Merge branch 'master' into explore-page-fix

Author: avi-007

bots/admin.py

@@ -445,11 +445,12 @@ class SavedRunAdmin(admin.ModelAdmin):
 
     readonly_fields = [
         "open_in_gooey",
-        "parent",
+        "view_parent_published_run",
         "view_bots",
         "price",
         "view_usage_cost",
         "transaction",
+        "parent",
         "created_at",
         "updated_at",
         "run_time",
@@ -485,7 +486,10 @@ def view_user(self, saved_run: SavedRun):
     view_user.short_description = "View User"
 
     def view_bots(self, saved_run: SavedRun):
-        return list_related_html_url(saved_run.botintegrations)
+        pr = saved_run.parent_published_run()
+        if not pr:
+            raise SavedRun.DoesNotExist
+        return list_related_html_url(pr.botintegrations)
 
     view_bots.short_description = "View Bots"
 

bots/models/saved_run.py

@@ -56,6 +56,7 @@ class SavedRun(models.Model):
         null=True,
         blank=True,
         related_name="children",
+        verbose_name="Parent Run",
     )
     parent_version = models.ForeignKey(
         "bots.PublishedRunVersion",

daras_ai_v2/asr.py

@@ -121,6 +121,17 @@
     "srp", "swe", "swh", "tam", "tel", "tgk", "tgl", "tha", "tur", "ukr", "urd", "uzn", "vie", "yor", "yue", "zul",
 }  # fmt: skip
 
+# Eleven Labs Scribe v1 - supports 99 languages with 3-letter ISO codes
+ELEVENLABS_SUPPORTED = {
+    "afr", "amh", "ara", "hye", "asm", "ast", "aze", "bel", "ben", "bos", "bul", "mya", "yue", "cat", "ceb", "nya",
+    "hrv", "ces", "dan", "nld", "eng", "est", "fil", "fin", "fra", "ful", "glg", "lug", "kat", "deu", "ell", "guj",
+    "hau", "heb", "hin", "hun", "isl", "ibo", "ind", "gle", "ita", "jpn", "jav", "kea", "kan", "kaz", "khm", "kor",
+    "kur", "kir", "lao", "lav", "lin", "lit", "luo", "ltz", "mkd", "msa", "mal", "mlt", "zho", "mri", "mar", "mon",
+    "nep", "nso", "nor", "oci", "ori", "pus", "fas", "pol", "por", "pan", "ron", "rus", "srp", "sna", "snd", "slk",
+    "slv", "som", "spa", "swa", "swe", "tam", "tgk", "tel", "tha", "tur", "ukr", "umb", "urd", "uzb", "vie", "cym",
+    "wol", "xho", "zul",
+}  # fmt: skip
+
 AZURE_SUPPORTED = {
     "af-ZA", "am-ET", "ar-AE", "ar-BH", "ar-DZ", "ar-EG", "ar-IL", "ar-IQ", "ar-JO", "ar-KW", "ar-LB", "ar-LY", "ar-MA",
     "ar-OM", "ar-PS", "ar-QA", "ar-SA", "ar-SY", "ar-TN", "ar-YE", "az-AZ", "bg-BG", "bn-IN", "bs-BA", "ca-ES", "cs-CZ",
@@ -260,6 +271,7 @@ class AsrModels(Enum):
     usm = "Chirp / USM (Google V2)"
     deepgram = "Deepgram"
     azure = "Azure Speech"
+    elevenlabs = "ElevenLabs Scribe v1"
     seamless_m4t_v2 = "Seamless M4T v2 (Facebook Research)"
     mms_1b_all = "Massively Multilingual Speech (MMS) (Facebook Research)"
 
@@ -329,6 +341,7 @@ def supports_input_prompt(self) -> bool:
     AsrModels.seamless_m4t_v2: "facebook/seamless-m4t-v2-large",
     AsrModels.mms_1b_all: "facebook/mms-1b-all",
     AsrModels.lelapa: "lelapa-vulavula",
+    AsrModels.elevenlabs: "elevenlabs-scribe-v1",
 }
 
 forced_asr_languages = {
@@ -354,6 +367,7 @@ def supports_input_prompt(self) -> bool:
     AsrModels.gcp_v1: GCP_V1_SUPPORTED,
     AsrModels.usm: CHIRP_SUPPORTED,
     AsrModels.deepgram: DEEPGRAM_SUPPORTED,
+    AsrModels.elevenlabs: ELEVENLABS_SUPPORTED,
     AsrModels.seamless_m4t_v2: SEAMLESS_v2_ASR_SUPPORTED,
     AsrModels.azure: AZURE_SUPPORTED,
     AsrModels.mms_1b_all: MMS_SUPPORTED,
@@ -971,6 +985,33 @@ def get_google_auth_session(*scopes: str) -> tuple[AuthorizedSession, str]:
         return AuthorizedSession(credentials=creds), project
 
 
+def elevenlabs_asr(audio_url: str, language: str = None) -> dict:
+    """
+    Call ElevenLabs Speech-to-Text API
+    """
+    audio_r = requests.get(audio_url)
+    raise_for_status(audio_r, is_user_url=True)
+
+    # Set up the files and form data for the multipart request
+    files = {"file": audio_r.content}
+    data = {"model_id": "scribe_v1"}
+    headers = {"xi-api-key": settings.ELEVEN_LABS_API_KEY}
+
+    # Language parameter is sent in the form data
+    if language:
+        data["language_code"] = language
+
+    response = requests.post(
+        "https://api.elevenlabs.io/v1/speech-to-text",
+        files=files,
+        headers=headers,
+        data=data,
+    )
+    raise_for_status(response)
+
+    return response.json()
+
+
 def run_asr(
     audio_url: str,
     selected_model: str,
@@ -1017,6 +1058,21 @@ def run_asr(
 
     if selected_model == AsrModels.azure:
         return azure_asr(audio_url, language)
+    elif selected_model == AsrModels.elevenlabs:
+        result = elevenlabs_asr(audio_url, language)
+        chunks = []
+        for word_data in result.get("words", []):
+            if word_data.get("type") == "word":
+                speaker = word_data.get("speaker_id", 0)
+            else:
+                speaker = None
+            chunk = {
+                "timestamp": (word_data["start"], word_data["end"]),
+                "text": word_data["text"],
+                "speaker": speaker,
+            }
+            chunks.append(chunk)
+        data = {"text": result["text"], "chunks": chunks}
     elif selected_model == AsrModels.whisper_large_v3:
         import replicate
 

daras_ai_v2/language_model.py

@@ -122,6 +122,8 @@ class LargeLanguageModels(Enum):
         price=1,
         is_vision_model=True,
         supports_json=True,
+        is_deprecated=True,
+        redirect_to="gpt_4_o",
     )
 
     # https://platform.openai.com/docs/models/o4-mini

daras_ai_v2/stable_diffusion.py

@@ -117,6 +117,7 @@ def _deprecated(cls):
             cls.openjourney,
             cls.analog_diffusion,
             cls.protogen_5_3,
+            cls.dall_e,
         }
 
 
@@ -444,18 +445,6 @@ def _get_gpt_image_1_img_size(width: int, height: int) -> tuple[int, int]:
         return 1024, 1536
 
 
-def prepare_init_image(
-    init_image_bytes: bytes, width: int, height: int
-) -> tuple[bytes, bytes]:
-    image = resize_img_pad(init_image_bytes, (width, height))
-    image = rgb_img_to_rgba(image)
-    mask = io.BytesIO()
-    Image.new("RGBA", (width, height), (0, 0, 0, 0)).save(mask, format="PNG")
-    mask = mask.getvalue()
-
-    return image, mask
-
-
 def img2img(
     *,
     selected_model: str,
@@ -468,12 +457,13 @@ def img2img(
     negative_prompt: str = None,
     guidance_scale: float,
     seed: int = 42,
+    gpt_image_1_quality: typing.Literal["low", "medium", "high"] | None = None,
 ):
     prompt_strength = prompt_strength or 0.7
     assert 0 <= prompt_strength <= 0.9, "Prompt Strength must be in range [0, 0.9]"
 
     match selected_model:
-        case Img2ImgModels.dall_e.name | Img2ImgModels.gpt_image_1.name:
+        case Img2ImgModels.gpt_image_1.name:
             from openai import NOT_GIVEN, OpenAI
 
             init_height, init_width, _ = bytes_to_cv2_img(init_image_bytes).shape
@@ -487,20 +477,19 @@ def img2img(
                 width, height = _get_gpt_image_1_img_size(init_width, init_height)
                 response_format = NOT_GIVEN
 
-            image, mask = prepare_init_image(
-                init_image_bytes, width=width, height=height
-            )
+            image = resize_img_pad(init_image_bytes, (width, height))
+            image = rgb_img_to_rgba(image)
 
             client = OpenAI()
             with capture_openai_content_policy_violation():
                 response = client.images.edit(
                     model=img2img_model_ids[Img2ImgModels[selected_model]],
                     prompt=prompt,
                     image=("image.png", image),
-                    mask=("mask.png", mask),
                     n=num_outputs,
                     size=f"{width}x{height}",
                     response_format=response_format,
+                    quality=gpt_image_1_quality,
                 )
 
             # Record usage costs if usage data is available

poetry.lock

@@ -82,7 +82,7 @@ anthropic = "^0.34.1"
 azure-cognitiveservices-speech = "^1.37.0"
 twilio = "^9.2.3"
 sentry-sdk = {version = "1.45.0", extras = ["loguru"]}
-gooey-gui = "0.5.5"
+gooey-gui = "0.6.0"
 django-safedelete = "^1.4.0"
 numexpr = "^2.10.1"
 django-csp = "^3.8"

pyproject.toml

@@ -102,13 +102,7 @@ def _render_results(results: list[AggFunctionResult]):
             + [
                 [
                     columns[i],
-                    dict(
-                        kind="number",
-                        readonly=True,
-                        displayData=str(values[i]),
-                        data=values[i],
-                        themeOverride=dict(bgCell=colors[i]),
-                    ),
+                    dict(value=values[i], style=dict(backgroundColor=colors[i])),
                     g[i].get("count", 1),
                 ]
                 for i in range(len(g))