feat(images): add local Stable Diffusion provider with A1111 and ComfyUI support

Author: jddunn

src/api/model.ts

@@ -51,6 +51,7 @@ const ENV_URL_MAP: Record<string, string> = {
   stability: 'STABILITY_BASE_URL',
   replicate: 'REPLICATE_BASE_URL',
   ollama: 'OLLAMA_BASE_URL',
+  'stable-diffusion-local': 'STABLE_DIFFUSION_LOCAL_BASE_URL',
 };
 
 /**
@@ -160,6 +161,15 @@ export function resolveMediaProvider(
     return { providerId, modelId, baseUrl };
   }
 
+  if (providerId === 'stable-diffusion-local') {
+    if (!baseUrl) {
+      throw new Error(
+        `No base URL for stable-diffusion-local. Set STABLE_DIFFUSION_LOCAL_BASE_URL or pass baseUrl.`,
+      );
+    }
+    return { providerId, modelId, baseUrl };
+  }
+
   const envVar = ENV_KEY_MAP[providerId];
   if (envVar && !apiKey) {
     throw new Error(`No API key for ${providerId}. Set ${envVar} or pass apiKey.`);

src/api/provider-defaults.ts

@@ -59,6 +59,9 @@ export const PROVIDER_DEFAULTS: Record<string, ProviderDefaults> = {
   replicate: {
     image: 'black-forest-labs/flux-1.1-pro',
   },
+  'stable-diffusion-local': {
+    image: 'v1-5-pruned-emaonly',
+  },
 };
 
 /** Env var keys checked for auto-detection, in priority order */
@@ -70,6 +73,7 @@ const AUTO_DETECT_ORDER: Array<{ envKey: string; provider: string }> = [
   { envKey: 'OLLAMA_BASE_URL', provider: 'ollama' },
   { envKey: 'STABILITY_API_KEY', provider: 'stability' },
   { envKey: 'REPLICATE_API_TOKEN', provider: 'replicate' },
+  { envKey: 'STABLE_DIFFUSION_LOCAL_BASE_URL', provider: 'stable-diffusion-local' },
 ];
 
 /**

src/core/images/IImageProvider.ts

@@ -1,4 +1,4 @@
-export type ImageProviderId = 'openai' | 'openrouter' | 'stability' | 'replicate' | (string & {});
+export type ImageProviderId = 'openai' | 'openrouter' | 'stability' | 'replicate' | 'stable-diffusion-local' | (string & {});
 export type ImageModality = 'image' | 'text';
 export type ImageBackground = 'transparent' | 'opaque' | 'auto';
 export type ImageOutputFormat = 'png' | 'jpeg' | 'jpg' | 'webp';
@@ -72,11 +72,39 @@ export interface ReplicateImageProviderOptions {
   extraBody?: Record<string, unknown>;
 }
 
+export interface StableDiffusionLocalImageProviderOptions {
+  /** Number of inference steps (default 25). */
+  steps?: number;
+  /** Classifier-free guidance scale (default 7.5). */
+  cfgScale?: number;
+  /** Random seed (-1 for random). */
+  seed?: number;
+  /** Sampler name (e.g. 'Euler a', 'DPM++ 2M Karras'). */
+  sampler?: string;
+  /** Negative prompt. */
+  negativePrompt?: string;
+  /** Image width in pixels (default 512). */
+  width?: number;
+  /** Image height in pixels (default 512). */
+  height?: number;
+  /** Number of images to generate (default 1). */
+  batchSize?: number;
+  /** ControlNet settings forwarded verbatim to the backend. */
+  controlnet?: Record<string, unknown>;
+  /** LoRA models to apply.  Injected into the prompt as `<lora:name:weight>`. */
+  loras?: Array<{ name: string; weight?: number }>;
+  /** Enable high-resolution fix (A1111 only). */
+  hrFix?: boolean;
+  /** Denoising strength for high-res fix or img2img (default 0.7). */
+  denoisingStrength?: number;
+}
+
 export interface ImageProviderOptionBag {
   openai?: OpenAIImageProviderOptions;
   openrouter?: OpenRouterImageProviderOptions;
   stability?: StabilityImageProviderOptions;
   replicate?: ReplicateImageProviderOptions;
+  'stable-diffusion-local'?: StableDiffusionLocalImageProviderOptions;
   [providerId: string]: unknown;
 }
 
@@ -118,7 +146,7 @@ export interface IImageProvider {
   shutdown?(): Promise<void>;
 }
 
-const BUILT_IN_IMAGE_PROVIDER_IDS = new Set(['openai', 'openrouter', 'stability', 'replicate']);
+const BUILT_IN_IMAGE_PROVIDER_IDS = new Set(['openai', 'openrouter', 'stability', 'replicate', 'stable-diffusion-local']);
 
 export function getImageProviderOptions<T extends object>(
   providerId: string,