argo-proxy 3.2.0a0__tar.gz → 3.2.1__tar.gz

{argo_proxy-3.2.0a0/src/argo_proxy.egg-info → argo_proxy-3.2.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: argo-proxy
-Version: 3.2.0a0
+Version: 3.2.1
 Summary: Proxy server to Argo API, OpenAI format compatible
 Author-email: Peng Ding <oaklight@gmx.com>
 License-Expression: MIT
@@ -15,7 +15,7 @@ Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: aiohttp>=3.12.2
-Requires-Dist: llm-rosetta>=0.7.0a0
+Requires-Dist: llm-rosetta>=0.7.0
 Requires-Dist: pydantic>=2.11.7
 Requires-Dist: tiktoken>=0.9.0
 Requires-Dist: tqdm>=4.67.1

{argo_proxy-3.2.0a0 → argo_proxy-3.2.1}/pyproject.toml

@@ -20,7 +20,7 @@ classifiers = [
 
 dependencies = [
     "aiohttp>=3.12.2",
-    "llm-rosetta>=0.7.0a0",
+    "llm-rosetta>=0.7.0",
     "pydantic>=2.11.7",
     "tiktoken>=0.9.0",
     "tqdm>=4.67.1",

{argo_proxy-3.2.0a0 → argo_proxy-3.2.1/src/argo_proxy.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: argo-proxy
-Version: 3.2.0a0
+Version: 3.2.1
 Summary: Proxy server to Argo API, OpenAI format compatible
 Author-email: Peng Ding <oaklight@gmx.com>
 License-Expression: MIT
@@ -15,7 +15,7 @@ Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: aiohttp>=3.12.2
-Requires-Dist: llm-rosetta>=0.7.0a0
+Requires-Dist: llm-rosetta>=0.7.0
 Requires-Dist: pydantic>=2.11.7
 Requires-Dist: tiktoken>=0.9.0
 Requires-Dist: tqdm>=4.67.1

{argo_proxy-3.2.0a0 → argo_proxy-3.2.1}/src/argo_proxy.egg-info/requires.txt

@@ -1,5 +1,5 @@
 aiohttp>=3.12.2
-llm-rosetta>=0.7.0a0
+llm-rosetta>=0.7.0
 pydantic>=2.11.7
 tiktoken>=0.9.0
 tqdm>=4.67.1

argo_proxy-3.2.1/src/argoproxy/__init__.py

@@ -0,0 +1 @@
+__version__ = "3.2.1"

{argo_proxy-3.2.0a0 → argo_proxy-3.2.1}/src/argoproxy/app.py

@@ -92,7 +92,6 @@ async def prepare_app(app):
     # Create optimized HTTP session
     resolve_overrides = getattr(app["config"], "resolve_overrides", None)
     http_session_manager = OptimizedHTTPSession(
-        user_agent=f"argo-proxy/{__version__}",
         resolve_overrides=resolve_overrides if resolve_overrides else None,
         **perf_config,
     )

{argo_proxy-3.2.0a0 → argo_proxy-3.2.1}/src/argoproxy/config/model.py

@@ -76,6 +76,10 @@ class ArgoConfig:
     # Model list auto-refresh
     model_refresh_interval_hours: float = 24  # 0 to disable
 
+    # Fallback models — used when an unrecognized model name is requested
+    default_chat_model: str = "argo:gpt-5.4-nano"
+    default_embed_model: str = "argo:text-embedding-3-small"
+
     # Image processing settings
     enable_payload_control: bool = False  # Enable automatic payload size control
     max_payload_size: int = 20  # MB default (total for all images)

{argo_proxy-3.2.0a0 → argo_proxy-3.2.1}/src/argoproxy/endpoints/dispatch.py

@@ -288,6 +288,11 @@ def _build_upstream_headers(
     """Build HTTP headers for the upstream API request."""
     headers: dict[str, str] = {"Content-Type": "application/json"}
 
+    # Pass through client's User-Agent to upstream (transparent proxy)
+    client_ua = request.headers.get("User-Agent")
+    if client_ua:
+        headers["User-Agent"] = client_ua
+
     if should_use_username_passthrough():
         api_key = _extract_client_credential(request, target_provider) or fallback_user
     else:

{argo_proxy-3.2.0a0 → argo_proxy-3.2.1}/src/argoproxy/models.py

@@ -51,6 +51,7 @@ _DEFAULT_CHAT_MODELS = flatten_mapping(
         "gpt51": "argo:gpt-5.1",
         "gpt52": "argo:gpt-5.2",
         "gpt54": "argo:gpt-5.4",
+        "gpt54nano": "argo:gpt-5.4-nano",
         "gpt55": "argo:gpt-5.5",
         # gemini
         "gemini25pro": "argo:gemini-2.5-pro",
@@ -308,7 +309,6 @@ async def get_upstream_model_list_async(
         async with aiohttp.ClientSession(
             connector=connector,
             timeout=client_timeout,
-            headers={"User-Agent": "argo-proxy/1.0"},
         ) as session:
             log_debug(f"Sending request to: {url}", context="models")
 
@@ -480,7 +480,7 @@ def _categorize_results(
             unavailable.clear()
         else:
             log_error(
-                "Proceeding without unavailable models. Subsequent calls to these models will be replaced with argo:gpt-5-nano",
+                "Proceeding without unavailable models. Subsequent calls to these models will use the configured default fallback",
                 context="models",
             )
 
@@ -781,14 +781,27 @@ class ModelRegistry:
             return model_name
 
         if model_type == "chat":
-            default_model = "argo:gpt-5-nano"
+            default_model = self._config.default_chat_model
         elif model_type == "embed":
-            default_model = "argo:text-embedding-3-small"
+            default_model = self._config.default_embed_model
+        else:
+            default_model = self._config.default_chat_model
+            log_warning(
+                f"Unknown model_type '{model_type}', using chat fallback",
+                context="ModelRegistry",
+            )
         log_warning(
             f"Model '{model_name}' not found in registry, falling back to {default_model}",
             context="ModelRegistry",
         )
-        return self.available_models[default_model]
+        if default_model in self.available_models:
+            return self.available_models[default_model]
+        # Default model itself not in registry — return its name as-is
+        log_warning(
+            f"Default fallback model '{default_model}' not in registry either",
+            context="ModelRegistry",
+        )
+        return default_model
 
     def as_openai_list(self) -> dict[str, Any]:
         # Mock data for available models

{argo_proxy-3.2.0a0 → argo_proxy-3.2.1}/src/argoproxy/performance.py

@@ -93,7 +93,6 @@ class OptimizedHTTPSession:
         read_timeout: Socket read timeout in seconds.
         total_timeout: Total request timeout in seconds.
         dns_cache_ttl: DNS cache TTL in seconds.
-        user_agent: User agent string.
         resolve_overrides: Optional dict mapping "host:port" to IP address
             for custom DNS resolution (similar to curl --resolve).
     """
@@ -108,7 +107,6 @@ class OptimizedHTTPSession:
         read_timeout: int = 30,
         total_timeout: int = 60,
         dns_cache_ttl: int = 300,
-        user_agent: str = "argo-proxy",
         resolve_overrides: dict[str, str] | None = None,
     ):
         connector_kwargs: dict = {
@@ -137,7 +135,6 @@ class OptimizedHTTPSession:
         )
 
         self.session: aiohttp.ClientSession | None = None
-        self.user_agent = user_agent
 
     async def create_session(self) -> aiohttp.ClientSession:
         """Create and return the HTTP session."""
@@ -145,7 +142,6 @@ class OptimizedHTTPSession:
             self.session = aiohttp.ClientSession(
                 connector=self.connector,
                 timeout=self.timeout,
-                headers={"User-Agent": self.user_agent},
             )
             log_debug(
                 f"HTTP session created: {self.connector.limit} total, "

argo_proxy-3.2.0a0/src/argoproxy/__init__.py

@@ -1 +0,0 @@
-__version__ = "3.2.0a0"