Initial commit: 2125_GCE project

gx_lain/runtime.py

@@ -50,25 +50,33 @@ def normalize_segments(
     Converts raw segment metadata from codex_lineage.segments into normalized form
     with required keys: id, start_line, end_line, text, symbolic_lane, semantic_role.
 
-    Assumptions:
-    - Raw segments have: id, start, end (0-based line indices)
-    - Optional: lane, role metadata
-    - Text extraction is stubbed (reserved for future payload analysis)
+    Text extraction: Decompresses payload and extracts text for each segment using
+    byte ranges from segment metadata (start_byte, end_byte).
 
     Args:
         manifest: GX manifest dict
         raw_segments: List of raw segment dicts from codex_lineage
-        payload: Compressed payload bytes (reserved for text extraction)
+        payload: Compressed payload bytes
 
     Returns:
         List of normalized segment dicts conforming to Segment schema
     """
+    from xic_extensions.gsz3_decompressor import GSZ3Decompressor, GSZ3DecompressionError
+    
     normalized = []
+    
+    # Decompress payload to get full text
+    try:
+        decompressed = GSZ3Decompressor.decompress(payload)
+    except GSZ3DecompressionError:
+        decompressed = ""
 
     for raw_seg in raw_segments:
         seg_id = str(raw_seg.get("id", "unknown"))
         start_line = int(raw_seg.get("start", 0))
         end_line = int(raw_seg.get("end", 0))
+        start_byte = int(raw_seg.get("start_byte", 0))
+        end_byte = int(raw_seg.get("end_byte", len(decompressed)))
 
         # Explicit lane from metadata, or infer
         explicit_lane = raw_seg.get("lane")
@@ -83,8 +91,15 @@ def normalize_segments(
         # Semantic role: infer from segment metadata
         semantic_role = _infer_semantic_role(raw_seg)
 
-        # Text: stub for now; in production extract from payload via byte ranges
-        text = f"[segment {seg_id}: lines {start_line}–{end_line}]"
+        # Extract text from decompressed payload using byte ranges
+        text = ""
+        if decompressed and start_byte < len(decompressed):
+            end = min(end_byte, len(decompressed))
+            text = decompressed[start_byte:end]
+        
+        # Fallback if no text extracted
+        if not text.strip():
+            text = f"[segment {seg_id}: lines {start_line}–{end_line}]"
 
         normalized_seg = {
             "id": seg_id,
@@ -93,7 +108,7 @@ def normalize_segments(
             "text": text,
             "symbolic_lane": symbolic_lane,
             "semantic_role": semantic_role,
-            "_raw": raw_seg,  # preserve original for debugging
+            "_raw": raw_seg,
         }
 
         normalized.append(normalized_seg)