Update app.py

app.py

@@ -10,6 +10,8 @@ import requests
 import gc
 import math
 import uuid
+import re
+from urllib.parse import quote
 from datetime import timedelta
 
 # --- CRITICAL ENVIRONMENT FIXES ---
@@ -77,6 +79,18 @@ ACTIVE_GEMINI_KEY = ENV_GEMINI_KEY if ENV_GEMINI_KEY else HARDCODED_GEMINI_KEY
 
 # --- HELPER FUNCTIONS ---
 
+def clean_json_response(text):
+    """Strips markdown code fences from Gemini output to prevent JSON errors."""
+    try:
+        # Remove ```json ... ``` or just ``` ... ```
+        pattern = r"```(?:json)?\s*(.*?)```"
+        match = re.search(pattern, text, re.DOTALL)
+        if match:
+            return match.group(1).strip()
+        return text.strip()
+    except Exception:
+        return text
+
 def format_timecode(seconds, fps=25):
     """Converts seconds to HH:MM:SS:FF (for EDL)."""
     td = timedelta(seconds=seconds)
@@ -114,10 +128,19 @@ def generate_cmx_edl(edl_title, segments, source_name, fps=25):
     return "\n".join(edl_lines)
 
 def generate_xml(sequence_name, segments, source_name, fps=25):
-    """Constructs a Hybrid XML compatible with both Premiere and Resolve."""
+    """Constructs a Robust XML for Premiere AND Resolve."""
     
+    # Static UUID to ensure all cuts are seen as the same master clip
     master_id = "masterfile-1"
     
+    # FPS Logic (Fix for 29.97 vs 30 issues)
+    timebase = int(round(fps))
+    is_ntsc = "TRUE" if fps % 1 != 0 else "FALSE"
+    
+    # URL Encode the filename to handle spaces (Critical for Premiere)
+    encoded_name = quote(source_name)
+    path_url = f"file://localhost/{encoded_name}.mov"
+
     lines = []
     lines.append('<?xml version="1.0" encoding="UTF-8"?>')
     lines.append('<!DOCTYPE xmeml>')
@@ -125,8 +148,8 @@ def generate_xml(sequence_name, segments, source_name, fps=25):
     lines.append('<sequence>')
     lines.append(f'\t<name>{sequence_name}</name>')
     lines.append('\t<rate>')
-    lines.append(f'\t\t<timebase>{int(fps)}</timebase>')
-    lines.append(f'\t\t<ntsc>{str(fps % 1 != 0).upper()}</ntsc>') 
+    lines.append(f'\t\t<timebase>{timebase}</timebase>')
+    lines.append(f'\t\t<ntsc>{is_ntsc}</ntsc>') 
     lines.append('\t</rate>')
     lines.append('\t<media>')
     
@@ -134,7 +157,7 @@ def generate_xml(sequence_name, segments, source_name, fps=25):
     lines.append('\t\t<video>')
     lines.append('\t\t\t<format>')
     lines.append('\t\t\t\t<samplecharacteristics>')
-    lines.append(f'\t\t\t\t\t<rate><timebase>{int(fps)}</timebase></rate>')
+    lines.append(f'\t\t\t\t\t<rate><timebase>{timebase}</timebase></rate>')
     lines.append('\t\t\t\t\t<width>1920</width>')
     lines.append('\t\t\t\t\t<height>1080</height>')
     lines.append('\t\t\t\t\t<anamorphic>FALSE</anamorphic>')
@@ -152,35 +175,34 @@ def generate_xml(sequence_name, segments, source_name, fps=25):
         tl_start = timeline_head_frames
         tl_end = tl_start + duration_frames
         
+        # Truncate clip names for Premiere safety (max 80 chars)
         raw_note = seg.get('note', 'Junior Editor Selection')
         clip_note = (raw_note[:75] + '..') if len(raw_note) > 75 else raw_note
 
         lines.append(f'\t\t\t\t<clipitem id="clipitem-v-{i}">')
         lines.append(f'\t\t\t\t\t<name>{clip_note}</name>')
         lines.append(f'\t\t\t\t\t<duration>{duration_frames}</duration>')
-        lines.append(f'\t\t\t\t\t<rate><timebase>{int(fps)}</timebase></rate>')
+        lines.append(f'\t\t\t\t\t<rate><timebase>{timebase}</timebase></rate>')
         lines.append(f'\t\t\t\t\t<start>{tl_start}</start>')
         lines.append(f'\t\t\t\t\t<end>{tl_end}</end>')
         lines.append(f'\t\t\t\t\t<in>{src_in_frames}</in>')
         lines.append(f'\t\t\t\t\t<out>{src_out_frames}</out>')
         lines.append(f'\t\t\t\t\t<masterclipid>{master_id}</masterclipid>')
         
-        # --- UNIVERSAL FILE DEFINITION ---
-        # 1. <pathurl>: Used by Resolve to find the file.
-        # 2. <media>: Used by Premiere to understand the file structure.
+        # --- FULL ROBUST FILE DEFINITION ---
+        # We repeat the full definition every time. It's verbose, but it's the only way
+        # to guarantee Premiere sees the metadata AND Resolve sees the path.
         lines.append(f'\t\t\t\t\t<file id="{master_id}">')
         lines.append(f'\t\t\t\t\t\t<name>{source_name}</name>')
-        lines.append(f'\t\t\t\t\t\t<pathurl>file://localhost/{source_name}.mov</pathurl>') 
-        lines.append(f'\t\t\t\t\t\t<rate><timebase>{int(fps)}</timebase></rate>')
+        lines.append(f'\t\t\t\t\t\t<pathurl>{path_url}</pathurl>') 
+        lines.append(f'\t\t\t\t\t\t<rate><timebase>{timebase}</timebase></rate>')
         lines.append(f'\t\t\t\t\t\t<duration>8640000</duration>') 
         lines.append(f'\t\t\t\t\t\t<timecode>')
-        lines.append(f'\t\t\t\t\t\t\t<rate><timebase>{int(fps)}</timebase></rate>')
+        lines.append(f'\t\t\t\t\t\t\t<rate><timebase>{timebase}</timebase></rate>')
         lines.append(f'\t\t\t\t\t\t\t<string>00:00:00:00</string>') 
         lines.append(f'\t\t\t\t\t\t\t<frame>0</frame>')
         lines.append(f'\t\t\t\t\t\t\t<displayformat>NDF</displayformat>')
         lines.append(f'\t\t\t\t\t\t</timecode>')
-        
-        # --- MEDIA BLOCK (Required by Premiere) ---
         lines.append(f'\t\t\t\t\t\t<media>')
         lines.append(f'\t\t\t\t\t\t\t<video>')
         lines.append(f'\t\t\t\t\t\t\t\t<samplecharacteristics><width>1920</width><height>1080</height></samplecharacteristics>')
@@ -190,9 +212,9 @@ def generate_xml(sequence_name, segments, source_name, fps=25):
         lines.append(f'\t\t\t\t\t\t\t\t<channelcount>2</channelcount>')
         lines.append(f'\t\t\t\t\t\t\t</audio>')
         lines.append(f'\t\t\t\t\t\t</media>')
-        # ------------------------------------------
-        
         lines.append(f'\t\t\t\t\t</file>')
+        # -----------------------------------
+        
         lines.append(f'\t\t\t\t</clipitem>')
         
         gap_seconds = seg.get('gap', 0.0)
@@ -221,15 +243,35 @@ def generate_xml(sequence_name, segments, source_name, fps=25):
         lines.append(f'\t\t\t\t<clipitem id="clipitem-a-{i}">')
         lines.append(f'\t\t\t\t\t<name>{clip_note}</name>')
         lines.append(f'\t\t\t\t\t<duration>{duration_frames}</duration>')
-        lines.append(f'\t\t\t\t\t<rate><timebase>{int(fps)}</timebase></rate>')
+        lines.append(f'\t\t\t\t\t<rate><timebase>{timebase}</timebase></rate>')
         lines.append(f'\t\t\t\t\t<start>{tl_start}</start>')
         lines.append(f'\t\t\t\t\t<end>{tl_end}</end>')
         lines.append(f'\t\t\t\t\t<in>{src_in_frames}</in>')
         lines.append(f'\t\t\t\t\t<out>{src_out_frames}</out>')
         lines.append(f'\t\t\t\t\t<masterclipid>{master_id}</masterclipid>')
         
-        # Reuse same file definition
-        lines.append(f'\t\t\t\t\t<file id="{master_id}"/>')
+        # Reuse same file definition (Verbose)
+        lines.append(f'\t\t\t\t\t<file id="{master_id}">')
+        lines.append(f'\t\t\t\t\t\t<name>{source_name}</name>')
+        lines.append(f'\t\t\t\t\t\t<pathurl>{path_url}</pathurl>') 
+        lines.append(f'\t\t\t\t\t\t<rate><timebase>{timebase}</timebase></rate>')
+        lines.append(f'\t\t\t\t\t\t<duration>8640000</duration>') 
+        lines.append(f'\t\t\t\t\t\t<timecode>')
+        lines.append(f'\t\t\t\t\t\t\t<rate><timebase>{timebase}</timebase></rate>')
+        lines.append(f'\t\t\t\t\t\t\t<string>00:00:00:00</string>') 
+        lines.append(f'\t\t\t\t\t\t\t<frame>0</frame>')
+        lines.append(f'\t\t\t\t\t\t\t<displayformat>NDF</displayformat>')
+        lines.append(f'\t\t\t\t\t\t</timecode>')
+        lines.append(f'\t\t\t\t\t\t<media>')
+        lines.append(f'\t\t\t\t\t\t\t<video>')
+        lines.append(f'\t\t\t\t\t\t\t\t<samplecharacteristics><width>1920</width><height>1080</height></samplecharacteristics>')
+        lines.append(f'\t\t\t\t\t\t\t</video>')
+        lines.append(f'\t\t\t\t\t\t\t<audio>')
+        lines.append(f'\t\t\t\t\t\t\t\t<samplecharacteristics><depth>16</depth><samplerate>48000</samplerate></samplecharacteristics>')
+        lines.append(f'\t\t\t\t\t\t\t\t<channelcount>2</channelcount>')
+        lines.append(f'\t\t\t\t\t\t\t</audio>')
+        lines.append(f'\t\t\t\t\t\t</media>')
+        lines.append(f'\t\t\t\t\t</file>')
         
         # SOURCE TRACK MAPPING (Simple Mono)
         lines.append('\t\t\t\t\t<sourcetrack>')
@@ -283,7 +325,9 @@ def call_gemini_for_edl(transcript_data, story_prompt, api_key):
         res = requests.post(url, json=payload)
         res.raise_for_status()
         result_json = res.json()
-        return json.loads(result_json['candidates'][0]['content']['parts'][0]['text'])
+        raw_text = result_json['candidates'][0]['content']['parts'][0]['text']
+        cleaned_text = clean_json_response(raw_text)
+        return json.loads(cleaned_text)
     except Exception as e:
         st.error(f"Senior Editor AI Error: {e}")
         return None
@@ -305,7 +349,7 @@ st.divider()
 
 with st.sidebar:
     st.header("Project Settings")
-    fps = st.number_input("Timeline FPS", value=25)
+    fps = st.number_input("Timeline FPS", value=25.0, format="%.3f")
     
     st.header("Export Settings")
     export_format = st.radio("Output Format", ["EDL", "XML (Multicam)"], index=0)