modelfunction-cli 0.1.2 → 0.1.3

package/README.md

@@ -5,7 +5,7 @@
 ## 安装
 
 ```bash
-npx -y modelfunction-cli@0.1.2 --version
+npx -y modelfunction-cli@0.1.3 --version
 # 或
 npm i -D modelfunction-cli
 npx mf --version
@@ -83,7 +83,7 @@ INTERNAL_LLM_KEY=... mf eval --suite mf.suite.json \
   "mcpServers": {
     "modelfunction": {
       "command": "npx",
-      "args": ["-y", "modelfunction-cli@0.1.2", "mcp"]
+      "args": ["-y", "modelfunction-cli@0.1.3", "mcp"]
     }
   }
 }
@@ -105,7 +105,7 @@ INTERNAL_LLM_KEY=... mf eval --suite mf.suite.json \
 
 ```bash
 # 1. 把 mf 挂成 MCP server
-claude mcp add modelfunction -- npx -y modelfunction-cli@0.1.2 mcp
+claude mcp add modelfunction -- npx -y modelfunction-cli@0.1.3 mcp
 ```
 
 ```text

package/lib/cases.js

@@ -167,15 +167,63 @@ export function suspectedPairs(tools) {
       if (union.size === 0) continue;
       let shared = 0;
       for (const token of union) if (a.has(token) && b.has(token)) shared += 1;
-      const similarity = shared / union.size;
+      const lexical = shared / union.size;
+      const structural = structuralScore(tools[i], tools[j]);
+      const similarity = Math.max(lexical, structural);
       if (similarity >= PAIR_SIMILARITY) {
-        pairs.push({ a: tools[i].name, b: tools[j].name, similarity });
+        pairs.push({
+          a: tools[i].name,
+          b: tools[j].name,
+          similarity,
+          basis: structural > lexical ? "structural" : "lexical"
+        });
       }
     }
   }
   return pairs.sort((x, y) => y.similarity - x.similarity);
 }
 
+// Tools that share part of the name (search / file / read...) and the same
+// parameter shape confuse the model even when the wording differs, which the
+// lexical score alone misses. Name order is unreliable: `tavily_search` leads
+// with the vendor, `search_docs` with the action.
+function structuralScore(x, y) {
+  const left = nameTokens(x.name);
+  const right = nameTokens(y.name);
+  const sharedName = left.some((token) => right.includes(token));
+  if (!sharedName) return 0;
+  const params = paramNames(x);
+  const otherParams = paramNames(y);
+  if (params.size === 0 && otherParams.size === 0) return 0;
+  const paramOverlap = jaccard(params, otherParams);
+  if (paramOverlap < 0.34) return 0;
+  const requiredOverlap = jaccard(
+    new Set(x.parameters?.required ?? []),
+    new Set(y.parameters?.required ?? [])
+  );
+  return 0.5 + 0.5 * (0.6 * paramOverlap + 0.4 * requiredOverlap);
+}
+
+function nameTokens(name) {
+  return String(name ?? "")
+    .replace(/([a-z0-9])([A-Z])/g, "$1_$2")
+    .toLowerCase()
+    .split(/[^a-z0-9]+/)
+    .filter(Boolean);
+}
+
+function paramNames(tool) {
+  return new Set(Object.keys(tool.parameters?.properties ?? {}));
+}
+
+function jaccard(a, b) {
+  if (a.size === 0 && b.size === 0) return 1;
+  const union = new Set([...a, ...b]);
+  let shared = 0;
+  for (const item of union) if (a.has(item) && b.has(item)) shared += 1;
+  return shared / union.size;
+}
+
 function tokens(text) {
   const set = new Set();
   const lowered = String(text ?? "").toLowerCase();

package/lib/lint.js

@@ -178,7 +178,8 @@ export function lintSuite(suite) {
         pair: [pair.a, pair.b]
       });
     } else if (pair.similarity >= NO_BOUNDARY_SIMILARITY) {
-      add("warn", "pair-no-boundary", pair.a, `与 ${pair.b} 的描述相似度 ${similarity}，且两边的 description 都没写“不适用于”边界，模型只能靠相似度硬猜`, {
+      const reason = pair.basis === "structural" ? "名字与参数形状接近" : "描述相近";
+      add("warn", "pair-no-boundary", pair.a, `与 ${pair.b} ${reason}（相似度 ${similarity}），且两边都没写“不适用于”边界，模型只能靠相似度硬猜`, {
         pair: [pair.a, pair.b],
         suggestion: `在 ${pair.a} 和 ${pair.b} 的 description 里各写一句“不适用于什么”，并补一条歧义用例`
       });

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "modelfunction-cli",
-  "version": "0.1.2",
+  "version": "0.1.3",
   "type": "module",
   "license": "MIT",
   "description": "mf — function-calling accuracy scorecard for AI agents",