Improve segmented vision matching quality

src/llm/vision_cache.py

@@ -13,6 +13,7 @@ import base64
 import json
 import logging
 import re
+import time
 import urllib.error
 import urllib.request
 from dataclasses import asdict
@@ -38,6 +39,17 @@ Return only compact JSON with these keys:
 subject, setting, composition, action_phase, distinctive_objects, lighting_color, negatives.
 Focus on stable visual facts and spatial layout. Ignore timecode overlays, subtitles, logos, compression, aspect ratio, and color grading differences."""
 
+_RETRYABLE_HTTP_CODES = {408, 409, 425, 429, 500, 502, 503, 504}
+_CREDIT_ERROR_PATTERNS = (
+    "insufficient credit",
+    "insufficient credits",
+    "no credits",
+    "out of credits",
+    "billing",
+    "quota exceeded",
+    "payment required",
+)
+
 
 def _cache_path(cfg: AppConfig) -> Path:
     return cfg.paths.cache_dir / "vision_descriptions.json"
@@ -133,13 +145,44 @@ def _call_vision_model(label: str, image_urls: list[str], cfg: AppConfig) -> str
 
     url = f"{vision.base_url.rstrip('/')}/chat/completions"
     req = urllib.request.Request(url, data=body, headers=headers, method="POST")
-    try:
-        with urllib.request.urlopen(req, timeout=vision.timeout_seconds) as resp:
-            data = json.loads(resp.read().decode("utf-8"))
-            return str(data["choices"][0]["message"]["content"]).strip()
-    except urllib.error.HTTPError as exc:
-        body_text = exc.read().decode(errors="replace")
-        raise RuntimeError(f"Vision HTTP {exc.code} from {url}:\n{body_text}") from exc
+    delays_s = (8.0, 20.0, 45.0, 90.0)
+    for attempt in range(len(delays_s) + 1):
+        try:
+            with urllib.request.urlopen(req, timeout=vision.timeout_seconds) as resp:
+                data = json.loads(resp.read().decode("utf-8"))
+                return str(data["choices"][0]["message"]["content"]).strip()
+        except urllib.error.HTTPError as exc:
+            body_text = exc.read().decode(errors="replace")
+            lowered = body_text.lower()
+            if exc.code == 402 or any(pattern in lowered for pattern in _CREDIT_ERROR_PATTERNS):
+                raise RuntimeError(f"Vision HTTP {exc.code} from {url}:\n{body_text}") from exc
+            if exc.code not in _RETRYABLE_HTTP_CODES or attempt >= len(delays_s):
+                raise RuntimeError(f"Vision HTTP {exc.code} from {url}:\n{body_text}") from exc
+            delay_s = delays_s[attempt]
+            logger.warning(
+                "Vision HTTP %d for %s; waiting %.0fs before retry %d/%d.",
+                exc.code,
+                label,
+                delay_s,
+                attempt + 1,
+                len(delays_s),
+            )
+            time.sleep(delay_s)
+        except urllib.error.URLError as exc:
+            if attempt >= len(delays_s):
+                raise RuntimeError(f"Vision request failed for {url}: {exc}") from exc
+            delay_s = delays_s[attempt]
+            logger.warning(
+                "Vision request failed for %s (%s); waiting %.0fs before retry %d/%d.",
+                label,
+                exc.reason,
+                delay_s,
+                attempt + 1,
+                len(delays_s),
+            )
+            time.sleep(delay_s)
+
+    raise RuntimeError(f"Vision request failed unexpectedly for {url}")
 
 
 def _description_key(kind: str, item_id: int, start_s: float, end_s: float, cfg: AppConfig) -> str: