prism-mcp-server 19.3.0 → 19.3.2

package/README.md

@@ -85,7 +85,9 @@ Every `prism_infer` call tracks which model handled it (local Ollama vs cloud) a
 ```
 📊 Inference Metrics (this session):
   Total calls: 12 — Local: 10 (83%) | Cloud: 2 (17%)
-  Tokens: 8,420 in + 3,150 out = 11,570 total
+  Prompt tokens: 7,840 evaluated / 8,420 submitted est.
+  Completion tokens: 3,150
+  Cloud tokens saved (est.): 11,570 — token volume handled locally instead of cloud
   Avg latency: 1,240ms
   By model:
     prism-coder:27b: 6 calls, 7,200 tokens, avg 1,800ms
@@ -93,6 +95,8 @@ Every `prism_infer` call tracks which model handled it (local Ollama vs cloud) a
     synalux-27b: 2 calls, 1,500 tokens, avg 1,100ms
 ```
 
+**Cloud tokens saved** is the honest routing metric — it accrues only when local Ollama handles a call that would otherwise have gone to Claude or the Synalux portal. A compact version appears inline after every 5th `prism_infer` call: `📊 local 10 (83%) · cloud 2 (17%) · ~11,570 tok · avg 1,240ms · 11,570 cloud tok saved`.
+
 Local calls use actual Ollama token counts (`prompt_eval_count` / `eval_count` from Ollama); cloud calls use char/4 estimates. Metrics are tracked locally — no portal dependency, no env vars, works offline. Per-call data is also forwarded to the Synalux portal as best-effort analytics (independent of the display).
 
 ### Session Drift Detection

package/dist/storage/sqlite.js

@@ -741,6 +741,7 @@ export class SqliteStorage {
         "archived_at", "deleted_at", "is_rollup", "role", "event_type",
         "created_at", "updated_at", "session_date", "importance", "title",
         "agent_name", "last_accessed_at", "confidence_score", "rollup_count",
+        "embedding",
     ]);
     parsePostgRESTFilters(params) {
         const conditions = [];

package/dist/tools/prismInferHandler.js

@@ -11,7 +11,7 @@
  *   3. Call /api/generate locally — return on success
  *   4. On local fail, if cloud_fallback=true:
  *        - exchange synalux_sk_ → JWT (cached)
- *        - POST synalux portal /api/v1/prism-aac/inference
+ *        - POST synalux portal /api/v1/prism/inference
  *        - portal runs its own cascade (9B/27B/Claude by tier)
  *   5. Return { output, backend, model_picked, ram_free_mb, latency_ms, used_cloud }
  *
@@ -260,7 +260,7 @@ async function callSynaluxInference(prompt, maxTokens, timeoutMs) {
     const jwt = await getSynaluxJwt();
     if (!jwt)
         return { ok: false, reason: "jwt_exchange_failed" };
-    const url = `${PRISM_SYNALUX_BASE_URL}/api/v1/prism-aac/inference`;
+    const url = `${PRISM_SYNALUX_BASE_URL}/api/v1/prism/inference`;
     try {
         let res = await fetch(url, {
             method: "POST",

package/dist/utils/inferenceMetrics.js

@@ -37,6 +37,7 @@ let promptTokensEvaluated = 0;
 let promptTokensSubmittedEst = 0;
 let totalCompletionTokens = 0;
 let totalLatencyMs = 0;
+let cloudTokensSavedEst = 0;
 export function recordInference(result) {
     if (result.backend === "safety_gate")
         return;
@@ -64,6 +65,9 @@ export function recordInference(result) {
     promptTokensSubmittedEst += submittedEst;
     totalCompletionTokens += ct;
     totalLatencyMs += result.latency_ms;
+    if (!result.used_cloud) {
+        cloudTokensSavedEst += submittedEst + ct;
+    }
     if (!byModel[key]) {
         byModel[key] = {
             calls: 0,
@@ -96,6 +100,7 @@ export function getInferenceSnapshot() {
         totalCompletionTokens,
         totalTokens: promptTokensSubmittedEst + totalCompletionTokens,
         avgLatencyMs: total > 0 ? Math.round(totalLatencyMs / total) : 0,
+        cloudTokensSavedEst,
         byModel: modelCopy,
     };
 }
@@ -106,6 +111,7 @@ export function resetInferenceMetrics() {
     promptTokensSubmittedEst = 0;
     totalCompletionTokens = 0;
     totalLatencyMs = 0;
+    cloudTokensSavedEst = 0;
     for (const key of Object.keys(byModel)) {
         delete byModel[key];
     }
@@ -142,7 +148,8 @@ export function formatInferenceMetrics(compact = false) {
         // see at least one rollup. Otherwise emit every N calls as the rolling summary.
         if (snap.totalCalls !== 1 && snap.totalCalls % every !== 0)
             return "";
-        return `📊 local ${snap.localCalls} (${snap.localPct}%) · cloud ${snap.cloudCalls} (${snap.cloudPct}%) · ~${snap.totalTokens.toLocaleString()} tok · avg ${snap.avgLatencyMs}ms`;
+        const savedStr = snap.cloudTokensSavedEst > 0 ? ` · ${snap.cloudTokensSavedEst.toLocaleString()} cloud tok saved` : "";
+        return `📊 local ${snap.localCalls} (${snap.localPct}%) · cloud ${snap.cloudCalls} (${snap.cloudPct}%) · ~${snap.totalTokens.toLocaleString()} tok · avg ${snap.avgLatencyMs}ms${savedStr}`;
     }
     // Full multi-line block (explicit inference_metrics tool call).
     // T2: show both evaluated (Ollama actual) and submitted estimate.
@@ -150,11 +157,15 @@ export function formatInferenceMetrics(compact = false) {
     const promptLine = snap.promptTokensEvaluated !== snap.promptTokensSubmittedEst
         ? `  Prompt tokens: ${snap.promptTokensEvaluated.toLocaleString()} evaluated / ${snap.promptTokensSubmittedEst.toLocaleString()} submitted est.`
         : `  Prompt tokens: ${snap.promptTokensEvaluated.toLocaleString()}`;
+    const savedLine = snap.cloudTokensSavedEst > 0
+        ? `  Cloud tokens saved (est.): ${snap.cloudTokensSavedEst.toLocaleString()} — token volume handled locally instead of cloud`
+        : `  Cloud tokens saved (est.): 0`;
     const lines = [
         `\n📊 Delegation Metrics — local-model calls this session (not host model spend):`,
         `  Total calls: ${snap.totalCalls} — Local: ${snap.localCalls} (${snap.localPct}%) | Cloud: ${snap.cloudCalls} (${snap.cloudPct}%)`,
         promptLine,
         `  Completion tokens: ${snap.totalCompletionTokens.toLocaleString()}`,
+        savedLine,
         `  Avg latency: ${snap.avgLatencyMs}ms`,
     ];
     const models = Object.entries(snap.byModel).sort((a, b) => b[1].calls - a[1].calls);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "prism-mcp-server",
-  "version": "19.3.0",
+  "version": "19.3.2",
   "mcpName": "io.github.dcostenco/prism-coder",
   "description": "Prism Coder — Cognitive memory + tool-calling intelligence for AI agents. Mind Palace persistent memory (BFCL Gold Certified, 100% Tool-Call Accuracy, 114 Agent Skills, PHI Guard, Tier Enforcement, Prompt-Based Skill Routing, Zero-Search HDC/HRR retrieval, HRR Semantic Drift Detection across BCBA/Coding/AAC domains, HIPAA-hardened local-first storage, SLERP-optimized GRPO alignment) plus the prism-coder 1.7B–32B open-weights LLM fleet.",
   "module": "index.ts",