local-openai2anthropic 0.3.7__py3-none-any.whl → 0.3.8__py3-none-any.whl

local_openai2anthropic/__init__.py

@@ -3,7 +3,7 @@
 local-openai2anthropic: A proxy server that converts Anthropic Messages API to OpenAI API.
 """
 
-__version__ = "0.3.7"
+__version__ = "0.3.8"
 
 from local_openai2anthropic.protocol import (
     AnthropicError,

local_openai2anthropic/main.py

@@ -101,7 +101,7 @@ def create_app(settings: Settings | None = None) -> FastAPI:
     app = FastAPI(
         title="local-openai2anthropic",
         description="A proxy server that converts Anthropic Messages API to OpenAI API",
-        version="0.3.7",
+        version="0.3.8",
         docs_url="/docs",
         redoc_url="/redoc",
     )
@@ -253,7 +253,7 @@ Examples:
     parser.add_argument(
         "--version",
         action="version",
-        version="%(prog)s 0.3.7",
+        version="%(prog)s 0.3.8",
     )
 
     # Create subparsers for commands

local_openai2anthropic/protocol.py

@@ -41,7 +41,7 @@ from anthropic.types.message_create_params import MessageCreateParams
 
 class UsageWithCache(BaseModel):
     """Extended usage with cache token support."""
-    
+
     input_tokens: int
     output_tokens: int
     cache_creation_input_tokens: Optional[int] = None
@@ -50,14 +50,14 @@ class UsageWithCache(BaseModel):
 
 class AnthropicError(BaseModel):
     """Error structure for Anthropic API."""
-    
+
     type: str
     message: str
 
 
 class AnthropicErrorResponse(BaseModel):
     """Error response structure for Anthropic API."""
-    
+
     type: str = "error"
     error: AnthropicError
 
@@ -70,6 +70,7 @@ class PingEvent(BaseModel):
 
 # Web Search Tool Types
 
+
 class ApproximateLocation(BaseModel):
     """Approximate user location for web search."""
 
@@ -122,7 +123,13 @@ class WebSearchToolResultError(BaseModel):
     """Error content for web search tool result."""
 
     type: Literal["web_search_tool_result_error"] = "web_search_tool_result_error"
-    error_code: Literal["max_uses_exceeded", "too_many_requests", "unavailable"]
+    error_code: Literal[
+        "invalid_input",
+        "max_uses_exceeded",
+        "query_too_long",
+        "too_many_requests",
+        "unavailable",
+    ]
 
 
 class WebSearchToolResult(BaseModel):
@@ -131,6 +138,7 @@ class WebSearchToolResult(BaseModel):
     type: Literal["web_search_tool_result"] = "web_search_tool_result"
     tool_use_id: str
     results: list[WebSearchResult] | WebSearchToolResultError  # 'results' for client
+    content: list[WebSearchResult] | WebSearchToolResultError | None = None  # 'content' for compatibility
 
 
 class WebSearchCitation(BaseModel):
@@ -168,12 +176,10 @@ __all__ = [
     "TextBlockParam",
     "ToolUseBlockParam",
     "ToolResultBlockParam",
-
     # Message types
     "Message",
     "MessageParam",
     "MessageCreateParams",
-
     # Streaming events
     "MessageStreamEvent",
     "MessageStartEvent",
@@ -183,23 +189,18 @@ __all__ = [
     "ContentBlockDeltaEvent",
     "ContentBlockStopEvent",
     "PingEvent",
-
     # Delta types
     "TextDelta",
     "BetaThinkingDelta",
-
     # Usage
     "UsageWithCache",
     "UsageWithServerToolUse",
     "MessageDeltaUsage",
-
     # Config
     "BetaThinkingConfigParam",
-
     # Error
     "AnthropicError",
     "AnthropicErrorResponse",
-
     # Web Search Tool Types
     "ApproximateLocation",
     "WebSearchToolDefinition",

local_openai2anthropic/server_tools/web_search.py

@@ -148,30 +148,39 @@ class WebSearchServerTool(ServerTool):
         blocks: list[dict[str, Any]] = []
 
         # 1. server_tool_use block - signals a server-side tool was invoked
-        blocks.append({
-            "type": "server_tool_use",
-            "id": call_id,
-            "name": cls.tool_name,
-            "input": call_args,
-        })
+        blocks.append(
+            {
+                "type": "server_tool_use",
+                "id": call_id,
+                "name": cls.tool_name,
+                "input": call_args,
+            }
+        )
 
         # 2. web_search_tool_result block - contains the search results
-        # Note: Claude Code client expects 'results' field (not 'content') for counting
+        # Provide both 'results' and 'content' for client compatibility.
         if result.success:
-            blocks.append({
-                "type": "web_search_tool_result",
-                "tool_use_id": call_id,
-                "results": result.content,
-            })
+            blocks.append(
+                {
+                    "type": "web_search_tool_result",
+                    "tool_use_id": call_id,
+                    "results": result.content,
+                    "content": result.content,
+                }
+            )
         else:
-            blocks.append({
-                "type": "web_search_tool_result",
-                "tool_use_id": call_id,
-                "results": {
-                    "type": "web_search_tool_result_error",
-                    "error_code": result.error_code,
-                },
-            })
+            error_payload = {
+                "type": "web_search_tool_result_error",
+                "error_code": result.error_code or "unavailable",
+            }
+            blocks.append(
+                {
+                    "type": "web_search_tool_result",
+                    "tool_use_id": call_id,
+                    "results": error_payload,
+                    "content": error_payload,
+                }
+            )
 
         return blocks
 
@@ -190,7 +199,7 @@ class WebSearchServerTool(ServerTool):
                     {
                         "url": item.get("url"),
                         "title": item.get("title"),
-                        "snippet": item.get("snippet"),
+                        "content": item.get("encrypted_content"),
                         "page_age": item.get("page_age"),
                     }
                     for item in result.content

local_openai2anthropic/tavily_client.py

@@ -57,12 +57,16 @@ class TavilyClient:
 
         Returns:
             Tuple of (list of WebSearchResult, error_code or None).
-            Error codes: "max_uses_exceeded", "too_many_requests", "unavailable"
+            Error codes: "invalid_input", "query_too_long", "too_many_requests", "unavailable"
         """
         if not self._enabled:
             logger.warning("Tavily search called but API key not configured")
             return [], "unavailable"
 
+        if not query or not query.strip():
+            logger.warning("Tavily search called with empty query")
+            return [], "invalid_input"
+
         url = f"{self.base_url}/search"
         headers = {
             "Content-Type": "application/json",
@@ -80,6 +84,14 @@ class TavilyClient:
             async with httpx.AsyncClient(timeout=self.timeout) as client:
                 response = await client.post(url, headers=headers, json=payload)
 
+                if response.status_code == 400:
+                    logger.warning("Tavily invalid request")
+                    return [], "invalid_input"
+
+                if response.status_code == 413:
+                    logger.warning("Tavily query too long")
+                    return [], "query_too_long"
+
                 if response.status_code == 429:
                     logger.warning("Tavily rate limit exceeded")
                     return [], "too_many_requests"
@@ -88,6 +100,11 @@ class TavilyClient:
                     logger.error(f"Tavily server error: {response.status_code}")
                     return [], "unavailable"
 
+                # Only raise for status codes we haven't handled (e.g., 401, 403)
+                if response.status_code >= 400:
+                    logger.error(f"Tavily client error: {response.status_code}")
+                    return [], "unavailable"
+
                 response.raise_for_status()
                 data = response.json()
 
@@ -102,7 +119,9 @@ class TavilyClient:
                     )
                     results.append(result)
 
-                logger.debug(f"Tavily search returned {len(results)} results for query: {query[:50]}...")
+                logger.debug(
+                    f"Tavily search returned {len(results)} results for query: {query[:50]}..."
+                )
                 return results, None
 
         except httpx.TimeoutException:

{local_openai2anthropic-0.3.7.dist-info → local_openai2anthropic-0.3.8.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: local-openai2anthropic
-Version: 0.3.7
+Version: 0.3.8
 Summary: A lightweight proxy server that converts Anthropic Messages API to OpenAI API
 Project-URL: Homepage, https://github.com/dongfangzan/local-openai2anthropic
 Project-URL: Repository, https://github.com/dongfangzan/local-openai2anthropic
@@ -109,37 +109,47 @@ Examples:
 
 > **Note:** If you're using [Ollama](https://ollama.com), it natively supports the Anthropic API format, so you don't need this proxy. Just point your Claude SDK directly to `http://localhost:11434/v1`.
 
-### 3. Start the Proxy
+### 3. Start the Proxy (Recommended)
 
-**Option A: Run in background (recommended)**
+Run the following command to start the proxy in background mode:
 
 ```bash
-export OA2A_OPENAI_BASE_URL=http://localhost:8000/v1  # Your local LLM endpoint
-export OA2A_OPENAI_API_KEY=dummy  # Any value, not used by local backends
+oa2a start
+```
+
+**First-time setup**: If `~/.oa2a/config.toml` doesn't exist, an interactive setup wizard will guide you through:
+- Enter your OpenAI API Key (for the local LLM backend)
+- Enter the base URL of your local LLM (e.g., `http://localhost:8000/v1`)
+- Configure server host and port (optional)
+- Set server API key for authentication (optional)
 
-oa2a start              # Start server in background
-# Server starts at http://localhost:8080
+After configuration, the server starts at `http://localhost:8080`.
 
-# View logs
+**Daemon management commands:**
+
+```bash
 oa2a logs               # Show last 50 lines of logs
 oa2a logs -f            # Follow logs in real-time (Ctrl+C to exit)
-
-# Check status
 oa2a status             # Check if server is running
-
-# Stop server
 oa2a stop               # Stop background server
-
-# Restart server
 oa2a restart            # Restart with same settings
 ```
 
+**Manual Configuration**
+
+You can also manually create/edit the config file at `~/.oa2a/config.toml`:
+
+```toml
+# OA2A Configuration File
+openai_api_key = "dummy"
+openai_base_url = "http://localhost:8000/v1"
+host = "0.0.0.0"
+port = 8080
+```
+
 **Option B: Run in foreground**
 
 ```bash
-export OA2A_OPENAI_BASE_URL=http://localhost:8000/v1
-export OA2A_OPENAI_API_KEY=dummy
-
 oa2a                    # Run server in foreground (blocking)
 # Press Ctrl+C to stop
 ```
@@ -195,15 +205,6 @@ You can configure [Claude Code](https://github.com/anthropics/claude-code) to us
 | `ANTHROPIC_DEFAULT_HAIKU_MODEL` | Default model for Haiku mode |
 | `ANTHROPIC_REASONING_MODEL` | Default model for reasoning tasks |
 
-2. **Or set environment variables** before running Claude Code:
-
-```bash
-export ANTHROPIC_BASE_URL=http://localhost:8080
-export ANTHROPIC_API_KEY=dummy-key
-
-claude
-```
-
 ### Complete Workflow Example
 
 Make sure `~/.claude/settings.json` is configured as described above.
@@ -215,10 +216,7 @@ vllm serve meta-llama/Llama-2-7b-chat-hf
 
 Terminal 2 - Start the proxy (background mode):
 ```bash
-export OA2A_OPENAI_BASE_URL=http://localhost:8000/v1
-export OA2A_OPENAI_API_KEY=dummy
-export OA2A_TAVILY_API_KEY="tvly-your-tavily-api-key"  # Optional: enable web search
-
+# First run: interactive setup wizard will guide you
 oa2a start
 ```
 
@@ -241,102 +239,43 @@ oa2a stop
 - ✅ **Streaming responses** - Real-time token streaming via SSE
 - ✅ **Tool calling** - Local LLM function calling support
 - ✅ **Vision models** - Multi-modal input for vision-capable models
-- ✅ **Web Search** - Give your local LLM internet access (see below)
+- ✅ **Web Search** - Built-in Tavily web search for local models
 - ✅ **Thinking mode** - Supports reasoning/thinking model outputs
 
 ---
 
-## Web Search Capability 🔍
-
-**Bridge the gap: Give your local LLM the web search power that Claude Code users enjoy!**
-
-When using locally-hosted models with Claude Code, you lose access to the built-in web search tool. This proxy fills that gap by providing a server-side web search implementation powered by [Tavily](https://tavily.com).
+## Web Search 🔍
 
-### The Problem
+Enable web search for your local LLM using [Tavily](https://tavily.com).
 
-| Scenario | Web Search Available? |
-|----------|----------------------|
-| Using Claude (cloud) in Claude Code | ✅ Built-in |
-| Using local vLLM/SGLang in Claude Code | ❌ Not available |
-| **Using this proxy + local LLM** | ✅ **Enabled via Tavily** |
+**Setup:**
 
-### How It Works
+1. Get a free API key at [tavily.com](https://tavily.com)
 
+2. Add to your config (`~/.oa2a/config.toml`):
+```toml
+tavily_api_key = "tvly-your-api-key"
 ```
-Claude Code → Anthropic SDK → This Proxy → Local LLM
-                                      ↓
-                                 Tavily API (Web Search)
-```
-
-The proxy intercepts `web_search_20250305` tool calls and handles them directly, regardless of whether your local model supports web search natively.
-
-### Setup Tavily Search
 
-1. **Get a free API key** at [tavily.com](https://tavily.com) - generous free tier available
+3. Use `web_search_20250305` tool in your app - the proxy handles search automatically.
 
-2. **Configure the proxy:**
-```bash
-export OA2A_OPENAI_BASE_URL=http://localhost:8000/v1
-export OA2A_OPENAI_API_KEY=dummy
-export OA2A_TAVILY_API_KEY="tvly-your-tavily-api-key"  # Enable web search
-
-oa2a
-```
-
-3. **Use in your app:**
-```python
-import anthropic
-
-client = anthropic.Anthropic(
-    base_url="http://localhost:8080",
-    api_key="dummy-key",
-)
-
-message = client.messages.create(
-    model="meta-llama/Llama-2-7b-chat-hf",
-    max_tokens=1024,
-    tools=[
-        {
-            "name": "web_search_20250305",
-            "description": "Search the web for current information",
-            "input_schema": {
-                "type": "object",
-                "properties": {
-                    "query": {"type": "string", "description": "Search query"},
-                },
-                "required": ["query"],
-            },
-        }
-    ],
-    messages=[{"role": "user", "content": "What happened in AI today?"}],
-)
-
-if message.stop_reason == "tool_use":
-    tool_use = message.content[-1]
-    print(f"Searching: {tool_use.input}")
-    # The proxy automatically calls Tavily and returns results
-```
-
-### Tavily Configuration Options
-
-| Variable | Default | Description |
-|----------|---------|-------------|
-| `OA2A_TAVILY_API_KEY` | - | Your Tavily API key ([get free at tavily.com](https://tavily.com)) |
-| `OA2A_TAVILY_MAX_RESULTS` | 5 | Number of search results to return |
-| `OA2A_TAVILY_TIMEOUT` | 30 | Search timeout in seconds |
-| `OA2A_WEBSEARCH_MAX_USES` | 5 | Max search calls per request |
+**Options:** `tavily_max_results` (default: 5), `tavily_timeout` (default: 30), `websearch_max_uses` (default: 5)
 
 ---
 
 ## Configuration
 
-| Variable | Required | Default | Description |
+Config file: `~/.oa2a/config.toml` (auto-created on first run)
+
+| Option | Required | Default | Description |
 |----------|----------|---------|-------------|
-| `OA2A_OPENAI_BASE_URL` | ✅ | - | Your local LLM's OpenAI-compatible endpoint |
-| `OA2A_OPENAI_API_KEY` | ✅ | - | Any value (local backends usually ignore this) |
-| `OA2A_PORT` | ❌ | 8080 | Proxy server port |
-| `OA2A_HOST` | ❌ | 0.0.0.0 | Proxy server host |
-| `OA2A_TAVILY_API_KEY` | ❌ | - | Enable web search ([tavily.com](https://tavily.com)) |
+| `openai_base_url` | ✅ | - | Local LLM endpoint (e.g., `http://localhost:8000/v1`) |
+| `openai_api_key` | ✅ | - | API key for local LLM |
+| `port` | ❌ | 8080 | Proxy port |
+| `host` | ❌ | 0.0.0.0 | Proxy host |
+| `api_key` | ❌ | - | Auth key for this proxy |
+| `tavily_api_key` | ❌ | - | Enable web search |
+| `log_level` | ❌ | INFO | DEBUG, INFO, WARNING, ERROR |
 
 ---
 

{local_openai2anthropic-0.3.7.dist-info → local_openai2anthropic-0.3.8.dist-info}/RECORD

@@ -1,25 +1,25 @@
-local_openai2anthropic/__init__.py,sha256=ykkqdWKgqvSffE2-GmDcZQFHJRAe0y4d4aeD3fUq_Os,1059
+local_openai2anthropic/__init__.py,sha256=gFCH8riHg-nyPD7S0l1TAAKYwgzp4BLzPdUEn4Ce5dM,1059
 local_openai2anthropic/__main__.py,sha256=K21u5u7FN8-DbO67TT_XDF0neGqJeFrVNkteRauCRQk,179
 local_openai2anthropic/config.py,sha256=Cjg6J7H7ydKtVSd5m0RlTj-YF6yht3TpF4LcyodqQP4,9621
 local_openai2anthropic/converter.py,sha256=og94I514M9km_Wbk9c1ddU6fyaQNEbpd2zfpfnBQaTQ,16029
 local_openai2anthropic/daemon.py,sha256=pZnRojGFcuIpR8yLDNjV-b0LJRBVhgRAa-dKeRRse44,10017
 local_openai2anthropic/daemon_runner.py,sha256=rguOH0PgpbjqNsKYei0uCQX8JQOQ1wmtQH1CtW95Dbw,3274
-local_openai2anthropic/main.py,sha256=FPCEATNPXvGpkszftdXJh0o0F5sUAOXo2zDagmsWGKI,12174
+local_openai2anthropic/main.py,sha256=AaG_MKznk_XA0PYFgTVby7ykOFhdOJGq7fB9YPx8cBM,12174
 local_openai2anthropic/openai_types.py,sha256=jFdCvLwtXYoo5gGRqOhbHQcVaxcsxNnCP_yFPIv7rG4,3823
-local_openai2anthropic/protocol.py,sha256=VW3B1YrbYg5UAo7PveQv0Ny5vfuNa6yG6IlHtkuyXiI,5178
+local_openai2anthropic/protocol.py,sha256=B6nL9z86vIEiuYPvndYc7DA_zrMh7GLzhNmO67M9f3g,5349
 local_openai2anthropic/router.py,sha256=gwSGCYQGd0tAj4B4cl30UDkIJDIfBP4D8T9KEMKnxyk,16196
-local_openai2anthropic/tavily_client.py,sha256=QsBhnyF8BFWPAxB4XtWCCpHCquNL5SW93-zjTTi4Meg,3774
+local_openai2anthropic/tavily_client.py,sha256=b_gyW0rLPIMNJYAFfeZkStbPYisuQzADCUZXzXRdyKo,4548
 local_openai2anthropic/server_tools/__init__.py,sha256=QlJfjEta-HOCtLe7NaY_fpbEKv-ZpInjAnfmSqE9tbk,615
 local_openai2anthropic/server_tools/base.py,sha256=pNFsv-jSgxVrkY004AHAcYMNZgVSO8ZOeCzQBUtQ3vU,5633
-local_openai2anthropic/server_tools/web_search.py,sha256=1C7lX_cm-tMaN3MsCjinEZYPJc_Hj4yAxYay9h8Zbvs,6543
+local_openai2anthropic/server_tools/web_search.py,sha256=98EYdpS9kAAoSHl9vdUABPCYqYFmo4YSPSBZSGyUI40,6812
 local_openai2anthropic/streaming/__init__.py,sha256=RFKYQnc0zlhWK-Dm7GZpmabmszbZhY5NcXaaSsQ7Sys,227
 local_openai2anthropic/streaming/handler.py,sha256=X8viml6b40p-vr-A4HlEi5iCqmTsIMyQgj3S2RfweVE,22033
 local_openai2anthropic/tools/__init__.py,sha256=OM_6YAwy3G1kbrF7n5NvmBwWPGO0hwq4xLrYZFMHANA,318
 local_openai2anthropic/tools/handler.py,sha256=SO8AmEUfNIg16s6jOKBaYdajYc0fiI8ciOoiKXIJe_c,14106
 local_openai2anthropic/utils/__init__.py,sha256=0Apd3lQCmWpQHol4AfjtQe6A3Cpex9Zn-8dyK_FU8Z0,372
 local_openai2anthropic/utils/tokens.py,sha256=TV3vGAjoGZeyo1xPvwb5jto43p1U1f4HteCApB86X0g,3187
-local_openai2anthropic-0.3.7.dist-info/METADATA,sha256=ZqgaeWvxJAKD1fDKC0XWiSwf3BI1DbH5UQpjvOzFewo,11270
-local_openai2anthropic-0.3.7.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
-local_openai2anthropic-0.3.7.dist-info/entry_points.txt,sha256=hdc9tSJUNxyNLXcTYye5SuD2K0bEQhxBhGnWTFup6ZM,116
-local_openai2anthropic-0.3.7.dist-info/licenses/LICENSE,sha256=X3_kZy3lJvd_xp8IeyUcIAO2Y367MXZc6aaRx8BYR_s,11369
-local_openai2anthropic-0.3.7.dist-info/RECORD,,
+local_openai2anthropic-0.3.8.dist-info/METADATA,sha256=uUbJJfhlsv-I54la3bD_Dkpmpe-pnUku60wgkXzXMVM,9276
+local_openai2anthropic-0.3.8.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
+local_openai2anthropic-0.3.8.dist-info/entry_points.txt,sha256=hdc9tSJUNxyNLXcTYye5SuD2K0bEQhxBhGnWTFup6ZM,116
+local_openai2anthropic-0.3.8.dist-info/licenses/LICENSE,sha256=X3_kZy3lJvd_xp8IeyUcIAO2Y367MXZc6aaRx8BYR_s,11369
+local_openai2anthropic-0.3.8.dist-info/RECORD,,