            # manifest for introspection but does not load the module.
-            # since they have their own single-select config (memory.provider
+            if manifest.kind == "exclusive":
-            # / context.engine), not the enabled list.
-            if enabled is None or manifest.name not in enabled:
                 loaded = LoadedPlugin(manifest=manifest, enabled=False)
-                loaded.error = "not enabled in config (run `hermes plugins enable {}` to activate)".format(
+                loaded.error = (
-                    manifest.name
+                    "exclusive plugin — activate via <category>.provider config"
                 )
-                self._plugins[manifest.name] = loaded
+                self._plugins[lookup_key] = loaded
                 logger.debug(
-                    "Skipping '%s' (not in plugins.enabled)", manifest.name
+                    "Skipping '%s' (exclusive, handled by category discovery)",
+                    lookup_key,
+                )
+                continue
+
+            # Built-in backends auto-load — they ship with hermes and must
+            # just work. Selection among them (e.g. which image_gen backend
+            # services calls) is driven by ``<category>.provider`` config,
+            # enforced by the tool wrapper.
+            if manifest.kind == "backend" and manifest.source == "bundled":
+                self._load_plugin(manifest)
+                continue
+
+            # Everything else (standalone, user-installed backends,
+            # entry-point plugins) is opt-in via plugins.enabled.
+            # Accept both the path-derived key and the legacy bare name
+            # so existing configs keep working.
+            is_enabled = (
+                enabled is not None
+                and (lookup_key in enabled or manifest.name in enabled)
+            )
+            if not is_enabled:
+                loaded = LoadedPlugin(manifest=manifest, enabled=False)
+                loaded.error = (
+                    "not enabled in config (run `hermes plugins enable {}` to activate)"
+                    .format(lookup_key)
+                )
+                self._plugins[lookup_key] = loaded
+                logger.debug(
+                    "Skipping '%s' (not in plugins.enabled)", lookup_key
                 )
                 continue
             self._load_plugin(manifest)
@@ -545,9 +630,37 @@ class PluginManager:
     ) -> List[PluginManifest]:
         """Read ``plugin.yaml`` manifests from subdirectories of *path*.
 
-        *skip_names* is an optional allow-list of names to ignore (used
+        Supports two layouts, mixed freely:
-        for the bundled scan to exclude ``memory`` / ``context_engine``
+
-        subdirs that have their own discovery path).
+        * **Flat** — ``<root>/<plugin-name>/plugin.yaml``. Key is
+          ``<plugin-name>`` (e.g. ``disk-cleanup``).
+        * **Category** — ``<root>/<category>/<plugin-name>/plugin.yaml``,
+          where the ``<category>`` directory itself has no ``plugin.yaml``.
+          Key is ``<category>/<plugin-name>`` (e.g. ``image_gen/openai``).
+          Depth is capped at two segments.
+
+        *skip_names* is an optional allow-list of names to ignore at the
+        top level (kept for back-compat; the current call sites no longer
+        pass it now that categories are first-class).
+        """
+        return self._scan_directory_level(
+            path, source, skip_names=skip_names, prefix="", depth=0
+        )
+
+    def _scan_directory_level(
+        self,
+        path: Path,
+        source: str,
+        *,
+        skip_names: Optional[Set[str]],
+        prefix: str,
+        depth: int,
+    ) -> List[PluginManifest]:
+        """Recursive implementation of :meth:`_scan_directory`.
+
+        ``prefix`` is the category path already accumulated ("" at root,
+        "image_gen" one level in). ``depth`` is the recursion depth; we
+        cap at 2 so ``<root>/a/b/c/`` is ignored.
         """
         manifests: List[PluginManifest] = []
         if not path.is_dir():
@@ -556,22 +669,73 @@ class PluginManager:
         for child in sorted(path.iterdir()):
             if not child.is_dir():
                 continue
-            if skip_names and child.name in skip_names:
+            if depth == 0 and skip_names and child.name in skip_names:
                 continue
             manifest_file = child / "plugin.yaml"
             if not manifest_file.exists():
                 manifest_file = child / "plugin.yml"
-            if not manifest_file.exists():
+
-                logger.debug("Skipping %s (no plugin.yaml)", child)
+            if manifest_file.exists():
+                manifest = self._parse_manifest(
+                    manifest_file, child, source, prefix
+                )
+                if manifest is not None:
+                    manifests.append(manifest)
                 continue
 
+            # No manifest at this level. If we're still within the depth
+            # cap, treat this directory as a category namespace and recurse
+            # one level in looking for children with manifests.
+            if depth >= 1:
+                logger.debug("Skipping %s (no plugin.yaml, depth cap reached)", child)
+                continue
+
+            sub_prefix = f"{prefix}/{child.name}" if prefix else child.name
+            manifests.extend(
+                self._scan_directory_level(
+                    child,
+                    source,
+                    skip_names=None,
+                    prefix=sub_prefix,
+                    depth=depth + 1,
+                )
+            )
+
+        return manifests
+
+    def _parse_manifest(
+        self,
+        manifest_file: Path,
+        plugin_dir: Path,
+        source: str,
+        prefix: str,
+    ) -> Optional[PluginManifest]:
+        """Parse a single ``plugin.yaml`` into a :class:`PluginManifest`.
+
+        Returns ``None`` on parse failure (logs a warning).
+        """
         try:
             if yaml is None:
                 logger.warning("PyYAML not installed – cannot load %s", manifest_file)
-                    continue
+                return None
             data = yaml.safe_load(manifest_file.read_text()) or {}
-                manifest = PluginManifest(
+
-                    name=data.get("name", child.name),
+            name = data.get("name", plugin_dir.name)
+            key = f"{prefix}/{plugin_dir.name}" if prefix else name
+
+            raw_kind = data.get("kind", "standalone")
+            if not isinstance(raw_kind, str):
+                raw_kind = "standalone"
+            kind = raw_kind.strip().lower()
+            if kind not in _VALID_PLUGIN_KINDS:
+                logger.warning(
+                    "Plugin %s: unknown kind '%s' (valid: %s); treating as 'standalone'",
+                    key, raw_kind, ", ".join(sorted(_VALID_PLUGIN_KINDS)),
+                )
+                kind = "standalone"
+
+            return PluginManifest(
+                name=name,
                 version=str(data.get("version", "")),
                 description=data.get("description", ""),
                 author=data.get("author", ""),
@@ -579,13 +743,13 @@ class PluginManager:
                 provides_tools=data.get("provides_tools", []),
                 provides_hooks=data.get("provides_hooks", []),
                 source=source,
-                    path=str(child),
+                path=str(plugin_dir),
+                kind=kind,
+                key=key,
             )
-                manifests.append(manifest)
         except Exception as exc:
             logger.warning("Failed to parse %s: %s", manifest_file, exc)
-
+            return None
-        return manifests
 
     # -----------------------------------------------------------------------
     # Entry-point scanning
@@ -609,6 +773,7 @@ class PluginManager:
                     name=ep.name,
                     source="entrypoint",
                     path=ep.value,
+                    key=ep.name,
                 )
                 manifests.append(manifest)
         except Exception as exc:
@@ -670,10 +835,16 @@ class PluginManager:
             loaded.error = str(exc)
             logger.warning("Failed to load plugin '%s': %s", manifest.name, exc)
 
-        self._plugins[manifest.name] = loaded
+        self._plugins[manifest.key or manifest.name] = loaded
 
     def _load_directory_module(self, manifest: PluginManifest) -> types.ModuleType:
-        """Import a directory-based plugin as ``hermes_plugins.<name>``."""
+        """Import a directory-based plugin as ``hermes_plugins.<slug>``.
+
+        The module slug is derived from ``manifest.key`` so category-namespaced
+        plugins (``image_gen/openai``) import as
+        ``hermes_plugins.image_gen__openai`` without colliding with any
+        future ``tts/openai``.
+        """
         plugin_dir = Path(manifest.path)  # type: ignore[arg-type]
         init_file = plugin_dir / "__init__.py"
         if not init_file.exists():
@@ -686,7 +857,9 @@ class PluginManager:
             ns_pkg.__package__ = _NS_PARENT
             sys.modules[_NS_PARENT] = ns_pkg
 
-        module_name = f"{_NS_PARENT}.{manifest.name.replace('-', '_')}"
+        key = manifest.key or manifest.name
+        slug = key.replace("/", "__").replace("-", "_")
+        module_name = f"{_NS_PARENT}.{slug}"
         spec = importlib.util.spec_from_file_location(
             module_name,
             init_file,
@@ -767,10 +940,12 @@ class PluginManager:
     def list_plugins(self) -> List[Dict[str, Any]]:
         """Return a list of info dicts for all discovered plugins."""
         result: List[Dict[str, Any]] = []
-        for name, loaded in sorted(self._plugins.items()):
+        for key, loaded in sorted(self._plugins.items()):
             result.append(
                 {
-                    "name": name,
+                    "name": loaded.manifest.name,
+                    "key": loaded.manifest.key or loaded.manifest.name,
+                    "kind": loaded.manifest.kind,
                     "version": loaded.manifest.version,
                     "description": loaded.manifest.description,
                     "source": loaded.manifest.source,

hermes_cli/setup.py

@@ -408,13 +408,36 @@ def _print_setup_summary(config: dict, hermes_home):
             ("Browser Automation", False, missing_browser_hint)
         )
 
-    # FAL (image generation)
+    # Image generation — FAL (direct or via Nous), or any plugin-registered
+    # provider (OpenAI, etc.)
     if subscription_features.image_gen.managed_by_nous:
         tool_status.append(("Image Generation (Nous subscription)", True, None))
     elif subscription_features.image_gen.available:
         tool_status.append(("Image Generation", True, None))
     else:
-        tool_status.append(("Image Generation", False, "FAL_KEY"))
+        # Fall back to probing plugin-registered providers so OpenAI-only
+        # setups don't show as "missing FAL_KEY".
+        _img_backend = None
+        try:
+            from agent.image_gen_registry import list_providers
+            from hermes_cli.plugins import _ensure_plugins_discovered
+
+            _ensure_plugins_discovered()
+            for _p in list_providers():
+                if _p.name == "fal":
+                    continue
+                try:
+                    if _p.is_available():
+                        _img_backend = _p.display_name
+                        break
+                except Exception:
+                    continue
+        except Exception:
+            pass
+        if _img_backend:
+            tool_status.append((f"Image Generation ({_img_backend})", True, None))
+        else:
+            tool_status.append(("Image Generation", False, "FAL_KEY or OPENAI_API_KEY"))
 
     # TTS — show configured provider
     tts_provider = config.get("tts", {}).get("provider", "edge")

hermes_cli/tools_config.py

@@ -847,6 +847,51 @@ def _configure_toolset(ts_key: str, config: dict):
         _configure_simple_requirements(ts_key)
 
 
+def _plugin_image_gen_providers() -> list[dict]:
+    """Build picker-row dicts from plugin-registered image gen providers.
+
+    Each returned dict looks like a regular ``TOOL_CATEGORIES`` provider
+    row but carries an ``image_gen_plugin_name`` marker so downstream
+    code (config writing, model picker) knows to route through the
+    plugin registry instead of the in-tree FAL backend.
+
+    FAL is skipped — it's already exposed by the hardcoded
+    ``TOOL_CATEGORIES["image_gen"]`` entries. When FAL gets ported to
+    a plugin in a follow-up PR, the hardcoded entries go away and this
+    function surfaces it alongside OpenAI automatically.
+    """
+    try:
+        from agent.image_gen_registry import list_providers
+        from hermes_cli.plugins import _ensure_plugins_discovered
+
+        _ensure_plugins_discovered()
+        providers = list_providers()
+    except Exception:
+        return []
+
+    rows: list[dict] = []
+    for provider in providers:
+        if getattr(provider, "name", None) == "fal":
+            # FAL has its own hardcoded rows today.
+            continue
+        try:
+            schema = provider.get_setup_schema()
+        except Exception:
+            continue
+        if not isinstance(schema, dict):
+            continue
+        rows.append(
+            {
+                "name": schema.get("name", provider.display_name),
+                "badge": schema.get("badge", ""),
+                "tag": schema.get("tag", ""),
+                "env_vars": schema.get("env_vars", []),
+                "image_gen_plugin_name": provider.name,
+            }
+        )
+    return rows
+
+
 def _visible_providers(cat: dict, config: dict) -> list[dict]:
     """Return provider entries visible for the current auth/config state."""
     features = get_nous_subscription_features(config)
@@ -857,6 +902,12 @@ def _visible_providers(cat: dict, config: dict) -> list[dict]:
         if provider.get("requires_nous_auth") and not features.nous_auth_present:
             continue
         visible.append(provider)
+
+    # Inject plugin-registered image_gen backends (OpenAI today, more
+    # later) so the picker lists them alongside FAL / Nous Subscription.
+    if cat.get("name") == "Image Generation":
+        visible.extend(_plugin_image_gen_providers())
+
     return visible
 
 
@@ -876,7 +927,24 @@ def _toolset_needs_configuration_prompt(ts_key: str, config: dict) -> bool:
         browser_cfg = config.get("browser", {})
         return not isinstance(browser_cfg, dict) or "cloud_provider" not in browser_cfg
     if ts_key == "image_gen":
-        return not fal_key_is_configured()
+        # Satisfied when the in-tree FAL backend is configured OR any
+        # plugin-registered image gen provider is available.
+        if fal_key_is_configured():
+            return False
+        try:
+            from agent.image_gen_registry import list_providers
+            from hermes_cli.plugins import _ensure_plugins_discovered
+
+            _ensure_plugins_discovered()
+            for provider in list_providers():
+                try:
+                    if provider.is_available():
+                        return False
+                except Exception:
+                    continue
+        except Exception:
+            pass
+        return True
 
     return not _toolset_has_keys(ts_key, config)
 
@@ -1095,6 +1163,88 @@ def _configure_imagegen_model(backend_name: str, config: dict) -> None:
     _print_success(f"  Model set to: {chosen}")
 
 
+def _plugin_image_gen_catalog(plugin_name: str):
+    """Return ``(catalog_dict, default_model_id)`` for a plugin provider.
+
+    ``catalog_dict`` is shaped like the legacy ``FAL_MODELS`` table —
+    ``{model_id: {"display", "speed", "strengths", "price", ...}}`` —
+    so the existing picker code paths work without change. Returns
+    ``({}, None)`` if the provider isn't registered or has no models.
+    """
+    try:
+        from agent.image_gen_registry import get_provider
+        from hermes_cli.plugins import _ensure_plugins_discovered
+
+        _ensure_plugins_discovered()
+        provider = get_provider(plugin_name)
+    except Exception:
+        return {}, None
+    if provider is None:
+        return {}, None
+    try:
+        models = provider.list_models() or []
+        default = provider.default_model()
+    except Exception:
+        return {}, None
+    catalog = {m["id"]: m for m in models if isinstance(m, dict) and "id" in m}
+    return catalog, default
+
+
+def _configure_imagegen_model_for_plugin(plugin_name: str, config: dict) -> None:
+    """Prompt the user to pick a model for a plugin-registered backend.
+
+    Writes selection to ``image_gen.model``. Mirrors
+    :func:`_configure_imagegen_model` but sources its catalog from the
+    plugin registry instead of :data:`IMAGEGEN_BACKENDS`.
+    """
+    catalog, default_model = _plugin_image_gen_catalog(plugin_name)
+    if not catalog:
+        return
+
+    cur_cfg = config.setdefault("image_gen", {})
+    if not isinstance(cur_cfg, dict):
+        cur_cfg = {}
+        config["image_gen"] = cur_cfg
+    current_model = cur_cfg.get("model") or default_model
+    if current_model not in catalog:
+        current_model = default_model
+
+    model_ids = list(catalog.keys())
+    ordered = [current_model] + [m for m in model_ids if m != current_model]
+
+    widths = {
+        "model": max(len(m) for m in model_ids),
+        "speed": max((len(catalog[m].get("speed", "")) for m in model_ids), default=6),
+        "strengths": max((len(catalog[m].get("strengths", "")) for m in model_ids), default=0),
+    }
+
+    print()
+    header = (
+        f"  {'Model':<{widths['model']}}  "
+        f"{'Speed':<{widths['speed']}}  "
+        f"{'Strengths':<{widths['strengths']}}  "
+        f"Price"
+    )
+    print(color(header, Colors.CYAN))
+
+    rows = []
+    for mid in ordered:
+        row = _format_imagegen_model_row(mid, catalog[mid], widths)
+        if mid == current_model:
+            row += "  ← currently in use"
+        rows.append(row)
+
+    idx = _prompt_choice(
+        f"  Choose {plugin_name} model:",
+        rows,
+        default=0,
+    )
+
+    chosen = ordered[idx]
+    cur_cfg["model"] = chosen
+    _print_success(f"  Model set to: {chosen}")
+
+
 def _configure_provider(provider: dict, config: dict):
     """Configure a single provider - prompt for API keys and set config."""
     env_vars = provider.get("env_vars", [])
@@ -1151,10 +1301,28 @@ def _configure_provider(provider: dict, config: dict):
         _print_success(f"  {provider['name']} - no configuration needed!")
         if managed_feature:
             _print_info("  Requests for this tool will be billed to your Nous subscription.")
+        # Plugin-registered image_gen provider: write image_gen.provider
+        # and route model selection to the plugin's own catalog.
+        plugin_name = provider.get("image_gen_plugin_name")
+        if plugin_name:
+            img_cfg = config.setdefault("image_gen", {})
+            if not isinstance(img_cfg, dict):
+                img_cfg = {}
+                config["image_gen"] = img_cfg
+            img_cfg["provider"] = plugin_name
+            _print_success(f"  image_gen.provider set to: {plugin_name}")
+            _configure_imagegen_model_for_plugin(plugin_name, config)
+            return
         # Imagegen backends prompt for model selection after backend pick.
         backend = provider.get("imagegen_backend")
         if backend:
             _configure_imagegen_model(backend, config)
+            # In-tree FAL is the only non-plugin backend today. Keep
+            # image_gen.provider clear so the dispatch shim falls through
+            # to the legacy FAL path.
+            img_cfg = config.setdefault("image_gen", {})
+            if isinstance(img_cfg, dict) and img_cfg.get("provider") not in (None, "", "fal"):
+                img_cfg["provider"] = "fal"
         return
 
     # Prompt for each required env var
@@ -1189,10 +1357,23 @@ def _configure_provider(provider: dict, config: dict):
 
     if all_configured:
         _print_success(f"  {provider['name']} configured!")
+        plugin_name = provider.get("image_gen_plugin_name")
+        if plugin_name:
+            img_cfg = config.setdefault("image_gen", {})
+            if not isinstance(img_cfg, dict):
+                img_cfg = {}
+                config["image_gen"] = img_cfg
+            img_cfg["provider"] = plugin_name
+            _print_success(f"  image_gen.provider set to: {plugin_name}")
+            _configure_imagegen_model_for_plugin(plugin_name, config)
+            return
         # Imagegen backends prompt for model selection after env vars are in.
         backend = provider.get("imagegen_backend")
         if backend:
             _configure_imagegen_model(backend, config)
+            img_cfg = config.setdefault("image_gen", {})
+            if isinstance(img_cfg, dict) and img_cfg.get("provider") not in (None, "", "fal"):
+                img_cfg["provider"] = "fal"
 
 
 def _configure_simple_requirements(ts_key: str):

plugins/image_gen/openai/__init__.py

@@ -0,0 +1,303 @@
+"""OpenAI image generation backend.
+
+Exposes OpenAI's ``gpt-image-2`` model at three quality tiers as an
+:class:`ImageGenProvider` implementation. The tiers are implemented as
+three virtual model IDs so the ``hermes tools`` model picker and the
+``image_gen.model`` config key behave like any other multi-model backend:
+
+    gpt-image-2-low     ~15s   fastest, good for iteration
+    gpt-image-2-medium  ~40s   default — balanced
+    gpt-image-2-high    ~2min  slowest, highest fidelity
+
+All three hit the same underlying API model (``gpt-image-2``) with a
+different ``quality`` parameter. Output is base64 JSON → saved under
+``$HERMES_HOME/cache/images/``.
+
+Selection precedence (first hit wins):
+
+1. ``OPENAI_IMAGE_MODEL`` env var (escape hatch for scripts / tests)
+2. ``image_gen.openai.model`` in ``config.yaml``
+3. ``image_gen.model`` in ``config.yaml`` (when it's one of our tier IDs)
+4. :data:`DEFAULT_MODEL` — ``gpt-image-2-medium``
+"""
+
+from __future__ import annotations
+
+import logging
+import os
+from typing import Any, Dict, List, Optional, Tuple
+
+from agent.image_gen_provider import (
+    DEFAULT_ASPECT_RATIO,
+    ImageGenProvider,
+    error_response,
+    resolve_aspect_ratio,
+    save_b64_image,
+    success_response,
+)
+
+logger = logging.getLogger(__name__)
+
+
+# ---------------------------------------------------------------------------
+# Model catalog
+# ---------------------------------------------------------------------------
+#
+# All three IDs resolve to the same underlying API model with a different
+# ``quality`` setting. ``api_model`` is what gets sent to OpenAI;
+# ``quality`` is the knob that changes generation time and output fidelity.
+
+API_MODEL = "gpt-image-2"
+
+_MODELS: Dict[str, Dict[str, Any]] = {
+    "gpt-image-2-low": {
+        "display": "GPT Image 2 (Low)",
+        "speed": "~15s",
+        "strengths": "Fast iteration, lowest cost",
+        "quality": "low",
+    },
+    "gpt-image-2-medium": {
+        "display": "GPT Image 2 (Medium)",
+        "speed": "~40s",
+        "strengths": "Balanced — default",
+        "quality": "medium",
+    },
+    "gpt-image-2-high": {
+        "display": "GPT Image 2 (High)",
+        "speed": "~2min",
+        "strengths": "Highest fidelity, strongest prompt adherence",
+        "quality": "high",
+    },
+}
+
+DEFAULT_MODEL = "gpt-image-2-medium"
+
+_SIZES = {
+    "landscape": "1536x1024",
+    "square": "1024x1024",
+    "portrait": "1024x1536",
+}
+
+
+def _load_openai_config() -> Dict[str, Any]:
+    """Read ``image_gen`` from config.yaml (returns {} on any failure)."""
+    try:
+        from hermes_cli.config import load_config
+
+        cfg = load_config()
+        section = cfg.get("image_gen") if isinstance(cfg, dict) else None
+        return section if isinstance(section, dict) else {}
+    except Exception as exc:
+        logger.debug("Could not load image_gen config: %s", exc)
+        return {}
+
+
+def _resolve_model() -> Tuple[str, Dict[str, Any]]:
+    """Decide which tier to use and return ``(model_id, meta)``."""
+    env_override = os.environ.get("OPENAI_IMAGE_MODEL")
+    if env_override and env_override in _MODELS:
+        return env_override, _MODELS[env_override]
+
+    cfg = _load_openai_config()
+    openai_cfg = cfg.get("openai") if isinstance(cfg.get("openai"), dict) else {}
+    candidate: Optional[str] = None
+    if isinstance(openai_cfg, dict):
+        value = openai_cfg.get("model")
+        if isinstance(value, str) and value in _MODELS:
+            candidate = value
+    if candidate is None:
+        top = cfg.get("model")
+        if isinstance(top, str) and top in _MODELS:
+            candidate = top
+
+    if candidate is not None:
+        return candidate, _MODELS[candidate]
+
+    return DEFAULT_MODEL, _MODELS[DEFAULT_MODEL]
+
+
+# ---------------------------------------------------------------------------
+# Provider
+# ---------------------------------------------------------------------------
+
+
+class OpenAIImageGenProvider(ImageGenProvider):
+    """OpenAI ``images.generate`` backend — gpt-image-2 at low/medium/high."""
+
+    @property
+    def name(self) -> str:
+        return "openai"
+
+    @property
+    def display_name(self) -> str:
+        return "OpenAI"
+
+    def is_available(self) -> bool:
+        if not os.environ.get("OPENAI_API_KEY"):
+            return False
+        try:
+            import openai  # noqa: F401
+        except ImportError:
+            return False
+        return True
+
+    def list_models(self) -> List[Dict[str, Any]]:
+        return [
+            {
+                "id": model_id,
+                "display": meta["display"],
+                "speed": meta["speed"],
+                "strengths": meta["strengths"],
+                "price": "varies",
+            }
+            for model_id, meta in _MODELS.items()
+        ]
+
+    def default_model(self) -> Optional[str]:
+        return DEFAULT_MODEL
+
+    def get_setup_schema(self) -> Dict[str, Any]:
+        return {
+            "name": "OpenAI",
+            "badge": "paid",
+            "tag": "gpt-image-2 at low/medium/high quality tiers",
+            "env_vars": [
+                {
+                    "key": "OPENAI_API_KEY",
+                    "prompt": "OpenAI API key",
+                    "url": "https://platform.openai.com/api-keys",
+                },
+            ],
+        }
+
+    def generate(
+        self,
+        prompt: str,
+        aspect_ratio: str = DEFAULT_ASPECT_RATIO,
+        **kwargs: Any,
+    ) -> Dict[str, Any]:
+        prompt = (prompt or "").strip()
+        aspect = resolve_aspect_ratio(aspect_ratio)
+
+        if not prompt:
+            return error_response(
+                error="Prompt is required and must be a non-empty string",
+                error_type="invalid_argument",
+                provider="openai",
+                aspect_ratio=aspect,
+            )
+
+        if not os.environ.get("OPENAI_API_KEY"):
+            return error_response(
+                error=(
+                    "OPENAI_API_KEY not set. Run `hermes tools` → Image "
+                    "Generation → OpenAI to configure, or `hermes setup` "
+                    "to add the key."
+                ),
+                error_type="auth_required",
+                provider="openai",
+                aspect_ratio=aspect,
+            )
+
+        try:
+            import openai
+        except ImportError:
+            return error_response(
+                error="openai Python package not installed (pip install openai)",
+                error_type="missing_dependency",
+                provider="openai",
+                aspect_ratio=aspect,
+            )
+
+        tier_id, meta = _resolve_model()
+        size = _SIZES.get(aspect, _SIZES["square"])
+
+        # gpt-image-2 returns b64_json unconditionally and REJECTS
+        # ``response_format`` as an unknown parameter. Don't send it.
+        payload: Dict[str, Any] = {
+            "model": API_MODEL,
+            "prompt": prompt,
+            "size": size,
+            "n": 1,
+            "quality": meta["quality"],
+        }
+
+        try:
+            client = openai.OpenAI()
+            response = client.images.generate(**payload)
+        except Exception as exc:
+            logger.debug("OpenAI image generation failed", exc_info=True)
+            return error_response(
+                error=f"OpenAI image generation failed: {exc}",
+                error_type="api_error",
+                provider="openai",
+                model=tier_id,
+                prompt=prompt,
+                aspect_ratio=aspect,
+            )
+
+        data = getattr(response, "data", None) or []
+        if not data:
+            return error_response(
+                error="OpenAI returned no image data",
+                error_type="empty_response",
+                provider="openai",
+                model=tier_id,
+                prompt=prompt,
+                aspect_ratio=aspect,
+            )
+
+        first = data[0]
+        b64 = getattr(first, "b64_json", None)
+        url = getattr(first, "url", None)
+        revised_prompt = getattr(first, "revised_prompt", None)
+
+        if b64:
+            try:
+                saved_path = save_b64_image(b64, prefix=f"openai_{tier_id}")
+            except Exception as exc:
+                return error_response(
+                    error=f"Could not save image to cache: {exc}",
+                    error_type="io_error",
+                    provider="openai",
+                    model=tier_id,
+                    prompt=prompt,
+                    aspect_ratio=aspect,
+                )
+            image_ref = str(saved_path)
+        elif url:
+            # Defensive — gpt-image-2 returns b64 today, but fall back
+            # gracefully if the API ever changes.
+            image_ref = url
+        else:
+            return error_response(
+                error="OpenAI response contained neither b64_json nor URL",
+                error_type="empty_response",
+                provider="openai",
+                model=tier_id,
+                prompt=prompt,
+                aspect_ratio=aspect,
+            )
+
+        extra: Dict[str, Any] = {"size": size, "quality": meta["quality"]}
+        if revised_prompt:
+            extra["revised_prompt"] = revised_prompt
+
+        return success_response(
+            image=image_ref,
+            model=tier_id,
+            prompt=prompt,
+            aspect_ratio=aspect,
+            provider="openai",
+            extra=extra,
+        )
+
+
+# ---------------------------------------------------------------------------
+# Plugin entry point
+# ---------------------------------------------------------------------------
+
+
+def register(ctx) -> None:
+    """Plugin entry point — wire ``OpenAIImageGenProvider`` into the registry."""
+    ctx.register_image_gen_provider(OpenAIImageGenProvider())

plugins/image_gen/openai/plugin.yaml

@@ -0,0 +1,7 @@
+name: openai
+version: 1.0.0
+description: "OpenAI image generation backend (gpt-image-2). Saves generated images to $HERMES_HOME/cache/images/."
+author: NousResearch
+kind: backend
+requires_env:
+  - OPENAI_API_KEY

tests/agent/test_image_gen_registry.py

@@ -0,0 +1,111 @@
+"""Tests for agent/image_gen_registry.py — provider registration & active lookup."""
+
+from __future__ import annotations
+
+import pytest
+
+from agent import image_gen_registry
+from agent.image_gen_provider import ImageGenProvider
+
+
+class _FakeProvider(ImageGenProvider):
+    def __init__(self, name: str, available: bool = True):
+        self._name = name
+        self._available = available
+
+    @property
+    def name(self) -> str:
+        return self._name
+
+    def is_available(self) -> bool:
+        return self._available
+
+    def generate(self, prompt, aspect_ratio="landscape", **kw):
+        return {"success": True, "image": f"{self._name}://{prompt}"}
+
+
+@pytest.fixture(autouse=True)
+def _reset_registry():
+    image_gen_registry._reset_for_tests()
+    yield
+    image_gen_registry._reset_for_tests()
+
+
+class TestRegisterProvider:
+    def test_register_and_lookup(self):
+        provider = _FakeProvider("fake")
+        image_gen_registry.register_provider(provider)
+        assert image_gen_registry.get_provider("fake") is provider
+
+    def test_rejects_non_provider(self):
+        with pytest.raises(TypeError):
+            image_gen_registry.register_provider("not a provider")  # type: ignore[arg-type]
+
+    def test_rejects_empty_name(self):
+        class Empty(ImageGenProvider):
+            @property
+            def name(self) -> str:
+                return ""
+
+            def generate(self, prompt, aspect_ratio="landscape", **kw):
+                return {}
+
+        with pytest.raises(ValueError):
+            image_gen_registry.register_provider(Empty())
+
+    def test_reregister_overwrites(self):
+        a = _FakeProvider("same")
+        b = _FakeProvider("same")
+        image_gen_registry.register_provider(a)
+        image_gen_registry.register_provider(b)
+        assert image_gen_registry.get_provider("same") is b
+
+    def test_list_is_sorted(self):
+        image_gen_registry.register_provider(_FakeProvider("zeta"))
+        image_gen_registry.register_provider(_FakeProvider("alpha"))
+        names = [p.name for p in image_gen_registry.list_providers()]
+        assert names == ["alpha", "zeta"]
+
+
+class TestGetActiveProvider:
+    def test_single_provider_autoresolves(self, tmp_path, monkeypatch):
+        monkeypatch.setenv("HERMES_HOME", str(tmp_path))
+        image_gen_registry.register_provider(_FakeProvider("solo"))
+        active = image_gen_registry.get_active_provider()
+        assert active is not None and active.name == "solo"
+
+    def test_fal_preferred_on_multi_without_config(self, tmp_path, monkeypatch):
+        monkeypatch.setenv("HERMES_HOME", str(tmp_path))
+        image_gen_registry.register_provider(_FakeProvider("fal"))
+        image_gen_registry.register_provider(_FakeProvider("openai"))
+        active = image_gen_registry.get_active_provider()
+        assert active is not None and active.name == "fal"
+
+    def test_explicit_config_wins(self, tmp_path, monkeypatch):
+        import yaml
+
+        monkeypatch.setenv("HERMES_HOME", str(tmp_path))
+        (tmp_path / "config.yaml").write_text(
+            yaml.safe_dump({"image_gen": {"provider": "openai"}})
+        )
+        image_gen_registry.register_provider(_FakeProvider("fal"))
+        image_gen_registry.register_provider(_FakeProvider("openai"))
+        active = image_gen_registry.get_active_provider()
+        assert active is not None and active.name == "openai"
+
+    def test_missing_configured_provider_falls_back(self, tmp_path, monkeypatch):
+        import yaml
+
+        monkeypatch.setenv("HERMES_HOME", str(tmp_path))
+        (tmp_path / "config.yaml").write_text(
+            yaml.safe_dump({"image_gen": {"provider": "replicate"}})
+        )
+        # Only FAL is registered — configured provider doesn't exist
+        image_gen_registry.register_provider(_FakeProvider("fal"))
+        active = image_gen_registry.get_active_provider()
+        # Falls back to FAL preference (legacy default) rather than None
+        assert active is not None and active.name == "fal"
+
+    def test_none_when_empty(self, tmp_path, monkeypatch):
+        monkeypatch.setenv("HERMES_HOME", str(tmp_path))
+        assert image_gen_registry.get_active_provider() is None

tests/hermes_cli/test_image_gen_picker.py

@@ -0,0 +1,174 @@
+"""Tests for plugin image_gen providers injecting themselves into the picker.
+
+Covers `_plugin_image_gen_providers`, `_visible_providers`, and
+`_toolset_needs_configuration_prompt` handling of plugin providers.
+"""
+
+from __future__ import annotations
+
+import pytest
+
+from agent import image_gen_registry
+from agent.image_gen_provider import ImageGenProvider
+
+
+class _FakeProvider(ImageGenProvider):
+    def __init__(self, name: str, available: bool = True, schema=None, models=None):
+        self._name = name
+        self._available = available
+        self._schema = schema or {
+            "name": name.title(),
+            "badge": "test",
+            "tag": f"{name} test tag",
+            "env_vars": [{"key": f"{name.upper()}_API_KEY", "prompt": f"{name} key"}],
+        }
+        self._models = models or [
+            {"id": f"{name}-model-v1", "display": f"{name} v1",
+             "speed": "~5s", "strengths": "test", "price": "$"},
+        ]
+
+    @property
+    def name(self) -> str:
+        return self._name
+
+    def is_available(self) -> bool:
+        return self._available
+
+    def list_models(self):
+        return list(self._models)
+
+    def default_model(self):
+        return self._models[0]["id"] if self._models else None
+
+    def get_setup_schema(self):
+        return dict(self._schema)
+
+    def generate(self, prompt, aspect_ratio="landscape", **kw):
+        return {"success": True, "image": f"{self._name}://{prompt}"}
+
+
+@pytest.fixture(autouse=True)
+def _reset_registry():
+    image_gen_registry._reset_for_tests()
+    yield
+    image_gen_registry._reset_for_tests()
+
+
+class TestPluginPickerInjection:
+    def test_plugin_providers_returns_registered(self, monkeypatch):
+        from hermes_cli import tools_config
+
+        image_gen_registry.register_provider(_FakeProvider("myimg"))
+
+        rows = tools_config._plugin_image_gen_providers()
+        names = [r["name"] for r in rows]
+        plugin_names = [r.get("image_gen_plugin_name") for r in rows]
+
+        assert "Myimg" in names
+        assert "myimg" in plugin_names
+
+    def test_fal_skipped_to_avoid_duplicate(self, monkeypatch):
+        from hermes_cli import tools_config
+
+        # Simulate a FAL plugin being registered — the picker already has
+        # hardcoded FAL rows in TOOL_CATEGORIES, so plugin-FAL must be
+        # skipped to avoid showing FAL twice.
+        image_gen_registry.register_provider(_FakeProvider("fal"))
+        image_gen_registry.register_provider(_FakeProvider("openai"))
+
+        rows = tools_config._plugin_image_gen_providers()
+        names = [r.get("image_gen_plugin_name") for r in rows]
+        assert "fal" not in names
+        assert "openai" in names
+
+    def test_visible_providers_includes_plugins_for_image_gen(self, monkeypatch):
+        from hermes_cli import tools_config
+
+        image_gen_registry.register_provider(_FakeProvider("someimg"))
+
+        cat = tools_config.TOOL_CATEGORIES["image_gen"]
+        visible = tools_config._visible_providers(cat, {})
+        plugin_names = [p.get("image_gen_plugin_name") for p in visible if p.get("image_gen_plugin_name")]
+        assert "someimg" in plugin_names
+
+    def test_visible_providers_does_not_inject_into_other_categories(self, monkeypatch):
+        from hermes_cli import tools_config
+
+        image_gen_registry.register_provider(_FakeProvider("someimg"))
+
+        # Browser category must NOT see image_gen plugins.
+        browser = tools_config.TOOL_CATEGORIES["browser"]
+        visible = tools_config._visible_providers(browser, {})
+        assert all(p.get("image_gen_plugin_name") is None for p in visible)
+
+
+class TestPluginCatalog:
+    def test_plugin_catalog_returns_models(self):
+        from hermes_cli import tools_config
+
+        image_gen_registry.register_provider(_FakeProvider("catimg"))
+
+        catalog, default = tools_config._plugin_image_gen_catalog("catimg")
+        assert "catimg-model-v1" in catalog
+        assert default == "catimg-model-v1"
+
+    def test_plugin_catalog_empty_for_unknown(self):
+        from hermes_cli import tools_config
+
+        catalog, default = tools_config._plugin_image_gen_catalog("does-not-exist")
+        assert catalog == {}
+        assert default is None
+
+
+class TestConfigPrompt:
+    def test_image_gen_satisfied_by_plugin_provider(self, monkeypatch, tmp_path):
+        """When a plugin provider reports is_available(), the picker should
+        not force a setup prompt on the user."""
+        from hermes_cli import tools_config
+
+        monkeypatch.setenv("HERMES_HOME", str(tmp_path))
+        monkeypatch.delenv("FAL_KEY", raising=False)
+
+        image_gen_registry.register_provider(_FakeProvider("avail-img", available=True))
+
+        assert tools_config._toolset_needs_configuration_prompt("image_gen", {}) is False
+
+    def test_image_gen_still_prompts_when_nothing_available(self, monkeypatch, tmp_path):
+        from hermes_cli import tools_config
+
+        monkeypatch.setenv("HERMES_HOME", str(tmp_path))
+        monkeypatch.delenv("FAL_KEY", raising=False)
+
+        image_gen_registry.register_provider(_FakeProvider("unavail-img", available=False))
+
+        assert tools_config._toolset_needs_configuration_prompt("image_gen", {}) is True
+
+
+class TestConfigWriting:
+    def test_picking_plugin_provider_writes_provider_and_model(self, monkeypatch, tmp_path):
+        """When a user picks a plugin-backed image_gen provider with no
+        env vars needed, ``_configure_provider`` should write both
+        ``image_gen.provider`` and ``image_gen.model``."""
+        from hermes_cli import tools_config
+
+        monkeypatch.setenv("HERMES_HOME", str(tmp_path))
+        image_gen_registry.register_provider(_FakeProvider("noenv", schema={
+            "name": "NoEnv",
+            "badge": "free",
+            "tag": "",
+            "env_vars": [],
+        }))
+
+        # Stub out the interactive model picker — no TTY in tests.
+        monkeypatch.setattr(tools_config, "_prompt_choice", lambda *a, **kw: 0)
+
+        config: dict = {}
+        provider_row = {
+            "name": "NoEnv",
+            "env_vars": [],
+            "image_gen_plugin_name": "noenv",
+        }
+        tools_config._configure_provider(provider_row, config)
+
+        assert config["image_gen"]["provider"] == "noenv"
+        assert config["image_gen"]["model"] == "noenv-model-v1"

tests/hermes_cli/test_plugin_scanner_recursion.py

@@ -0,0 +1,357 @@
+"""Tests for PR1 pluggable image gen: scanner recursion, kinds, path keys.
+
+Covers ``_scan_directory`` recursion into category namespaces
+(``plugins/image_gen/openai/``), ``kind`` parsing, path-derived registry
+keys, and the new gate logic (bundled backends auto-load; user backends
+still opt-in; exclusive kind skipped; unknown kinds → standalone warning).
+"""
+
+from __future__ import annotations
+
+from pathlib import Path
+from typing import Any, Dict
+
+import pytest
+import yaml
+
+from hermes_cli.plugins import PluginManager, PluginManifest
+
+
+# ── Helpers ────────────────────────────────────────────────────────────────
+
+
+def _write_plugin(
+    root: Path,
+    segments: list[str],
+    *,
+    manifest_extra: Dict[str, Any] | None = None,
+    register_body: str = "pass",
+) -> Path:
+    """Create a plugin dir at ``root/<segments...>/`` with plugin.yaml + __init__.py.
+
+    ``segments`` lets tests build both flat (``["my-plugin"]``) and
+    category-namespaced (``["image_gen", "openai"]``) layouts.
+    """
+    plugin_dir = root
+    for seg in segments:
+        plugin_dir = plugin_dir / seg
+    plugin_dir.mkdir(parents=True, exist_ok=True)
+
+    manifest = {
+        "name": segments[-1],
+        "version": "0.1.0",
+        "description": f"Test plugin {'/'.join(segments)}",
+    }
+    if manifest_extra:
+        manifest.update(manifest_extra)
+    (plugin_dir / "plugin.yaml").write_text(yaml.dump(manifest))
+    (plugin_dir / "__init__.py").write_text(
+        f"def register(ctx):\n    {register_body}\n"
+    )
+    return plugin_dir
+
+
+def _enable(hermes_home: Path, name: str) -> None:
+    """Append ``name`` to ``plugins.enabled`` in ``<hermes_home>/config.yaml``."""
+    cfg_path = hermes_home / "config.yaml"
+    cfg: dict = {}
+    if cfg_path.exists():
+        try:
+            cfg = yaml.safe_load(cfg_path.read_text()) or {}
+        except Exception:
+            cfg = {}
+    plugins_cfg = cfg.setdefault("plugins", {})
+    enabled = plugins_cfg.setdefault("enabled", [])
+    if isinstance(enabled, list) and name not in enabled:
+        enabled.append(name)
+    cfg_path.write_text(yaml.safe_dump(cfg))
+
+
+# ── Scanner recursion ──────────────────────────────────────────────────────
+
+
+class TestCategoryNamespaceRecursion:
+    def test_category_namespace_discovered(self, tmp_path, monkeypatch):
+        """``<root>/image_gen/openai/plugin.yaml`` is discovered with key
+        ``image_gen/openai`` when the ``image_gen`` parent has no manifest."""
+        import os
+        hermes_home = Path(os.environ["HERMES_HOME"])  # set by hermetic conftest fixture
+        user_plugins = hermes_home / "plugins"
+
+        _write_plugin(user_plugins, ["image_gen", "openai"])
+        _enable(hermes_home, "image_gen/openai")
+
+        mgr = PluginManager()
+        mgr.discover_and_load()
+
+        assert "image_gen/openai" in mgr._plugins
+        loaded = mgr._plugins["image_gen/openai"]
+        assert loaded.manifest.key == "image_gen/openai"
+        assert loaded.manifest.name == "openai"
+        assert loaded.enabled is True
+
+    def test_flat_plugin_key_matches_name(self, tmp_path, monkeypatch):
+        """Flat plugins keep their bare name as the key (back-compat)."""
+        import os
+        hermes_home = Path(os.environ["HERMES_HOME"])  # set by hermetic conftest fixture
+        user_plugins = hermes_home / "plugins"
+
+        _write_plugin(user_plugins, ["my-plugin"])
+        _enable(hermes_home, "my-plugin")
+
+        mgr = PluginManager()
+        mgr.discover_and_load()
+
+        assert "my-plugin" in mgr._plugins
+        assert mgr._plugins["my-plugin"].manifest.key == "my-plugin"
+
+    def test_depth_cap_two(self, tmp_path, monkeypatch):
+        """Plugins nested three levels deep are not discovered.
+
+        ``<root>/a/b/c/plugin.yaml`` should NOT be picked up — cap is
+        two segments.
+        """
+        import os
+        hermes_home = Path(os.environ["HERMES_HOME"])  # set by hermetic conftest fixture
+        user_plugins = hermes_home / "plugins"
+
+        _write_plugin(user_plugins, ["a", "b", "c"])
+
+        mgr = PluginManager()
+        mgr.discover_and_load()
+
+        non_bundled = [
+            k for k, p in mgr._plugins.items()
+            if p.manifest.source != "bundled"
+        ]
+        assert non_bundled == []
+
+    def test_category_dir_with_manifest_is_leaf(self, tmp_path, monkeypatch):
+        """If ``image_gen/plugin.yaml`` exists, ``image_gen`` itself IS the
+        plugin and its children are ignored."""
+        import os
+        hermes_home = Path(os.environ["HERMES_HOME"])  # set by hermetic conftest fixture
+        user_plugins = hermes_home / "plugins"
+
+        # parent has a manifest → stop recursing
+        _write_plugin(user_plugins, ["image_gen"])
+        # child also has a manifest — should NOT be found because we stop
+        # at the parent.
+        _write_plugin(user_plugins, ["image_gen", "openai"])
+        _enable(hermes_home, "image_gen")
+        _enable(hermes_home, "image_gen/openai")
+
+        mgr = PluginManager()
+        mgr.discover_and_load()
+
+        # The bundled plugins/image_gen/openai/ exists in the repo — filter
+        # it out so we're only asserting on the user-dir layout.
+        user_plugins_in_registry = {
+            k for k, p in mgr._plugins.items() if p.manifest.source != "bundled"
+        }
+        assert "image_gen" in user_plugins_in_registry
+        assert "image_gen/openai" not in user_plugins_in_registry
+
+
+# ── Kind parsing ───────────────────────────────────────────────────────────
+
+
+class TestKindField:
+    def test_default_kind_is_standalone(self, tmp_path, monkeypatch):
+        import os
+        hermes_home = Path(os.environ["HERMES_HOME"])  # set by hermetic conftest fixture
+        _write_plugin(hermes_home / "plugins", ["p1"])
+        _enable(hermes_home, "p1")
+
+        mgr = PluginManager()
+        mgr.discover_and_load()
+
+        assert mgr._plugins["p1"].manifest.kind == "standalone"
+
+    @pytest.mark.parametrize("kind", ["backend", "exclusive", "standalone"])
+    def test_valid_kinds_parsed(self, kind, tmp_path, monkeypatch):
+        import os
+        hermes_home = Path(os.environ["HERMES_HOME"])  # set by hermetic conftest fixture
+        _write_plugin(
+            hermes_home / "plugins",
+            ["p1"],
+            manifest_extra={"kind": kind},
+        )
+        # Not all kinds auto-load, but manifest should parse.
+        _enable(hermes_home, "p1")
+
+        mgr = PluginManager()
+        mgr.discover_and_load()
+
+        assert "p1" in mgr._plugins
+        assert mgr._plugins["p1"].manifest.kind == kind
+
+    def test_unknown_kind_falls_back_to_standalone(self, tmp_path, monkeypatch, caplog):
+        import os
+        hermes_home = Path(os.environ["HERMES_HOME"])  # set by hermetic conftest fixture
+        _write_plugin(
+            hermes_home / "plugins",
+            ["p1"],
+            manifest_extra={"kind": "bogus"},
+        )
+        _enable(hermes_home, "p1")
+
+        with caplog.at_level("WARNING"):
+            mgr = PluginManager()
+            mgr.discover_and_load()
+
+        assert mgr._plugins["p1"].manifest.kind == "standalone"
+        assert any(
+            "unknown kind" in rec.getMessage() for rec in caplog.records
+        )
+
+
+# ── Gate logic ─────────────────────────────────────────────────────────────
+
+
+class TestBackendGate:
+    def test_user_backend_still_gated_by_enabled(self, tmp_path, monkeypatch):
+        """User-installed ``kind: backend`` plugins still require opt-in —
+        they're not trusted by default."""
+        import os
+        hermes_home = Path(os.environ["HERMES_HOME"])  # set by hermetic conftest fixture
+        user_plugins = hermes_home / "plugins"
+
+        _write_plugin(
+            user_plugins,
+            ["image_gen", "fancy"],
+            manifest_extra={"kind": "backend"},
+        )
+        # Do NOT opt in.
+
+        mgr = PluginManager()
+        mgr.discover_and_load()
+
+        loaded = mgr._plugins["image_gen/fancy"]
+        assert loaded.enabled is False
+        assert "not enabled" in (loaded.error or "")
+
+    def test_user_backend_loads_when_enabled(self, tmp_path, monkeypatch):
+        import os
+        hermes_home = Path(os.environ["HERMES_HOME"])  # set by hermetic conftest fixture
+        user_plugins = hermes_home / "plugins"
+
+        _write_plugin(
+            user_plugins,
+            ["image_gen", "fancy"],
+            manifest_extra={"kind": "backend"},
+        )
+        _enable(hermes_home, "image_gen/fancy")
+
+        mgr = PluginManager()
+        mgr.discover_and_load()
+
+        assert mgr._plugins["image_gen/fancy"].enabled is True
+
+    def test_exclusive_kind_skipped(self, tmp_path, monkeypatch):
+        """``kind: exclusive`` plugins are recorded but not loaded — the
+        category's own discovery system handles them (memory today)."""
+        import os
+        hermes_home = Path(os.environ["HERMES_HOME"])  # set by hermetic conftest fixture
+        _write_plugin(
+            hermes_home / "plugins",
+            ["some-backend"],
+            manifest_extra={"kind": "exclusive"},
+        )
+        _enable(hermes_home, "some-backend")
+
+        mgr = PluginManager()
+        mgr.discover_and_load()
+
+        loaded = mgr._plugins["some-backend"]
+        assert loaded.enabled is False
+        assert "exclusive" in (loaded.error or "")
+
+
+# ── Bundled backend auto-load (integration with real bundled plugin) ────────
+
+
+class TestBundledBackendAutoLoad:
+    def test_bundled_image_gen_openai_autoloads(self, tmp_path, monkeypatch):
+        """The bundled ``plugins/image_gen/openai/`` plugin loads without
+        any opt-in — it's ``kind: backend`` and shipped in-repo."""
+        import os
+        hermes_home = Path(os.environ["HERMES_HOME"])  # set by hermetic conftest fixture
+
+        mgr = PluginManager()
+        mgr.discover_and_load()
+
+        assert "image_gen/openai" in mgr._plugins
+        loaded = mgr._plugins["image_gen/openai"]
+        assert loaded.manifest.source == "bundled"
+        assert loaded.manifest.kind == "backend"
+        assert loaded.enabled is True, f"error: {loaded.error}"
+
+
+# ── PluginContext.register_image_gen_provider ───────────────────────────────
+
+
+class TestRegisterImageGenProvider:
+    def test_accepts_valid_provider(self, tmp_path, monkeypatch):
+        from agent import image_gen_registry
+        from agent.image_gen_provider import ImageGenProvider
+
+        image_gen_registry._reset_for_tests()
+
+        class FakeProvider(ImageGenProvider):
+            @property
+            def name(self) -> str:
+                return "fake-test"
+
+            def generate(self, prompt, aspect_ratio="landscape", **kw):
+                return {"success": True, "image": "test://fake"}
+
+        import os
+        hermes_home = Path(os.environ["HERMES_HOME"])  # set by hermetic conftest fixture
+        plugin_dir = _write_plugin(
+            hermes_home / "plugins",
+            ["my-img-plugin"],
+            register_body=(
+                "from agent.image_gen_provider import ImageGenProvider\n"
+                "    class P(ImageGenProvider):\n"
+                "        @property\n"
+                "        def name(self): return 'fake-ctx'\n"
+                "        def generate(self, prompt, aspect_ratio='landscape', **kw):\n"
+                "            return {'success': True, 'image': 'x://y'}\n"
+                "    ctx.register_image_gen_provider(P())"
+            ),
+        )
+        _enable(hermes_home, "my-img-plugin")
+
+        mgr = PluginManager()
+        mgr.discover_and_load()
+
+        assert mgr._plugins["my-img-plugin"].enabled is True
+        assert image_gen_registry.get_provider("fake-ctx") is not None
+
+        image_gen_registry._reset_for_tests()
+
+    def test_rejects_non_provider(self, tmp_path, monkeypatch, caplog):
+        from agent import image_gen_registry
+
+        image_gen_registry._reset_for_tests()
+
+        import os
+        hermes_home = Path(os.environ["HERMES_HOME"])  # set by hermetic conftest fixture
+        _write_plugin(
+            hermes_home / "plugins",
+            ["bad-img-plugin"],
+            register_body="ctx.register_image_gen_provider('not a provider')",
+        )
+        _enable(hermes_home, "bad-img-plugin")
+
+        with caplog.at_level("WARNING"):
+            mgr = PluginManager()
+            mgr.discover_and_load()
+
+        # Plugin loaded (register returned normally) but nothing was
+        # registered in the provider registry.
+        assert mgr._plugins["bad-img-plugin"].enabled is True
+        assert image_gen_registry.get_provider("not a provider") is None
+
+        image_gen_registry._reset_for_tests()

tests/plugins/image_gen/test_openai_provider.py

@@ -0,0 +1,243 @@
+"""Tests for the bundled OpenAI image_gen plugin (gpt-image-2, three tiers)."""
+
+from __future__ import annotations
+
+from pathlib import Path
+from types import SimpleNamespace
+from unittest.mock import MagicMock, patch
+
+import pytest
+
+import plugins.image_gen.openai as openai_plugin
+
+
+# 1×1 transparent PNG — valid bytes for save_b64_image()
+_PNG_HEX = (
+    "89504e470d0a1a0a0000000d49484452000000010000000108060000001f15c4"
+    "890000000d49444154789c6300010000000500010d0a2db40000000049454e44"
+    "ae426082"
+)
+
+
+def _b64_png() -> str:
+    import base64
+    return base64.b64encode(bytes.fromhex(_PNG_HEX)).decode()
+
+
+def _fake_response(*, b64=None, url=None, revised_prompt=None):
+    item = SimpleNamespace(b64_json=b64, url=url, revised_prompt=revised_prompt)
+    return SimpleNamespace(data=[item])
+
+
+@pytest.fixture(autouse=True)
+def _tmp_hermes_home(tmp_path, monkeypatch):
+    monkeypatch.setenv("HERMES_HOME", str(tmp_path))
+    yield tmp_path
+
+
+@pytest.fixture
+def provider(monkeypatch):
+    monkeypatch.setenv("OPENAI_API_KEY", "test-key")
+    return openai_plugin.OpenAIImageGenProvider()
+
+
+def _patched_openai(fake_client: MagicMock):
+    fake_openai = MagicMock()
+    fake_openai.OpenAI.return_value = fake_client
+    return patch.dict("sys.modules", {"openai": fake_openai})
+
+
+# ── Metadata ────────────────────────────────────────────────────────────────
+
+
+class TestMetadata:
+    def test_name(self, provider):
+        assert provider.name == "openai"
+
+    def test_default_model(self, provider):
+        assert provider.default_model() == "gpt-image-2-medium"
+
+    def test_list_models_three_tiers(self, provider):
+        ids = [m["id"] for m in provider.list_models()]
+        assert ids == ["gpt-image-2-low", "gpt-image-2-medium", "gpt-image-2-high"]
+
+    def test_catalog_entries_have_display_speed_strengths(self, provider):
+        for entry in provider.list_models():
+            assert entry["display"].startswith("GPT Image 2")
+            assert entry["speed"]
+            assert entry["strengths"]
+
+
+# ── Availability ────────────────────────────────────────────────────────────
+
+
+class TestAvailability:
+    def test_no_api_key_unavailable(self, monkeypatch):
+        monkeypatch.delenv("OPENAI_API_KEY", raising=False)
+        assert openai_plugin.OpenAIImageGenProvider().is_available() is False
+
+    def test_api_key_set_available(self, monkeypatch):
+        monkeypatch.setenv("OPENAI_API_KEY", "test")
+        assert openai_plugin.OpenAIImageGenProvider().is_available() is True
+
+
+# ── Model resolution ────────────────────────────────────────────────────────
+
+
+class TestModelResolution:
+    def test_default_is_medium(self):
+        model_id, meta = openai_plugin._resolve_model()
+        assert model_id == "gpt-image-2-medium"
+        assert meta["quality"] == "medium"
+
+    def test_env_var_override(self, monkeypatch):
+        monkeypatch.setenv("OPENAI_IMAGE_MODEL", "gpt-image-2-high")
+        model_id, meta = openai_plugin._resolve_model()
+        assert model_id == "gpt-image-2-high"
+        assert meta["quality"] == "high"
+
+    def test_env_var_unknown_falls_back(self, monkeypatch):
+        monkeypatch.setenv("OPENAI_IMAGE_MODEL", "bogus-tier")
+        model_id, _ = openai_plugin._resolve_model()
+        assert model_id == openai_plugin.DEFAULT_MODEL
+
+    def test_config_openai_model(self, tmp_path):
+        import yaml
+        (tmp_path / "config.yaml").write_text(
+            yaml.safe_dump({"image_gen": {"openai": {"model": "gpt-image-2-low"}}})
+        )
+        model_id, meta = openai_plugin._resolve_model()
+        assert model_id == "gpt-image-2-low"
+        assert meta["quality"] == "low"
+
+    def test_config_top_level_model(self, tmp_path):
+        """``image_gen.model: gpt-image-2-high`` also works (top-level)."""
+        import yaml
+        (tmp_path / "config.yaml").write_text(
+            yaml.safe_dump({"image_gen": {"model": "gpt-image-2-high"}})
+        )
+        model_id, meta = openai_plugin._resolve_model()
+        assert model_id == "gpt-image-2-high"
+        assert meta["quality"] == "high"
+
+
+# ── Generate ────────────────────────────────────────────────────────────────
+
+
+class TestGenerate:
+    def test_empty_prompt_rejected(self, provider):
+        result = provider.generate("", aspect_ratio="square")
+        assert result["success"] is False
+        assert result["error_type"] == "invalid_argument"
+
+    def test_missing_api_key(self, monkeypatch):
+        monkeypatch.delenv("OPENAI_API_KEY", raising=False)
+        result = openai_plugin.OpenAIImageGenProvider().generate("a cat")
+        assert result["success"] is False
+        assert result["error_type"] == "auth_required"
+
+    def test_b64_saves_to_cache(self, provider, tmp_path):
+        import base64
+        png_bytes = bytes.fromhex(_PNG_HEX)
+        fake_client = MagicMock()
+        fake_client.images.generate.return_value = _fake_response(b64=_b64_png())
+
+        with _patched_openai(fake_client):
+            result = provider.generate("a cat", aspect_ratio="landscape")
+
+        assert result["success"] is True
+        assert result["model"] == "gpt-image-2-medium"
+        assert result["aspect_ratio"] == "landscape"
+        assert result["provider"] == "openai"
+        assert result["quality"] == "medium"
+
+        saved = Path(result["image"])
+        assert saved.exists()
+        assert saved.parent == tmp_path / "cache" / "images"
+        assert saved.read_bytes() == png_bytes
+
+        call_kwargs = fake_client.images.generate.call_args.kwargs
+        # All tiers hit the single underlying API model.
+        assert call_kwargs["model"] == "gpt-image-2"
+        assert call_kwargs["quality"] == "medium"
+        assert call_kwargs["size"] == "1536x1024"
+        # gpt-image-2 rejects response_format — we must NOT send it.
+        assert "response_format" not in call_kwargs
+
+    @pytest.mark.parametrize("tier,expected_quality", [
+        ("gpt-image-2-low", "low"),
+        ("gpt-image-2-medium", "medium"),
+        ("gpt-image-2-high", "high"),
+    ])
+    def test_tier_maps_to_quality(self, provider, monkeypatch, tier, expected_quality):
+        monkeypatch.setenv("OPENAI_IMAGE_MODEL", tier)
+        fake_client = MagicMock()
+        fake_client.images.generate.return_value = _fake_response(b64=_b64_png())
+
+        with _patched_openai(fake_client):
+            result = provider.generate("a cat")
+
+        assert result["model"] == tier
+        assert result["quality"] == expected_quality
+        assert fake_client.images.generate.call_args.kwargs["quality"] == expected_quality
+        # Always the same underlying API model regardless of tier.
+        assert fake_client.images.generate.call_args.kwargs["model"] == "gpt-image-2"
+
+    @pytest.mark.parametrize("aspect,expected_size", [
+        ("landscape", "1536x1024"),
+        ("square", "1024x1024"),
+        ("portrait", "1024x1536"),
+    ])
+    def test_aspect_ratio_mapping(self, provider, aspect, expected_size):
+        fake_client = MagicMock()
+        fake_client.images.generate.return_value = _fake_response(b64=_b64_png())
+
+        with _patched_openai(fake_client):
+            provider.generate("a cat", aspect_ratio=aspect)
+
+        assert fake_client.images.generate.call_args.kwargs["size"] == expected_size
+
+    def test_revised_prompt_passed_through(self, provider):
+        fake_client = MagicMock()
+        fake_client.images.generate.return_value = _fake_response(
+            b64=_b64_png(), revised_prompt="A photo of a cat",
+        )
+
+        with _patched_openai(fake_client):
+            result = provider.generate("a cat")
+
+        assert result["revised_prompt"] == "A photo of a cat"
+
+    def test_api_error_returns_error_response(self, provider):
+        fake_client = MagicMock()
+        fake_client.images.generate.side_effect = RuntimeError("boom")
+
+        with _patched_openai(fake_client):
+            result = provider.generate("a cat")
+
+        assert result["success"] is False
+        assert result["error_type"] == "api_error"
+        assert "boom" in result["error"]
+
+    def test_empty_response_data(self, provider):
+        fake_client = MagicMock()
+        fake_client.images.generate.return_value = SimpleNamespace(data=[])
+
+        with _patched_openai(fake_client):
+            result = provider.generate("a cat")
+
+        assert result["success"] is False
+        assert result["error_type"] == "empty_response"
+
+    def test_url_fallback_if_api_changes(self, provider):
+        """Defensive: if OpenAI ever returns URL instead of b64, pass through."""
+        fake_client = MagicMock()
+        fake_client.images.generate.return_value = _fake_response(
+            b64=None, url="https://example.com/img.png",
+        )
+
+        with _patched_openai(fake_client):
+            result = provider.generate("a cat")
+
+        assert result["success"] is True
+        assert result["image"] == "https://example.com/img.png"

tools/image_generation_tool.py

@@ -774,13 +774,40 @@ def check_fal_api_key() -> bool:
 
 
 def check_image_generation_requirements() -> bool:
-    """True if FAL credentials and fal_client SDK are both available."""
+    """True if any image gen backend is available.
+
+    Providers are considered in this order:
+
+    1. The in-tree FAL backend (FAL_KEY or managed gateway).
+    2. Any plugin-registered provider whose ``is_available()`` returns True.
+
+    Plugins win only when the in-tree FAL path is NOT ready, which matches
+    the historical behavior: shipping hermes with a FAL key configured
+    should still expose the tool. The active selection among ready
+    providers is resolved per-call by ``image_gen.provider``.
+    """
     try:
-        if not check_fal_api_key():
+        if check_fal_api_key():
-            return False
             fal_client  # noqa: F401 — SDK presence check
             return True
     except ImportError:
+        pass
+
+    # Probe plugin providers. Discovery is idempotent and cheap.
+    try:
+        from agent.image_gen_registry import list_providers
+        from hermes_cli.plugins import _ensure_plugins_discovered
+
+        _ensure_plugins_discovered()
+        for provider in list_providers():
+            try:
+                if provider.is_available():
+                    return True
+            except Exception:
+                continue
+    except Exception:
+        pass
+
     return False
 
 
@@ -827,10 +854,11 @@ from tools.registry import registry, tool_error
 IMAGE_GENERATE_SCHEMA = {
     "name": "image_generate",
     "description": (
-        "Generate high-quality images from text prompts using FAL.ai. "
+        "Generate high-quality images from text prompts. The underlying "
-        "The underlying model is user-configured (default: FLUX 2 Klein 9B, "
+        "backend (FAL, OpenAI, etc.) and model are user-configured and not "
-        "sub-1s generation) and is not selectable by the agent. Returns a "
+        "selectable by the agent. Returns either a URL or an absolute file "
-        "single image URL. Display it using markdown: ![description](URL)"
+        "path in the `image` field; display it with markdown "
+        "![description](url-or-path) and the gateway will deliver it."
     ),
     "parameters": {
         "type": "object",
@@ -851,13 +879,104 @@ IMAGE_GENERATE_SCHEMA = {
 }
 
 
+def _read_configured_image_provider():
+    """Return the value of ``image_gen.provider`` from config.yaml, or None.
+
+    We only consult the plugin registry when this is explicitly set — an
+    unset value keeps users on the legacy in-tree FAL path even when other
+    providers happen to be registered (e.g. a user has OPENAI_API_KEY set
+    for other features but never asked for OpenAI image gen).
+    """
+    try:
+        from hermes_cli.config import load_config
+        cfg = load_config()
+        section = cfg.get("image_gen") if isinstance(cfg, dict) else None
+        if isinstance(section, dict):
+            value = section.get("provider")
+            if isinstance(value, str) and value.strip():
+                return value.strip()
+    except Exception as exc:
+        logger.debug("Could not read image_gen.provider: %s", exc)
+    return None
+
+
+def _dispatch_to_plugin_provider(prompt: str, aspect_ratio: str):
+    """Route the call to a plugin-registered provider when one is selected.
+
+    Returns a JSON string on dispatch, or ``None`` to fall through to the
+    built-in FAL path.
+
+    Dispatch only fires when ``image_gen.provider`` is explicitly set AND
+    it does not point to ``fal`` (FAL still lives in-tree in this PR;
+    a later PR ports it into ``plugins/image_gen/fal/``). Any other value
+    that matches a registered plugin provider wins.
+    """
+    configured = _read_configured_image_provider()
+    if not configured or configured == "fal":
+        return None
+
+    try:
+        # Import locally so plugin discovery isn't triggered just by
+        # importing this module (tests rely on that).
+        from agent.image_gen_registry import get_provider
+        from hermes_cli.plugins import _ensure_plugins_discovered
+
+        _ensure_plugins_discovered()
+        provider = get_provider(configured)
+    except Exception as exc:
+        logger.debug("image_gen plugin dispatch skipped: %s", exc)
+        return None
+
+    if provider is None:
+        return json.dumps({
+            "success": False,
+            "image": None,
+            "error": (
+                f"image_gen.provider='{configured}' is set but no plugin "
+                f"registered that name. Run `hermes plugins list` to see "
+                f"available image gen backends."
+            ),
+            "error_type": "provider_not_registered",
+        })
+
+    try:
+        result = provider.generate(prompt=prompt, aspect_ratio=aspect_ratio)
+    except Exception as exc:
+        logger.warning(
+            "Image gen provider '%s' raised: %s",
+            getattr(provider, "name", "?"), exc,
+        )
+        return json.dumps({
+            "success": False,
+            "image": None,
+            "error": f"Provider '{getattr(provider, 'name', '?')}' error: {exc}",
+            "error_type": "provider_exception",
+        })
+    if not isinstance(result, dict):
+        return json.dumps({
+            "success": False,
+            "image": None,
+            "error": "Provider returned a non-dict result",
+            "error_type": "provider_contract",
+        })
+    return json.dumps(result)
+
+
 def _handle_image_generate(args, **kw):
     prompt = args.get("prompt", "")
     if not prompt:
         return tool_error("prompt is required for image generation")
+    aspect_ratio = args.get("aspect_ratio", DEFAULT_ASPECT_RATIO)
+
+    # Route to a plugin-registered provider if one is active (and it's
+    # not the in-tree FAL path).
+    dispatched = _dispatch_to_plugin_provider(prompt, aspect_ratio)
+    if dispatched is not None:
+        return dispatched
+
     return image_generate_tool(
         prompt=prompt,
-        aspect_ratio=args.get("aspect_ratio", DEFAULT_ASPECT_RATIO),
+        aspect_ratio=aspect_ratio,
     )