auto-model-router 0.4.6 → 0.4.7

package/.omp-plugin/marketplace.json

@@ -7,14 +7,14 @@
   },
   "metadata": {
     "description": "auto-model-router: a local cost/complexity-aware model router for Oh My Pi, backed by OpenRouter",
-    "version": "0.4.6",
+    "version": "0.4.7",
     "pluginRoot": "."
   },
   "plugins": [
     {
       "name": "auto-model-router",
       "description": "Local cost/complexity-aware model router for Oh My Pi, backed by OpenRouter. Runs in-process, routes per turn by price and task complexity, with budget caps, mid-stream escalation, and cache-aware hysteresis.",
-      "version": "0.4.6",
+      "version": "0.4.7",
       "author": {
         "name": "drewappling",
         "email": "drewappling@gmail.com"

package/hermes-plugin/native/__init__.py

@@ -273,9 +273,11 @@ def _why_text(e: Dict[str, Any]) -> str:
     pt, ct = usage.get("promptTokens", 0) or 0, usage.get("cachedTokens", 0) or 0
     cache = f"{round(100 * ct / pt)}%" if pt else "n/a"
     cost = e.get("reportedUsd")
+    if cost is None:
+        cost = e.get("predictedUsd") or 0
     lines = [
         f"last turn: {e.get('servedSlug') or e.get('slug')} [{e.get('tier')}] · {e.get('classificationSource')} (confidence {e.get('confidence')})",
-        f"cost ${cost if cost is not None else e.get('predictedUsd')} · cache hit {cache} · latency {e.get('latencyMs')}ms",
+        f"cost ${float(cost):.5f} · cache hit {cache} · latency {e.get('latencyMs')}ms",
     ]
     for r in e.get("reasons") or []:
         lines.append(f"  - {r}")

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "auto-model-router",
-  "version": "0.4.6",
+  "version": "0.4.7",
   "private": false,
   "description": "Local cost/complexity-aware model router for Oh My Pi, backed by OpenRouter",
   "type": "module",

package/src/server/digest.ts

@@ -121,6 +121,14 @@ interface RecentDigest {
 	atMs: number;
 	ledgerId: string;
 	rerun: boolean;
+	/**
+	 * Whether the call that PRODUCED this digest has been seen. A tool_result
+	 * digest is made before the next request, and that request's last
+	 * assistant message carries the producing call; it must not count as a
+	 * re-run. A compaction digest covers a call already in history, so its
+	 * origin counts as seen from the start.
+	 */
+	originSeen: boolean;
 }
 const RERUN_WINDOW_MS = 2 * 3_600_000;
 const RECENT_PER_SESSION = 50;
@@ -261,7 +269,7 @@ export function createDigester(deps: DigesterDeps): Digester {
 			if (text === "" || text.length >= inputBytes * 0.9) return { digested: false, reason: "digest did not shrink the output" };
 			if (req.ompSessionId !== "") {
 				const list = recent.get(req.ompSessionId) ?? [];
-				list.push({ tool: req.toolName.toLowerCase(), arg: primaryArg(JSON.stringify(req.input)), atMs: startedAt, ledgerId: entry.id, rerun: false });
+				list.push({ tool: req.toolName.toLowerCase(), arg: primaryArg(JSON.stringify(req.input)), atMs: startedAt, ledgerId: entry.id, rerun: false, originSeen: source === "compaction" });
 				recent.set(req.ompSessionId, list.slice(-RECENT_PER_SESSION));
 			}
 			return {
@@ -284,6 +292,11 @@ export function createDigester(deps: DigesterDeps): Digester {
 				if (arg === null) continue;
 				for (const d of list) {
 					if (d.rerun || d.tool !== tool || d.arg !== arg || nowMs - d.atMs > RERUN_WINDOW_MS) continue;
+					if (!d.originSeen) {
+						// The producing call, arriving in the next request's history.
+						d.originSeen = true;
+						continue;
+					}
 					d.rerun = true;
 					marked++;
 					try {

package/test/digest.test.ts

@@ -198,6 +198,9 @@ describe("createDigester", () => {
 		expect(dg.noteToolCalls("omp-1", [{ name: "read", argsJson: '{"path":"src/b.ts"}' }, { name: "grep", argsJson: '{"pattern":"src/a.ts"}' }])).toBe(0);
 		expect(dg.noteToolCalls("omp-2", [{ name: "read", argsJson: '{"path":"src/a.ts"}' }])).toBe(0);
 		expect(row().wasted).toBe(false);
+		// The next request carries the call that PRODUCED the digest in its last assistant message: not a re-run.
+		expect(dg.noteToolCalls("omp-1", [{ name: "read", argsJson: '{"path":"src/a.ts","offset":1}' }])).toBe(0);
+		expect(row().wasted).toBe(false);
 		// The same read again (case-insensitive tool name, any other args): the agent wanted the full output.
 		expect(dg.noteToolCalls("omp-1", [{ name: "Read", argsJson: '{"path":"src/a.ts","limit":50}' }])).toBe(1);
 		expect(row().wasted).toBe(true);