sync: PR #28 - tkinter GUI restore, proxy feature toggles, CROF guard fixes, gRPC test skip

translate-proxy.py

@@ -311,6 +311,9 @@ FORCE_MODEL = ""
 BGP_ROUTES = []
 CAVEMAN_MODE = False
 RTK_COMPRESSION = False
+AUTO_COMPACT = False
+ADAPTIVE_COMPACT = False
+TOOL_OUTPUT_TRUNCATION = True
 PROMPT_ENHANCER = False
 PROMPT_ENHANCER_MODE = "offline"
 PROMPT_ENHANCER_MODEL = ""
@@ -1179,7 +1182,7 @@ def _init_runtime():
     global MODELS, CC_VERSION, REASONING_ENABLED, REASONING_EFFORT, BGP_ROUTES
     global _api_key_pool, PROMPT_ENHANCER
     global VISION_FALLBACK_URL, VISION_FALLBACK_MODEL, VISION_FALLBACK_KEY
-    global CAVEMAN_MODE, RTK_COMPRESSION
+    global CAVEMAN_MODE, RTK_COMPRESSION, AUTO_COMPACT, ADAPTIVE_COMPACT, TOOL_OUTPUT_TRUNCATION
 
     CONFIG = load_config()
     PORT = CONFIG["port"]
@@ -1215,6 +1218,11 @@ def _init_runtime():
     BGP_ROUTES = CONFIG.get("bgp_routes", [])
     CAVEMAN_MODE = CONFIG.get("caveman_mode", False) or os.environ.get("CAVEMAN_MODE") == "1"
     RTK_COMPRESSION = CONFIG.get("rtk_compression", False) or os.environ.get("RTK_COMPRESSION") == "1"
+    AUTO_COMPACT = CONFIG.get("auto_compact", False) or os.environ.get("AUTO_COMPACT") == "1"
+    ADAPTIVE_COMPACT = CONFIG.get("adaptive_compact", False) or os.environ.get("ADAPTIVE_COMPACT") == "1"
+    TOOL_OUTPUT_TRUNCATION = CONFIG.get("tool_output_truncation", True)
+    if os.environ.get("TOOL_OUTPUT_TRUNCATION") == "0":
+        TOOL_OUTPUT_TRUNCATION = False
     _api_key_pool = None
     if API_KEY and "," in API_KEY and not OAUTH_PROVIDER.startswith("google") and BACKEND not in ("codebuff", "freebuff"):
         _api_key_pool = APIKeyPool(BACKEND, API_KEY)
@@ -1962,6 +1970,8 @@ def _crof_item_limit(model):
     return min(per_model, _CROF_ADAPTIVE["global_item_limit"])
 
 def _crof_compact_for_retry(input_data, model, aggression=0):
+    if "crof.ai" not in TARGET_URL:
+        return input_data
     limit = _crof_item_limit(model)
     if not isinstance(input_data, list) or len(input_data) < 2:
         return input_data
@@ -2149,10 +2159,11 @@ def _compact_input(input_data):
             compressed_data.append(item)
         input_data = compressed_data
 
-    if not isinstance(input_data, list) or len(input_data) <= _MAX_INPUT_ITEMS:
+    # Skip compaction entirely — just do optional tool output truncation
+    if not AUTO_COMPACT or not isinstance(input_data, list) or len(input_data) <= _MAX_INPUT_ITEMS:
         out = []
         for item in input_data:
-            if isinstance(item, dict) and item.get("type") == "function_call_output":
+            if TOOL_OUTPUT_TRUNCATION and isinstance(item, dict) and item.get("type") == "function_call_output":
                 o = item.get("output", "")
                 if len(o) > _MAX_TOOL_OUTPUT_CHARS:
                     item = dict(item)
@@ -2191,7 +2202,7 @@ def _compact_input(input_data):
         return head + tail
 
     for item in tail:
-        if isinstance(item, dict) and item.get("type") == "function_call_output":
+        if TOOL_OUTPUT_TRUNCATION and isinstance(item, dict) and item.get("type") == "function_call_output":
             o = item.get("output", "")
             if len(o) > _MAX_TOOL_OUTPUT_CHARS:
                 item["output"] = o[:_MAX_TOOL_OUTPUT_CHARS] + f"\n... [truncated {len(o) - _MAX_TOOL_OUTPUT_CHARS} chars]"
@@ -5703,7 +5714,6 @@ class Handler(http.server.BaseHTTPRequestHandler):
         raw_types = [i.get("type") for i in raw_input] if isinstance(raw_input, list) else "str"
         resolved_types = [i.get("type") for i in input_data] if isinstance(input_data, list) else "str"
 
-        print(f"[{_sid}] prev_id={prev_id} raw={raw_types} resolved={resolved_types}", file=sys.stderr)
         with open(_log_path, "a", encoding="utf-8") as _lf:
             _lf.write(f"\n{'='*60}\n{_ts} [session={_sid}] REQUEST {self.path}\n")
             _lf.write(f"  prev_id={prev_id}\n")
@@ -5807,7 +5817,7 @@ class Handler(http.server.BaseHTTPRequestHandler):
                 body["input"] = input_data
 
         compacted = False
-        if policy.get("compaction") and isinstance(input_data, list) and "claude" not in model.lower():
+        if ADAPTIVE_COMPACT and policy.get("compaction") and isinstance(input_data, list) and "claude" not in model.lower():
             input_data, compacted = _adaptive_compact(input_data, model, policy)
             if compacted:
                 body = dict(body)
@@ -5819,7 +5829,7 @@ class Handler(http.server.BaseHTTPRequestHandler):
             body["input"] = input_data
 
         crof_limit = _crof_item_limit(model)
-        _crof_eligible = True
+        _crof_eligible = "crof.ai" in TARGET_URL
         if _crof_eligible and not compacted and isinstance(input_data, list):
             _needs_compact = len(input_data) > crof_limit
             max_tok = _get_model_max_tokens(model)
@@ -6771,7 +6781,7 @@ class Handler(http.server.BaseHTTPRequestHandler):
             body["input"] = input_data
 
         compacted = False
-        if policy.get("compaction") and isinstance(input_data, list) and "claude" not in model.lower():
+        if ADAPTIVE_COMPACT and policy.get("compaction") and isinstance(input_data, list) and "claude" not in model.lower():
             input_data, compacted = _adaptive_compact(input_data, model, policy)
             if compacted:
                 body = dict(body)
@@ -7662,9 +7672,10 @@ class Handler(http.server.BaseHTTPRequestHandler):
             has_content = False
             has_message = False
             has_tool_call = False
+            _last_stream_usage = {}
 
             def _observe_event(event):
-                nonlocal last_resp_id, last_output, last_status, finish_reason, has_content, has_message, has_tool_call
+                nonlocal last_resp_id, last_output, last_status, finish_reason, has_content, has_message, has_tool_call, _last_stream_usage
                 for line in event.strip().split("\n"):
                     if line.startswith("data: "):
                         try:
@@ -7677,6 +7688,9 @@ class Handler(http.server.BaseHTTPRequestHandler):
                                 has_tool_call = any(o.get("type") == "function_call" for o in (last_output or []))
                                 has_message = any(o.get("type") == "message" for o in (last_output or []))
                                 has_content = has_message or has_tool_call
+                                resp_usage = d.get("response", {}).get("usage")
+                                if resp_usage:
+                                    _last_stream_usage = resp_usage
                         except Exception:
                             pass
 
@@ -9044,7 +9058,7 @@ def main():
         allow_reuse_address = True
         daemon_threads = True
         request_queue_size = 64
-    BIND_HOST = getattr(_args, "host", None) or os.environ.get("CODEX_HOST", "127.0.0.1")
+    BIND_HOST = os.environ.get("CODEX_HOST", "127.0.0.1")
     SERVER = ReusableHTTPServer((BIND_HOST, PORT), Handler)
     print(f"translate-proxy ({BACKEND}) listening on http://{BIND_HOST}:{PORT}", flush=True)
     print(f"Target: {TARGET_URL}", flush=True)