lm-deluge 0.0.51__tar.gz → 0.0.53__tar.gz

{lm_deluge-0.0.51/src/lm_deluge.egg-info → lm_deluge-0.0.53}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lm_deluge
-Version: 0.0.51
+Version: 0.0.53
 Summary: Python utility for using LLM API models.
 Author-email: Benjamin Anderson <ben@trytaylor.ai>
 Requires-Python: >=3.10

{lm_deluge-0.0.51 → lm_deluge-0.0.53}/pyproject.toml

@@ -3,7 +3,7 @@ requires = ["setuptools", "wheel"]
 
 [project]
 name = "lm_deluge"
-version = "0.0.51"
+version = "0.0.53"
 authors = [{ name = "Benjamin Anderson", email = "ben@trytaylor.ai" }]
 description = "Python utility for using LLM API models."
 readme = "README.md"

{lm_deluge-0.0.51 → lm_deluge-0.0.53}/src/lm_deluge/api_requests/anthropic.py

@@ -60,7 +60,8 @@ def _build_anthropic_request(
             "type": "enabled",
             "budget_tokens": budget,
         }
-        request_json.pop("top_p")
+        if "top_p" in request_json:
+            request_json["top_p"] = max(request_json["top_p"], 0.95)
         request_json["temperature"] = 1.0
         request_json["max_tokens"] += budget
     else:
@@ -70,6 +71,11 @@ def _build_anthropic_request(
     if system_message is not None:
         request_json["system"] = system_message
 
+    # handle temp + top_p for opus 4.1/sonnet 4.5
+    if model.name in ["claude-sonnet-4-5-20250929", "claude-opus-4-1-20250805"]:
+        if "temperature" in request_json and "top_p" in request_json:
+            request_json.pop("top_p")
+
     if tools:
         mcp_servers = []
         tool_definitions = []
@@ -89,6 +95,9 @@ def _build_anthropic_request(
                     _add_beta(base_headers, "computer-use-2025-01-24")
                 elif tool["type"] == "code_execution_20250522":
                     _add_beta(base_headers, "code-execution-2025-05-22")
+                elif tool["type"] in ["memory_20250818", "clear_tool_uses_20250919"]:
+                    _add_beta(base_headers, "context-management-2025-06-27")
+
             elif isinstance(tool, MCPServer):
                 _add_beta(base_headers, "mcp-client-2025-04-04")
                 mcp_servers.append(tool.for_anthropic())

{lm_deluge-0.0.51 → lm_deluge-0.0.53}/src/lm_deluge/api_requests/openai.py

@@ -1,5 +1,6 @@
 import json
 import os
+import traceback as tb
 import warnings
 from types import SimpleNamespace
 
@@ -341,7 +342,13 @@ class OpenAIResponsesRequest(APIRequestBase):
                                     elif content_item.get("type") == "refusal":
                                         parts.append(Text(content_item["refusal"]))
                             elif item.get("type") == "reasoning":
-                                parts.append(Thinking(item["summary"]["text"]))
+                                summary = item["summary"]
+                                if not summary:
+                                    continue
+                                if isinstance(summary, list) and len(summary) > 0:
+                                    summary = summary[0]
+                                assert isinstance(summary, dict), "summary isn't a dict"
+                                parts.append(Thinking(summary["text"]))
                             elif item.get("type") == "function_call":
                                 parts.append(
                                     ToolCall(
@@ -432,6 +439,9 @@ class OpenAIResponsesRequest(APIRequestBase):
                 except Exception as e:
                     is_error = True
                     error_message = f"Error parsing {self.model.name} responses API response: {str(e)}"
+                    print("got data:", data)
+                    traceback = tb.format_exc()
+                    print(f"Error details:\n{traceback}")
 
         elif mimetype and "json" in mimetype.lower():
             print("is_error True, json response")

{lm_deluge-0.0.51 → lm_deluge-0.0.53}/src/lm_deluge/api_requests/response.py

@@ -14,7 +14,7 @@ class APIResponse:
     # request information
     id: int  # should be unique to the request within a given prompt-processing call
     model_internal: str  # our internal model tag
-    prompt: Conversation | dict
+    prompt: Conversation | dict  # dict if converted to log
     sampling_params: SamplingParams
 
     # http response information

{lm_deluge-0.0.51 → lm_deluge-0.0.53}/src/lm_deluge/client.py

@@ -357,6 +357,8 @@ class _LLMClient(BaseModel):
         prompts = prompts_to_conversations(prompts)
         ids = list(range(len(prompts)))
         results: list[APIResponse | None] = [None for _ in range(len(prompts))]
+        contexts: list[RequestContext | None] = [None for _ in range(len(prompts))]
+        inflight_tasks: set[asyncio.Task[None]] = set()
         # Use existing tracker if client has been opened; otherwise open/close automatically
         tracker: StatusTracker
         tracker_preopened = self._tracker is not None
@@ -419,12 +421,14 @@ class _LLMClient(BaseModel):
                 )
 
                 # Launch simplified request processing
+                contexts[next_context.task_id] = next_context
+
                 async def process_and_store(ctx: RequestContext):
                     try:
                         response = await self.process_single_request(ctx, retry_queue)
                         results[ctx.task_id] = response
-                    except Exception as e:
-                        # Create an error response for validation errors and other exceptions
+                    except BaseException as exc:
+                        # Capture cancellations and other BaseExceptions before fallback response fires.
                         error_response = APIResponse(
                             id=ctx.task_id,
                             model_internal=ctx.model_name,
@@ -432,14 +436,17 @@ class _LLMClient(BaseModel):
                             sampling_params=ctx.sampling_params,
                             status_code=None,
                             is_error=True,
-                            error_message=str(e),
+                            error_message=f"{type(exc).__name__}: {exc}",
+                            raw_response={"exception_repr": repr(exc)},
                         )
                         results[ctx.task_id] = error_response
-                        # Mark task as completed so the main loop can finish
                         if ctx.status_tracker:
                             ctx.status_tracker.task_failed(ctx.task_id)
+                        raise
 
-                asyncio.create_task(process_and_store(next_context))
+                task = asyncio.create_task(process_and_store(next_context))
+                inflight_tasks.add(task)
+                task.add_done_callback(inflight_tasks.discard)
                 next_context = None  # Reset after successful dispatch
                 next_is_retry = False
 
@@ -456,9 +463,37 @@ class _LLMClient(BaseModel):
             # Yield briefly to allow in-flight tasks to progress
             await asyncio.sleep(min(0.01, seconds_to_sleep_each_loop))
 
+        if inflight_tasks:
+            await asyncio.gather(*inflight_tasks, return_exceptions=True)
+
         if not tracker_preopened:
             self.close()
 
+        for idx, response in enumerate(results):
+            if response is None:
+                ctx = contexts[idx]
+                prompt = ctx.prompt if ctx else prompts[idx]
+                sampling_params = (
+                    ctx.sampling_params
+                    if ctx
+                    else self.sampling_params[0]
+                    if self.sampling_params
+                    else SamplingParams()
+                )
+                model_name = ctx.model_name if ctx else self.model_names[0]
+                assert isinstance(
+                    prompt, Conversation
+                ), "expected prompt to be a conversation"
+                results[idx] = APIResponse(
+                    id=idx,
+                    model_internal=model_name,
+                    prompt=prompt,
+                    sampling_params=sampling_params,
+                    status_code=None,
+                    is_error=True,
+                    error_message="Internal error: no response produced.",
+                )
+
         if return_completions_only:
             return [r.completion if r is not None else None for r in results]
 

{lm_deluge-0.0.51 → lm_deluge-0.0.53}/src/lm_deluge/models/anthropic.py

@@ -10,6 +10,18 @@ ANTHROPIC_MODELS = {
     #                                                                   ░███
     #                                                                   █████
     #
+    "claude-4.5-sonnet": {
+        "id": "claude-4.5-sonnet",
+        "name": "claude-sonnet-4-5-20250929",
+        "api_base": "https://api.anthropic.com/v1",
+        "api_key_env_var": "ANTHROPIC_API_KEY",
+        "supports_json": False,
+        "api_spec": "anthropic",
+        "input_cost": 3.0,
+        "output_cost": 15.0,
+        "requests_per_minute": 4_000,
+        "tokens_per_minute": 400_000,
+    },
     "claude-4.1-opus": {
         "id": "claude-4.1-opus",
         "name": "claude-opus-4-1-20250805",

{lm_deluge-0.0.51 → lm_deluge-0.0.53/src/lm_deluge.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lm_deluge
-Version: 0.0.51
+Version: 0.0.53
 Summary: Python utility for using LLM API models.
 Author-email: Benjamin Anderson <ben@trytaylor.ai>
 Requires-Python: >=3.10