ccproxy-api 0.2.4__tar.gz → 0.2.6__tar.gz

{ccproxy_api-0.2.4 → ccproxy_api-0.2.6}/PKG-INFO

@@ -1,42 +1,42 @@
 Metadata-Version: 2.4
 Name: ccproxy-api
-Version: 0.2.4
+Version: 0.2.6
 Summary: API server that provides an Anthropic and OpenAI compatible interface over Claude Code, allowing to use your Claude OAuth account or over the API.
 License-File: LICENSE
 Requires-Python: >=3.11
-Requires-Dist: aiofiles>=24.1.0
-Requires-Dist: fastapi[standard]>=0.115.14
+Requires-Dist: aiofiles>=25.1.0
+Requires-Dist: fastapi[standard]>=0.119.0
 Requires-Dist: httpx[http2]>=0.28.1
 Requires-Dist: packaging>=25.0
-Requires-Dist: pydantic-settings>=2.4.0
-Requires-Dist: pydantic>=2.8.0
+Requires-Dist: pydantic-settings>=2.11.0
+Requires-Dist: pydantic>=2.12.2
 Requires-Dist: pyjwt>=2.10.1
-Requires-Dist: rich-toolkit>=0.14.8
-Requires-Dist: rich>=13.0.0
+Requires-Dist: rich-toolkit>=0.15.1
+Requires-Dist: rich>=14.2.0
 Requires-Dist: sortedcontainers>=2.4.0
 Requires-Dist: structlog>=25.4.0
-Requires-Dist: typer>=0.16.0
-Requires-Dist: typing-extensions>=4.0.0
-Requires-Dist: uvicorn>=0.34.0
+Requires-Dist: typer>=0.19.2
+Requires-Dist: typing-extensions>=4.15.0
+Requires-Dist: uvicorn>=0.37.0
 Provides-Extra: plugins-claude
-Requires-Dist: claude-agent-sdk>=0.1.0; extra == 'plugins-claude'
-Requires-Dist: keyring>=25.0.0; extra == 'plugins-claude'
+Requires-Dist: claude-agent-sdk>=0.1.4; extra == 'plugins-claude'
+Requires-Dist: keyring>=25.7.0; extra == 'plugins-claude'
 Requires-Dist: qrcode>=8.2; extra == 'plugins-claude'
 Provides-Extra: plugins-codex
 Requires-Dist: pyjwt>=2.10.1; extra == 'plugins-codex'
 Requires-Dist: qrcode>=8.2; extra == 'plugins-codex'
 Provides-Extra: plugins-mcp
-Requires-Dist: fastapi-mcp>=0.3.7; extra == 'plugins-mcp'
+Requires-Dist: fastapi-mcp>=0.4.0; extra == 'plugins-mcp'
 Provides-Extra: plugins-metrics
-Requires-Dist: prometheus-client>=0.22.1; extra == 'plugins-metrics'
+Requires-Dist: prometheus-client>=0.23.1; extra == 'plugins-metrics'
 Provides-Extra: plugins-storage
 Requires-Dist: duckdb-engine>=0.17.0; extra == 'plugins-storage'
-Requires-Dist: duckdb<1.4.0,>=1.1.0; extra == 'plugins-storage'
-Requires-Dist: sqlalchemy>=2.0.0; extra == 'plugins-storage'
-Requires-Dist: sqlmodel>=0.0.24; extra == 'plugins-storage'
+Requires-Dist: duckdb>=1.3.2; extra == 'plugins-storage'
+Requires-Dist: sqlalchemy>=2.0.44; extra == 'plugins-storage'
+Requires-Dist: sqlmodel>=0.0.27; extra == 'plugins-storage'
 Provides-Extra: plugins-tui
-Requires-Dist: aioconsole>=0.8.1; extra == 'plugins-tui'
-Requires-Dist: textual>=3.7.1; extra == 'plugins-tui'
+Requires-Dist: aioconsole>=0.8.2; extra == 'plugins-tui'
+Requires-Dist: textual>=6.3.0; extra == 'plugins-tui'
 Description-Content-Type: text/markdown
 
 # CCProxy API Server

{ccproxy_api-0.2.4 → ccproxy_api-0.2.6}/ccproxy/core/_version.py

@@ -28,7 +28,7 @@ version_tuple: VERSION_TUPLE
 commit_id: COMMIT_ID
 __commit_id__: COMMIT_ID
 
-__version__ = version = '0.2.4'
-__version_tuple__ = version_tuple = (0, 2, 4)
+__version__ = version = '0.2.6'
+__version_tuple__ = version_tuple = (0, 2, 6)
 
 __commit_id__ = commit_id = None

{ccproxy_api-0.2.4 → ccproxy_api-0.2.6}/ccproxy/core/plugins/factories.py

@@ -14,6 +14,7 @@ from fastapi import APIRouter
 from ccproxy.models.provider import ProviderConfig
 from ccproxy.services.adapters.base import BaseAdapter
 from ccproxy.services.adapters.http_adapter import BaseHTTPAdapter
+from ccproxy.services.adapters.mock_adapter import MockAdapter
 from ccproxy.services.interfaces import (
     IMetricsCollector,
     IRequestTracer,
@@ -104,6 +105,7 @@ class BaseProviderPluginFactory(ProviderPluginFactory):
     cli_commands: list[CliCommandSpec] = []
     cli_arguments: list[CliArgumentSpec] = []
     tool_accumulator_class: type | None = None
+    use_mock_adapter_in_bypass_mode: bool = True
 
     def __init__(self) -> None:
         """Initialize factory with manifest built from class attributes."""
@@ -215,6 +217,24 @@ class BaseProviderPluginFactory(ProviderPluginFactory):
         Returns:
             Adapter instance
         """
+        settings = context.get("settings")
+        service_container = context.get("service_container")
+        if settings and getattr(settings.server, "bypass_mode", False):
+            if not service_container:
+                raise RuntimeError(
+                    f"Cannot initialize plugin '{self.plugin_name}' in bypass mode: "
+                    "service container is required to create mock adapter. "
+                    "This is likely a configuration issue."
+                )
+            logger.warning(
+                "plugin_bypass_mode_enabled",
+                plugin=self.plugin_name,
+                adapter=self.adapter_class.__name__,
+                category="lifecycle",
+            )
+            if self.use_mock_adapter_in_bypass_mode:
+                return MockAdapter(service_container.get_mock_handler())
+
         # Extract services from context (one-time extraction)
         http_pool_manager: HTTPPoolManager | None = cast(
             "HTTPPoolManager | None", context.get("http_pool_manager")
@@ -232,7 +252,6 @@ class BaseProviderPluginFactory(ProviderPluginFactory):
         config = context.get("config")
 
         # Get all adapter dependencies from service container
-        service_container = context.get("service_container")
         if not service_container:
             raise RuntimeError("Service container is required for adapter services")
 
@@ -268,6 +287,8 @@ class BaseProviderPluginFactory(ProviderPluginFactory):
                 if hasattr(context, "get")
                 else None,
             }
+            if settings and getattr(settings.server, "bypass_mode", False):
+                adapter_kwargs["mock_handler"] = service_container.get_mock_handler()
             if self.tool_accumulator_class:
                 adapter_kwargs["tool_accumulator_class"] = self.tool_accumulator_class
 
@@ -303,6 +324,9 @@ class BaseProviderPluginFactory(ProviderPluginFactory):
                 "model_mapper": context.get("model_mapper")
                 if hasattr(context, "get")
                 else None,
+                "mock_handler": service_container.get_mock_handler()
+                if settings and getattr(settings.server, "bypass_mode", False)
+                else None,
             }
             if self.tool_accumulator_class:
                 non_http_adapter_kwargs["tool_accumulator_class"] = (

{ccproxy_api-0.2.4 → ccproxy_api-0.2.6}/ccproxy/llms/formatters/anthropic_to_openai/_helpers.py

@@ -42,3 +42,33 @@ def build_openai_tool_call(
             arguments=str(args_str),
         ),
     )
+
+
+def build_openai_tool_call_chunk(
+    *,
+    index: int,
+    tool_id: str | None,
+    tool_name: str | None,
+    tool_input: Any,
+    arguments: Any = None,
+    fallback_index: int = 0,
+) -> openai_models.ToolCallChunk:
+    args_str = (
+        arguments
+        if isinstance(arguments, str) and arguments
+        else serialize_tool_arguments(tool_input)
+    )
+    call_id = (
+        tool_id if isinstance(tool_id, str) and tool_id else f"call_{fallback_index}"
+    )
+    name = tool_name if isinstance(tool_name, str) and tool_name else "function"
+
+    return openai_models.ToolCallChunk(
+        index=index,
+        id=str(call_id),
+        type="function",
+        function=openai_models.FunctionCall(
+            name=str(name),
+            arguments=str(args_str),
+        ),
+    )

{ccproxy_api-0.2.4 → ccproxy_api-0.2.6}/ccproxy/llms/formatters/anthropic_to_openai/streams.py

@@ -27,7 +27,7 @@ from ccproxy.llms.models import anthropic as anthropic_models
 from ccproxy.llms.models import openai as openai_models
 from ccproxy.llms.streaming.accumulators import ClaudeAccumulator
 
-from ._helpers import build_openai_tool_call
+from ._helpers import build_openai_tool_call_chunk
 from .requests import _build_responses_payload_from_anthropic_request
 from .responses import convert__anthropic_usage_to_openai_responses__usage
 
@@ -88,10 +88,10 @@ def _anthropic_delta_to_text(
     return None
 
 
-def _build_openai_tool_call(
+def _build_openai_tool_call_chunk(
     accumulator: ClaudeAccumulator,
     block_index: int,
-) -> openai_models.ToolCall | None:
+) -> openai_models.ToolCallChunk | None:
     for tool_call in accumulator.get_complete_tool_calls():
         if tool_call.get("index") != block_index:
             continue
@@ -102,7 +102,8 @@ def _build_openai_tool_call(
         tool_name = function_payload.get("name") or tool_call.get("name")
         arguments = function_payload.get("arguments")
 
-        return build_openai_tool_call(
+        return build_openai_tool_call_chunk(
+            index=tool_call.get("index", block_index),
             tool_id=tool_call.get("id"),
             tool_name=tool_name,
             tool_input=tool_call.get("input", {}),
@@ -1413,7 +1414,7 @@ class AnthropicToOpenAIChatStreamAdapter:
                         continue
                     if block_index in emitted_tool_indices:
                         continue
-                    tool_call = _build_openai_tool_call(accumulator, block_index)
+                    tool_call = _build_openai_tool_call_chunk(accumulator, block_index)
                     if tool_call is None:
                         continue
                     emitted_tool_indices.add(block_index)

{ccproxy_api-0.2.4 → ccproxy_api-0.2.6}/ccproxy/llms/formatters/context.py

@@ -11,6 +11,9 @@ _INSTRUCTIONS_VAR: ContextVar[str | None] = ContextVar(
     "formatter_instructions", default=None
 )
 _TOOLS_VAR: ContextVar[list[Any] | None] = ContextVar("formatter_tools", default=None)
+_OPENAI_THINKING_XML_VAR: ContextVar[bool | None] = ContextVar(
+    "formatter_openai_thinking_xml", default=None
+)
 
 
 def register_request(request: Any | None, instructions: str | None = None) -> None:
@@ -114,3 +117,24 @@ def get_last_request_tools() -> list[Any] | None:
 
     cached = _TOOLS_VAR.get()
     return list(cached) if cached else None
+
+
+def register_openai_thinking_xml(enabled: bool | None) -> None:
+    """Cache OpenAI thinking serialization preference for active conversions.
+
+    Args:
+        enabled: Whether thinking blocks should be serialized with XML wrappers.
+            ``None`` means downstream conversion logic should use its default.
+
+    Note:
+        The value is stored in a ``ContextVar``, so concurrent async requests
+        keep independent preferences without leaking into each other.
+    """
+
+    _OPENAI_THINKING_XML_VAR.set(enabled)
+
+
+def get_openai_thinking_xml() -> bool | None:
+    """Return the OpenAI thinking serialization preference for active conversions."""
+
+    return _OPENAI_THINKING_XML_VAR.get()

{ccproxy_api-0.2.4 → ccproxy_api-0.2.6}/ccproxy/llms/formatters/openai_to_openai/responses.py

@@ -15,6 +15,7 @@ from ccproxy.llms.formatters.common import (
     convert_openai_responses_usage_to_completion_usage,
     merge_thinking_segments,
 )
+from ccproxy.llms.formatters.context import get_openai_thinking_xml
 from ccproxy.llms.models import openai as openai_models
 
 from ._helpers import (
@@ -333,6 +334,10 @@ def convert__openai_responses_to_openai_chat__response(
     response: openai_models.ResponseObject,
 ) -> openai_models.ChatCompletionResponse:
     """Convert an OpenAI ResponseObject to a ChatCompletionResponse."""
+    include_thinking = get_openai_thinking_xml()
+    if include_thinking is None:
+        include_thinking = True
+
     text_segments: list[str] = []
     added_reasoning: set[tuple[str, str]] = set()
     tool_calls: list[openai_models.ToolCall] = []
@@ -353,7 +358,7 @@ def convert__openai_responses_to_openai_chat__response(
                     if thinking_text and len(thinking_text) > 30
                     else thinking_text,
                 )
-                if thinking_text:
+                if include_thinking and thinking_text:
                     key = (signature or "", thinking_text)
                     if key not in added_reasoning:
                         text_segments.append(_wrap_thinking(signature, thinking_text))

{ccproxy_api-0.2.4 → ccproxy_api-0.2.6}/ccproxy/llms/formatters/openai_to_openai/streams.py

@@ -27,16 +27,14 @@ from ccproxy.llms.formatters.context import (
     get_last_instructions,
     get_last_request,
     get_last_request_tools,
+    get_openai_thinking_xml,
     register_request,
     register_request_tools,
 )
 from ccproxy.llms.models import openai as openai_models
 from ccproxy.llms.streaming.accumulators import OpenAIAccumulator
 
-from ._helpers import (
-    _convert_tools_chat_to_responses,
-    _get_attr,
-)
+from ._helpers import _convert_tools_chat_to_responses, _get_attr
 from .requests import _build_responses_payload_from_chat_request
 from .responses import (
     _collect_reasoning_segments,
@@ -61,6 +59,10 @@ class OpenAIResponsesToChatStreamAdapter:
         async def generator() -> AsyncGenerator[
             openai_models.ChatCompletionChunk, None
         ]:
+            include_thinking = get_openai_thinking_xml()
+            if include_thinking is None:
+                include_thinking = True
+
             model_id = ""
             role_sent = False
 
@@ -387,7 +389,8 @@ class OpenAIResponsesToChatStreamAdapter:
 
                     # Emit initial tool call chunk to surface id/name information
                     if not state.initial_emitted:
-                        tool_call = openai_models.ToolCall(
+                        tool_call = openai_models.ToolCallChunk(
+                            index=state.index,
                             id=state.id,
                             type="function",
                             function=openai_models.FunctionCall(
@@ -440,7 +443,8 @@ class OpenAIResponsesToChatStreamAdapter:
                             state.name = guessed
 
                     if state.initial_emitted:
-                        tool_call = openai_models.ToolCall(
+                        tool_call = openai_models.ToolCallChunk(
+                            index=state.index,
                             id=state.id,
                             type="function",
                             function=openai_models.FunctionCall(
@@ -492,7 +496,8 @@ class OpenAIResponsesToChatStreamAdapter:
                             if guessed:
                                 state.name = guessed
 
-                        tool_call = openai_models.ToolCall(
+                        tool_call = openai_models.ToolCallChunk(
+                            index=state.index,
                             id=state.id,
                             type="function",
                             function=openai_models.FunctionCall(
@@ -537,7 +542,7 @@ class OpenAIResponsesToChatStreamAdapter:
                             for entry in summary_list:
                                 text = _get_attr(entry, "text")
                                 signature = _get_attr(entry, "signature")
-                                if isinstance(text, str) and text:
+                                if include_thinking and isinstance(text, str) and text:
                                     chunk_text = _wrap_thinking(signature, text)
                                     sequence_counter += 1
                                     yield openai_models.ChatCompletionChunk(
@@ -584,7 +589,8 @@ class OpenAIResponsesToChatStreamAdapter:
                             if guessed:
                                 state.name = guessed
                         if not state.arguments_emitted:
-                            tool_call = openai_models.ToolCall(
+                            tool_call = openai_models.ToolCallChunk(
+                                index=state.index,
                                 id=state.id,
                                 type="function",
                                 function=openai_models.FunctionCall(
@@ -614,7 +620,8 @@ class OpenAIResponsesToChatStreamAdapter:
 
                     # Emit a patch chunk if the name was never surfaced earlier
                     if state.name and not state.name_emitted:
-                        tool_call = openai_models.ToolCall(
+                        tool_call = openai_models.ToolCallChunk(
+                            index=state.index,
                             id=state.id,
                             type="function",
                             function=openai_models.FunctionCall(

{ccproxy_api-0.2.4 → ccproxy_api-0.2.6}/ccproxy/llms/models/anthropic.py

@@ -352,8 +352,16 @@ class ThinkingConfigDisabled(ThinkingConfigBase):
     type: Literal["disabled"] = Field(default="disabled", alias="type")
 
 
+class ThinkingConfigAdaptive(ThinkingConfigBase):
+    """Configuration for adaptive thinking (Claude 4-6+)."""
+
+    type: Literal["adaptive"] = Field(default="adaptive", alias="type")
+    display: Literal["summarized", "omitted"] | None = None
+
+
 ThinkingConfig = Annotated[
-    ThinkingConfigEnabled | ThinkingConfigDisabled, Field(discriminator="type")
+    ThinkingConfigEnabled | ThinkingConfigDisabled | ThinkingConfigAdaptive,
+    Field(discriminator="type"),
 ]
 
 

{ccproxy_api-0.2.4 → ccproxy_api-0.2.6}/ccproxy/llms/models/openai.py

@@ -185,11 +185,22 @@ class FunctionCall(LlmBaseModel):
 
 
 class ToolCall(LlmBaseModel):
+    """Non-streaming tool call (ChatCompletionMessageToolCall)."""
+
     id: str
     type: Literal["function"] = Field(default="function")
     function: FunctionCall
 
 
+class ToolCallChunk(LlmBaseModel):
+    """Streaming tool call delta (ChoiceDeltaToolCall)."""
+
+    index: int
+    id: str | None = None
+    type: Literal["function"] | None = None
+    function: FunctionCall | None = None
+
+
 class ChatMessage(LlmBaseModel):
     """
     A message within a chat conversation.
@@ -309,7 +320,7 @@ class ChatCompletionResponse(LlmBaseModel):
 class DeltaMessage(LlmBaseModel):
     role: Literal["assistant"] | None = None
     content: str | list[Any] | None = None
-    tool_calls: list[ToolCall] | None = None
+    tool_calls: list[ToolCallChunk] | None = None
     audio: dict[str, Any] | None = None
     reasoning: ResponseMessageReasoning | None = None
 

{ccproxy_api-0.2.4 → ccproxy_api-0.2.6}/ccproxy/plugins/claude_api/adapter.py

@@ -85,15 +85,14 @@ class ClaudeAPIAdapter(BaseHTTPAdapter):
         # Always set Authorization from OAuth-managed access token
         filtered_headers["authorization"] = f"Bearer {token_value}"
 
-        # PATCH: Add Computer Use beta headers for Anthropic API
-        # These are required for browser automation tools to work
+        # Minimal beta tags required for OAuth-based Claude Code auth
         filtered_headers["anthropic-version"] = "2023-06-01"
-        filtered_headers["anthropic-beta"] = "computer-use-2025-01-24"
+        filtered_headers["anthropic-beta"] = "claude-code-20250219,oauth-2025-04-20"
 
-        # Add CLI headers if available, but never allow overriding auth
+        # Add CLI headers if available, but never allow overriding auth or beta
         cli_headers = self._collect_cli_headers()
         if cli_headers:
-            blocked_overrides = {"authorization", "x-api-key"}
+            blocked_overrides = {"authorization", "x-api-key", "anthropic-beta"}
             for key, value in cli_headers.items():
                 lk = key.lower()
                 if lk in blocked_overrides: