From 9befd9c28d4a348e4d46effb57e85458bdbfe9a9 Mon Sep 17 00:00:00 2001
From: David Kifer <dkifer19@gmail.com>
Date: Mon, 12 Jan 2026 16:01:22 -0500
Subject: [PATCH] added model c hange

---
 .../__pycache__/diarizer.cpython-313.pyc      | Bin 0 -> 3376 bytes
 .../ai_transcriber_v2/diarizer.py             |  24 +++++++--------
 .../documentation/HOW_TO_RUN_BACKGROUND.md    |  28 ++++++++++++++++++
 .../ai_transcriber_v2/translator.py           |   2 +-
 .../ai_transcriber_v2/utils.py                |   2 +-
 5 files changed, 40 insertions(+), 16 deletions(-)
 create mode 100644 video_transcription/ai_transcriber_v2/__pycache__/diarizer.cpython-313.pyc

diff --git a/video_transcription/ai_transcriber_v2/__pycache__/diarizer.cpython-313.pyc b/video_transcription/ai_transcriber_v2/__pycache__/diarizer.cpython-313.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..f530c1755660207f8b9f4c8559b3253ad7de03ee
GIT binary patch
literal 3376
zcmbUjOKcm*b(Y*ESIZ@-h_n<_AFfr|5+i9P)gM@F?Lx8X$W_D`Tr*KBr^JdJQ5#W9
zVwQAd6#>#-N}p^MHL4axqhkv6&|}b3dx*~>q(aQb1q{TO-egEHissTc%Uvy_I55x!
zIq%KP`_G&A<}er(5PaQtu=b-MLSNH~^LcFK@B$!@kb)Fu4v`jQ2xHAX#A87(@hYt1
ziJ=b}!YbYvB3zM4aQ~(v(s{2fi2^_T>HM$+<0CWxQWeh}`YEG(C6x7yP#X9O5<-hU
z3xy%WJD3YI0T*{Lg}cbKa0wOJSMY<sCqKpK@(-ZJkh7oZbTAiYA};P=Z5&Z&&4rn6
z7k97;#?_ehW==cnE>!$kkHSxTbhb@;%-wgeOuvJ<@S=@6QB}_(lw~t<m%_o?WJ}JP
z3kQ+X;n?Xw=~d_RviJ&F6hcbiyw{bIWe3p$qln-kH-U6tmdpIW)$CwOPzg;l5i~o2
zPz^GtvaWYzeHrlQ<d^kl&YfHXul*S(2yL6$4kawN*)l(cPyj(5`i9UMw8%Q7exwc}
zSbs%Ey(}Z!q{ksAm(dk(94#{-{f##1V+iTO{~^6ID}>N_j~mT6QdnJ>5sGXJE;EkC
z#A_QDgj<*_m&uwT85OLqVj>kZm1rAkRV(XK&Cv9sRJo_>dbx^awN}u|vOs7ZNH8XU
zORM0LrsIw26d`4jJV7Y|lcecJRV|fpL6)w!Nj6rsN(C0v8`o^!*;=s(+HR<MEZr`z
zVqF@Uy^*{9&Ri;;up}*)Yx?&{*?9BqvRa_Fw(D@rlSUj5CTu6i+ngLfFHa?8S>8Cd
z{(GqyldDq8Og3LDs2f5{8CARC1ql!mQLnA#Y$e8RSP0zHaH)_hsY|$&@R}S1gH%l~
z)(fWJrrU^F5s(TsLPfQimXs3RX2>zV6@Eff&<xyMtCUIAVl@3#qUPb~Okb&7ECS>(
zJr0xA^yM-&ndDc@pn-{|mh6)?JJmuVXQ<0KS1gs6)RJKaExO!YObj@BQ@l;t-nD1p
zWl~<tRR{)EaNq?(3$ZC`Mozt}YT#W7o1SXf^cC=3Essq>t72lAWtct{PFgRJ4qzd)
zxtJZP^*pZ73`+z|!OE&!MXjz7+6qj5WjSYsx#cQfv9rt!w1vj>6KvE<Rl{VfHKLnL
zj?lYiGT5L{azC?Q)pB5pZsdu!gh}r14<}lJFp(?w%%DBVS$2?N;1=O;l%a2Yf)03=
zi~frreIXz*`h*WZ2;C3u^L-7zZ>N80kIy&5k!Nh+*XduT_t~BX+p`n9v&a7Yz|+q~
zpYg$czPG{m?(<^}e(W!JU!G~mGy8J7A*Xl7-`V4D9r#eRZ$C2Hh>SKxx!Dyv;81kz
z0C}TAoj($h_}YWy{p5aNun`#C4-7Q|LrpRIVDA3hAGuG(KZx7!?!G>@CyqaJ?4sJX
zKKv~9nfTXhJAIe;_~|dZ2evjE-Kl!uX;-xFZ%U*4(xrxU>F*bIq)WTfY(wm;Uw`;s
zQ;a-F-%mGB_kSAueXQBj`@8kc^=9wjb3fY`dV$zr__-IIxyc-aP-jm)^q->?gT$Ug
z<l#;RWnqth|I2XXfP<}Xj%I=E)Jw?X)1RODJhIb&Wskr5Bpmt1BLA;J3*fJOsSx@{
z_}bOf74$_YkxF{Mm}cmBMYwU^``ABtL-szFIl%MIdjz=LS40ygk5HCzihBX5vS<kY
z&?|lhRn-ICli`3$_X6gzFnFQp9Mnf&mUfh%z@9(rx2br*T2vj2_FDp>h07Oh?NFuZ
z>&5yOZXT}&Tv~-)_}e%n{_&o;@HqKfc+r4_`O9FP2zvbrXIa9{&$K$uLPp_eGS1(z
zI}rGg@TnS@Ak1S@#D?_ViUu`=NCqyhVZCZdA8FN<Z<Rp1#7I{CF{CBczy%2kdsSP*
zQW@@gNv+6&Fhhz)QlNLnxhR!t?pW$bNi(X6<agCA{$kjwXTwQpQl1=>hT)n6K4sxm
zyk3RL@T@lc@t6eV@dHai+a3O{hU@Je!g1<bTBeS*Ov=-M+YVAi%B|GQTq={BoE-l6
z13}1OS{Q7%s0T?YrLU-Z9wOj+c_Mbl@}yOK4?3M&v;0|oRWE;}Tg}g^H3^UDT~pU#
z(DX#mjJQIN9W$dZa~Dc7rGUDfYqbNI(}o}}VFQ+0eb@|u?3{f>CI`iuasaRz2+S!*
z(B8LK1J0E0bvhI1D^@o*==001i+~aOEHclm(QL}KTT1ijxMo(WWE$v_@HhSh!zbuP
zHxi|0cyNawJoKTivvuxiSI?GzQ-%W5dFmnF6k(-Ds*BC;o_gTxQ;B-2+1<PK)}~&6
zs~J7Lb$0XZdg{oB`lW6EZr@aW{z<HV3-888>vK=ygWFTvtBrFryYV-7#J;ED_&>wr
z%>nsgpcx<9e!mf)_>K6y2X()35JP<<b#0e_{rM=^a3CWwzQYb2f`usINv9J&(^ED~
zuW`>H^x00j=s<6j<<6fGp&qqD!2fuyT&R`sRdN;9=p{D>U^wtH409MljPMnLU+60&
R9<gIg^zaUIj)`0R{{zB`DR2M)

literal 0
HcmV?d00001

diff --git a/video_transcription/ai_transcriber_v2/diarizer.py b/video_transcription/ai_transcriber_v2/diarizer.py
index 934324d..1a74307 100644
--- a/video_transcription/ai_transcriber_v2/diarizer.py
+++ b/video_transcription/ai_transcriber_v2/diarizer.py
@@ -6,31 +6,28 @@
 
 import os
 import sys
+import tracker
+import torch
 
 def diarize_audio(audio_path, num_speakers=None, hf_token=None):
     """
     Performs speaker diarization using pyannote.audio.
-    
-    Args:
-        audio_path (str): Path to audio file.
-        num_speakers (int, optional): Number of speakers if known.
-        hf_token (str): HuggingFace Auth Token.
-        
-    Returns:
-        list: List of segments with speaker labels [(start, end, speaker), ...].
     """
     try:
         from pyannote.audio import Pipeline
     except ImportError:
-        print("Error: pyannote.audio not installed. Diarization skipped.")
+        tracker.logger.error("Error: pyannote.audio not installed. Diarization skipped.")
         return None
 
     if not hf_token:
-        print("Error: HuggingFace Token (HF_TOKEN) not found. Diarization skipped.")
+        tracker.logger.error("Error: HuggingFace Token (HF_TOKEN) not found. Diarization skipped.")
         return None
 
-    print(f"Loading Diarization Pipeline (pyannote/speaker-diarization-3.1)...")
+    tracker.logger.info(f"Loading Diarization Pipeline (pyannote/speaker-diarization-3.1)...")
     try:
+        # Fix for PyTorch 2.6+ weights_only issue
+        torch.serialization.add_safe_globals([torch.torch_version.TorchVersion])
+        
         # Note: 'use_auth_token' was deprecated in favor of 'token' in recent versions
         pipeline = Pipeline.from_pretrained(
             "pyannote/speaker-diarization-3.1",
@@ -38,11 +35,10 @@ def diarize_audio(audio_path, num_speakers=None, hf_token=None):
         )
         
         # Move to GPU if available
-        import torch
         if torch.cuda.is_available():
             pipeline.to(torch.device("cuda"))
         
-        print(f"Diarizing {audio_path}...")
+        tracker.logger.info(f"Diarizing {audio_path}...")
         diarization = pipeline(audio_path, num_speakers=num_speakers)
         
         results = []
@@ -56,7 +52,7 @@ def diarize_audio(audio_path, num_speakers=None, hf_token=None):
         return results
 
     except Exception as e:
-        print(f"Error during diarization: {e}")
+        tracker.logger.error(f"Error during diarization: {e}")
         return None
 
 def merge_diarization_with_transcript(transcript_segments, diarization_segments):
diff --git a/video_transcription/ai_transcriber_v2/documentation/HOW_TO_RUN_BACKGROUND.md b/video_transcription/ai_transcriber_v2/documentation/HOW_TO_RUN_BACKGROUND.md
index 2547ba0..6bd7b3b 100644
--- a/video_transcription/ai_transcriber_v2/documentation/HOW_TO_RUN_BACKGROUND.md
+++ b/video_transcription/ai_transcriber_v2/documentation/HOW_TO_RUN_BACKGROUND.md
@@ -60,3 +60,31 @@ Here are the two best ways to do this.
     ```bash
     tail -f job_output.log
     ```
+
+---
+
+## Method 3: Rescuing a Running Job (The `disown` Method)
+**Use this if:** You already started the wizard or script in a normal terminal and now realize you need to disconnect without killing it.
+
+1.  **Pause the running job:**
+    Inside the terminal where the script is active, press **`Ctrl` + `Z`**.
+    *   The script will pause, and you'll see: `[1]+  Stopped ...`
+
+2.  **Move it to the background:**
+    Type the following command and press Enter:
+    ```bash
+    bg
+    ```
+    *   The script will resume running in the background.
+
+3.  **Disown the job:**
+    Tell the terminal to "forget" about the job so it doesn't kill it when you disconnect:
+    ```bash
+    disown -h %1
+    ```
+    *(Note: Use `%1` if your job number was `[1]`, `%2` if it was `[2]`, etc.)*
+
+4.  **Disconnect:**
+    You can now type `exit` or close your SSH window. The job will continue on the server.
+
+**Warning:** Unlike `tmux`, you cannot "reattach" to see the progress bar again. You must monitor the logs in the `logs/` folder to check its status.
diff --git a/video_transcription/ai_transcriber_v2/translator.py b/video_transcription/ai_transcriber_v2/translator.py
index 1118fb5..4ea1223 100644
--- a/video_transcription/ai_transcriber_v2/translator.py
+++ b/video_transcription/ai_transcriber_v2/translator.py
@@ -14,7 +14,7 @@ from utils import LANGUAGE_MAP
 # Define a retry decorator
 # ... (retry_policy remains)
 
-def translate_via_ollama(source_srt_content, target_language="English", model="llama3"):
+def translate_via_ollama(source_srt_content, target_language="English", model="dolphin-llama3"):
     """
     Translates SRT content using a local Ollama model (Line-by-Line for progress).
     Includes retries and debug logging.
diff --git a/video_transcription/ai_transcriber_v2/utils.py b/video_transcription/ai_transcriber_v2/utils.py
index 7750f4f..499bbb0 100644
--- a/video_transcription/ai_transcriber_v2/utils.py
+++ b/video_transcription/ai_transcriber_v2/utils.py
@@ -70,7 +70,7 @@ def verify_file_not_empty(file_path):
         return True
     return False
 
-def ensure_ollama_running(model_name="llama3"):
+def ensure_ollama_running(model_name="dolphin-llama3"):
     """
     Checks if Ollama is running. If not, attempts to start it.
     Supports Flatpak by escaping to host via flatpak-spawn.