@@ -325,6 +325,7 @@ GOOGLE_API_KEY = env.str('GOOGLE_API_KEY', default='defaultapikey')
 SERPER_API_KEY = env.str('SERPER_API_KEY', 'defaultapikey')
 FLUX_API_KEY = env.str('FLUX_API_KEY', 'defaultapikey')
 OPENROUTER_API_KEY = env.str('OPENROUTER_API_KEY', 'defaultapikey')
+FAL_API_KEY = env.str('FAL_API_KEY', 'defaultapikey')
 
 OPENAI_PROXY_HOST = env.str('OPENAI_PROXY_HOST', 'neuron-proxy:8080')
 UPSCALE_MULTIPLIER_HOST = env.str('UPSCALE_MULTIPLIER_HOST', 'packet:8080')
@@ -9,8 +9,8 @@ def migrate_nm_type(apps, schema_editor):
     NeuronModel = apps.get_model('ml_model', 'NeuronModel')
     models = NeuronModel.objects.all()
     for model in models:
-        model.type = model.category.slug
-    NeuronModel.objects.bulk_update(models, fields=['type'])
+        model.types = [model.category.slug]
+    NeuronModel.objects.bulk_update(models, fields=['types'])
 
 def migrate_versions_to_deployments(apps, schema_editor):
     Deployment = apps.get_model('ml_model', 'Deployment')
@@ -0,0 +1,18 @@
+# Generated by Django 5.0.11 on 2025-05-27 07:29
+
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('ml_model', '0055_remove_neuronmodel_category_and_more'),
+    ]
+
+    operations = [
+        migrations.AlterField(
+            model_name='deployment',
+            name='runner_import_path',
+            field=models.CharField(choices=[('ml_model.runners:FalAIRunner', 'FalAI'), ('ml_model.runners:OpenAIGPTRunner', 'OpenAIGPT'), ('ml_model.runners:OpenrouterRunner', 'Openrouter'), ('ml_model.runners:ReplicateAudioRunner', 'ReplicateAudio'), ('ml_model.runners:ReplicateImageRunner', 'ReplicateImage'), ('ml_model.runners:ReplicateTextRunner', 'ReplicateText'), ('ml_model.runners:ReplicateVideoRunner', 'ReplicateVideo')], max_length=100, verbose_name='Runner'),
+        ),
+    ]
@@ -6,6 +6,7 @@ from ml_model.runners.replicate import (
     ReplicateTextRunner,
     ReplicateVideoRunner,
 )
+from ml_model.runners.falai import FalAIRunner
 
 __all__ = [
     'OpenAIGPTRunner',
@@ -14,4 +15,5 @@ __all__ = [
     'ReplicateAudioRunner',
     'ReplicateImageRunner',
     'ReplicateVideoRunner',
+    'FalAIRunner'
 ]
@@ -0,0 +1,47 @@
+import logging
+from io import BytesIO, StringIO
+from typing import Any, Iterable, Generator
+
+import httpx
+from django.conf import settings
+
+from ml_model.runners.base import BaseRunner
+
+logger = logging.getLogger(__name__)
+
+
+class FalAIRunner(BaseRunner):
+    @classmethod
+    def generate(
+            cls,
+            content: str | None = None,
+            file: BytesIO | StringIO | None = None,
+            parameters: dict[str, Any] = {},
+            history: Iterable['Message'] = [],
+            scrape_results: list[StringIO] | list[BytesIO] = [],
+    ) -> Generator[str, Any, None]:
+        try:
+            version = parameters.pop('model')
+            model_owner = parameters.pop('model_owner')
+            official = version and model_owner
+            payload = {'prompt': content, **parameters}
+            if not official:
+                payload.update({'model_version': version})
+            with httpx.Client(
+                    base_url="https://fal.run/",
+                    headers={"Authorization": f"Key {settings.FAL_API_KEY}"},
+                    timeout=600,
+            ) as client:
+                with client.stream(
+                        'POST', f'{model_owner}/{version}/stream', json=payload, headers={'Accept': 'text/event-stream'}
+                ) as stream:
+                    for chunk in stream.iter_text():
+                        if 'images' in chunk:
+                            yield f'\n{chunk.split('"')[-1]}'
+                        elif '"' in chunk:
+                            yield chunk.split('\"')[0]
+                        else:
+                            yield chunk
+        except httpx.TimeoutException as exc:
+            logger.exception(exc)
+            raise Exception('Timeout happened')
@@ -128,20 +128,26 @@ class MediaAPIView(APIView):
         serializer = MessageSerializer(data=request.data)
         try:
             serializer.is_valid(raise_exception=True)
-            chat = self.manager.objects.prefetch_related('model', 'model__inferences').get(
-                user=request.user, model__slug=model
+            gallery, _ = self.manager.objects.get_or_create(
+                user=request.user,
+                model__slug=model,
+                defaults={
+                    'user': request.user,
+                    'model': NeuronModel.objects.get(slug=model),
+                },
             )
+            gallery = self.manager.objects.prefetch_related('model', 'model__inferences').get(pk=gallery.pk)
             info = serializer.validated_data.pop('info')
             inference_slug = info.pop('inference')
             if not any(
                 [
                     inference.slug == inference_slug and inference.enabled
-                    for inference in chat.model.inferences.all()
+                    for inference in gallery.model.inferences.all()
                 ]
             ):
                 raise Exception(_('Enabled inference not found in model'))
             input_message = Message.objects.create(
-                **serializer.validated_data, info=info, content_object=chat, from_model=False
+                **serializer.validated_data, info=info, content_object=gallery, from_model=False
             )
         except Exception as exc:
             return Response({'detail': str(exc)}, status=400)
@@ -154,7 +160,7 @@ class MediaAPIView(APIView):
                 history_ids=[
                     id
                     async for id in Message.objects.filter(
-                        object_id=chat.uid, content_type__model='chat', is_deleted=False, is_sent=True
+                        object_id=gallery.uid, content_type__model='chat', is_deleted=False, is_sent=True
                     )[:10].values_list('uid', flat=True)
                 ],
             )
@@ -171,7 +177,8 @@ class MediaAPIView(APIView):
                     if chunk:
                         content += chunk
                         yield f'id: {output_slot_id}\nevent: output\ndata: {"".join(chunk).replace("\n", "\\n")}\n\n'
-
+                    else:
+                        await asyncio.sleep(0.1)
                 yield f'id: {output_slot_id}\nevent: done\ndata: [DONE]\n\n'
             except Exception as exc:
                 input_message.is_sent = False