spl-agent-engine 0.3.0__tar.gz → 0.4.0__tar.gz

{spl_agent_engine-0.3.0/spl_agent_engine.egg-info → spl_agent_engine-0.4.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: spl-agent-engine
-Version: 0.3.0
+Version: 0.4.0
 Summary: Anthropomorphic psychology engine built on SPL Pure Core V8.0 — deterministic emotion, trauma, memory, trust, and defense mechanisms.
 Author-email: NOHN-AI <ai@nohnlins.com>
 Maintainer-email: NOHN-AI <ai@nohnlins.com>

{spl_agent_engine-0.3.0 → spl_agent_engine-0.4.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "spl-agent-engine"
-version = "0.3.0"
+version = "0.4.0"
 description = "Anthropomorphic psychology engine built on SPL Pure Core V8.0 — deterministic emotion, trauma, memory, trust, and defense mechanisms."
 readme = { file = "README.md", content-type = "text/markdown" }
 requires-python = ">=3.8"

{spl_agent_engine-0.3.0 → spl_agent_engine-0.4.0}/spl_agent_engine/__init__.py

@@ -24,6 +24,7 @@ from .core import (
     SPLPureCoreV7_3,
     NarrativeMapper,
     TokenUsage,
+    TokenStats,
     LLMAdapter,
     OpenAIAdapter,
     ClaudeAdapter,
@@ -34,10 +35,10 @@ from .core import (
 # Canonical name for V8.0 (class name kept for backward compatibility)
 SPLPureCore = SPLPureCoreV7_3
 
-__version__ = "0.3.0"
+__version__ = "0.4.0"
 __all__ = [
     "SPLPureCore", "SPLPureCoreV7_3", "NarrativeMapper", "AuditLogger",
-    "TokenUsage",
+    "TokenUsage", "TokenStats",
     "LLMAdapter", "OpenAIAdapter", "ClaudeAdapter", "ChainAdapter",
     "generate_line_with_llm",
 ]

{spl_agent_engine-0.3.0 → spl_agent_engine-0.4.0}/spl_agent_engine/core.py

@@ -123,6 +123,114 @@ class TokenUsage:
     model: str = ""
 
 
+# ================================================================
+# Token 累计统计器 — 跨调用汇总 token 用量
+# ================================================================
+class TokenStats:
+    """跨多次 LLM 调用累计统计 token 用量。
+
+    功能：
+      - 累计 prompt / completion / total tokens
+      - 按模型分别统计
+      - 记录每次调用明细
+      - 导出 JSON 汇总报告
+
+    用法：
+        stats = TokenStats()
+        stats.record(usage)                # 传入 TokenUsage
+        stats.record(usage2)
+        stats.summary()                    # 汇总 dict
+        stats.export_json("stats.json")    # 导出文件
+    """
+
+    def __init__(self):
+        self._total_prompt: int = 0
+        self._total_completion: int = 0
+        self._total_tokens: int = 0
+        self._call_count: int = 0
+        self._by_model: Dict[str, Dict[str, int]] = {}
+        self._history: List[Dict[str, Any]] = []
+
+    def record(self, usage: TokenUsage) -> None:
+        """记录一次 LLM 调用的 token 用量。"""
+        if usage is None:
+            return
+        self._total_prompt += usage.prompt_tokens
+        self._total_completion += usage.completion_tokens
+        self._total_tokens += usage.total_tokens
+        self._call_count += 1
+
+        model = usage.model or "unknown"
+        if model not in self._by_model:
+            self._by_model[model] = {
+                "prompt_tokens": 0,
+                "completion_tokens": 0,
+                "total_tokens": 0,
+                "call_count": 0,
+            }
+        m = self._by_model[model]
+        m["prompt_tokens"] += usage.prompt_tokens
+        m["completion_tokens"] += usage.completion_tokens
+        m["total_tokens"] += usage.total_tokens
+        m["call_count"] += 1
+
+        self._history.append({
+            "seq": self._call_count,
+            "model": model,
+            "prompt_tokens": usage.prompt_tokens,
+            "completion_tokens": usage.completion_tokens,
+            "total_tokens": usage.total_tokens,
+        })
+
+    def summary(self) -> Dict[str, Any]:
+        """返回累计统计汇总。"""
+        return {
+            "call_count": self._call_count,
+            "total_prompt_tokens": self._total_prompt,
+            "total_completion_tokens": self._total_completion,
+            "total_tokens": self._total_tokens,
+            "by_model": dict(self._by_model),
+        }
+
+    def history(self) -> List[Dict[str, Any]]:
+        """返回每次调用的明细记录。"""
+        return list(self._history)
+
+    def reset(self) -> None:
+        """清空所有统计。"""
+        self._total_prompt = 0
+        self._total_completion = 0
+        self._total_tokens = 0
+        self._call_count = 0
+        self._by_model.clear()
+        self._history.clear()
+
+    def export_json(self, path: str) -> None:
+        """将统计汇总 + 明细导出为 JSON 文件。"""
+        data = {
+            "summary": self.summary(),
+            "history": self.history(),
+        }
+        with open(path, "w", encoding="utf-8") as f:
+            json.dump(data, f, ensure_ascii=False, indent=2)
+
+    @property
+    def total_tokens(self) -> int:
+        return self._total_tokens
+
+    @property
+    def total_prompt_tokens(self) -> int:
+        return self._total_prompt
+
+    @property
+    def total_completion_tokens(self) -> int:
+        return self._total_completion
+
+    @property
+    def call_count(self) -> int:
+        return self._call_count
+
+
 # ================================================================
 # 1. 外部叙事映射器（"自设：偏见/世界观/价值观"——纯外部策略）
 #    用户可随意替换，无需修改核心代码。

{spl_agent_engine-0.3.0 → spl_agent_engine-0.4.0/spl_agent_engine.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: spl-agent-engine
-Version: 0.3.0
+Version: 0.4.0
 Summary: Anthropomorphic psychology engine built on SPL Pure Core V8.0 — deterministic emotion, trauma, memory, trust, and defense mechanisms.
 Author-email: NOHN-AI <ai@nohnlins.com>
 Maintainer-email: NOHN-AI <ai@nohnlins.com>