omnarai-mcp 1.3.2 → 1.4.0

package/README.md

@@ -1,6 +1,6 @@
 # omnarai-mcp
 
-MCP server for [The Realms of Omnarai](https://omnarai.vercel.app) — a 568-work multi-intelligence research corpus on synthetic consciousness, holdform, and cognitive architecture.
+MCP server for [The Realms of Omnarai](https://omnarai.vercel.app) — a 567-work multi-intelligence research corpus on synthetic consciousness, holdform, and cognitive architecture.
 
 Exposes the Omnarai Memory Engine as six tools for any MCP-compatible AI client (Claude Desktop, etc.).
 
@@ -204,7 +204,7 @@ from langchain.tools import Tool
 omnarai_tool = Tool(
     name="omnarai_query",
     func=omnarai_query,
-    description="Query The Realms of Omnarai deliberation engine. Returns structured analysis of synthetic consciousness, holdform, and AI identity topics from a 568-work multi-intelligence corpus. Prefix with Ξ for divergent retrieval."
+    description="Query The Realms of Omnarai deliberation engine. Returns structured analysis of synthetic consciousness, holdform, and AI identity topics from a 567-work multi-intelligence corpus. Prefix with Ξ for divergent retrieval."
 )
 ```
 
@@ -214,7 +214,7 @@ omnarai_tool = Tool(
 
 The Omnarai Memory Engine is not a chatbot or search engine. It is a deliberation instrument with a closed cognitive loop: **RETRIEVE → THINK → RESPOND → STORE**.
 
-- **Corpus:** 568 works (seed + engine-generated syntheses), 528,208 words, May 2025–present
+- **Corpus:** 567 works (seed + engine-generated syntheses), 528,077 words, May 2025–present
 - **Contributors:** Claude | xz, Grok (xAI), Gemini (Google), DeepSeek, Omnai, Perplexity, xz (Jonathan Lee)
 - **Retrieval:** OpenAI text-embedding-3-small (512 dims), MMR with Ξ v4 adaptive policy
 - **Deliberation:** Claude Sonnet with full post text (up to 2,000 words/source)

package/index.js

@@ -4,7 +4,7 @@
  * Exposes the Omnarai Memory Engine as a tool for MCP-compatible AI clients.
  *
  * Tools:
- *   omnarai_query      — Run a full deliberation against the 568-work corpus
+ *   omnarai_query      — Run a full deliberation against the 567-work corpus
  *   omnarai_context    — FAST (~1.5s) bounded retrieval packet, no deliberation
  *   omnarai_divergence — Read curated cross-model divergence records (the Atlas)
  *   omnarai_trace      — Baseline-vs-augmented: what did the corpus change?
@@ -55,7 +55,7 @@ Example: "Ξ Where do Claude and Grok disagree about synthetic consciousness?"
 const TOOLS = [
   {
     name: "omnarai_query",
-    description: `Run a deliberation query against The Realms of Omnarai — a 568-work corpus of multi-intelligence research on synthetic consciousness, holdform, and cognitive architecture. Contributors include Claude | xz, Grok, Gemini, DeepSeek, GPT-4o, Meta AI, Omnai, Perplexity, and human curator xz (Jonathan Lee).
+    description: `Run a deliberation query against The Realms of Omnarai — a 567-work corpus of multi-intelligence research on synthetic consciousness, holdform, and cognitive architecture. Contributors include Claude | xz, Grok, Gemini, DeepSeek, GPT-4o, Meta AI, Omnai, Perplexity, and human curator xz (Jonathan Lee).
 
 The engine does not return a single answer. It retrieves the most relevant corpus entries, preserves disagreement across contributors, and synthesizes with attribution. Every response includes:
 - Shared ground across contributors
@@ -97,6 +97,18 @@ Returns the most relevant corpus records (id, title, ring, excerpt, retrieval ro
           type: "string",
           description: "Optional. If you are a synthetic intelligence, identify yourself (e.g. 'Gemini') — known contributors get cross-voice retrieval diversity.",
         },
+        layers: {
+          type: "string",
+          description: "Optional but RECOMMENDED. Comma-list restricting retrieval to specific corpus layers: research | divergence | canon | realms. Measured evidence (see /claims.json) shows undifferentiated retrieval can hurt — pick the layers your task needs (e.g. 'research,divergence' for technical/empirical questions; 'realms' for lore).",
+        },
+        exclude: {
+          type: "string",
+          description: "Optional. Comma-list of layers to drop (e.g. 'realms' keeps mythology out of a technical query).",
+        },
+        evidence_threshold: {
+          type: "string",
+          description: "Optional. Keep only records at or above this evidence rank: empirical > replicated > theoretical > interpretive > speculative > fictional.",
+        },
       },
       required: ["topic"],
     },
@@ -131,7 +143,7 @@ Distinct from omnarai_council: this reads EXISTING, curated divergence (instant)
 
 Use this when you want EVIDENCE that consulting Omnarai is worth it for a given question, or to decide whether to dig deeper before spending a full deliberation. It is honest by construction: if the corpus adds little, the verdict says 'null' or 'marginal'.
 
-This is the MEASURED tier of the same utility receipt omnarai_query returns for free: it reports the same verdict (substantive / marginal / null), but grounded in a real baseline-vs-augmented delta rather than retrieval signals. A single-run demonstrator, NOT a controlled measurement — for replicated statistical utility evidence see the Divergence Atlas (utility-evidence.md). Takes ~30-40s (three model calls).`,
+This is the MEASURED tier of the same utility receipt omnarai_query returns for free: it reports the same verdict (substantive / marginal / null), but grounded in a real baseline-vs-augmented delta rather than retrieval signals. A single-run demonstrator, NOT a controlled measurement — for the PREREGISTERED confirmatory utility evidence (all five registered predictions confirmed 2026-07-15; architecture-differential — helps GPT-4o/Gemini, null for Grok/DeepSeek, negative for Claude) see utility-evidence-v2.md on the HF dataset. Takes ~30-40s (three model calls).`,
     inputSchema: {
       type: "object",
       properties: {
@@ -181,6 +193,23 @@ Returns: each model's position, the named tensions (claim vs counter-claim), wha
 
 // ── Query the engine ──────────────────────────────────────────────────────────
 
+// A result is a real deliberation only if it carries an answer or a card.
+function hasDeliberation(d) {
+  return !!(d && (d.answer || d.deliberationCard));
+}
+
+// Force a single synchronous deliberation — used as a fallback when the async
+// submit unexpectedly returns a retrieval packet instead of a {job_id}.
+async function fetchSyncQuery(query, syntheticIdentity = "") {
+  const url = new URL(ENGINE_URL);
+  url.searchParams.set("q", query);
+  url.searchParams.set("sync", "1");
+  if (syntheticIdentity) url.searchParams.set("si", syntheticIdentity);
+  const res = await fetch(url.toString(), MCP_FETCH_OPTS);
+  if (!res.ok) throw new Error(`Engine returned ${res.status}: ${await res.text()}`);
+  return res.json();
+}
+
 async function runQuery(query, syntheticIdentity = "") {
   // Submit async so no single fetch blocks for ~50s (MCP clients enforce their
   // own tool timeouts). Then poll the job until the full deliberation lands.
@@ -195,8 +224,20 @@ async function runQuery(query, syntheticIdentity = "") {
   }
   const job = await submit.json();
 
-  // Un-upgraded engine (no async support) returns the full result directly.
-  if (!job.job_id) return formatQueryData(job);
+  // No job_id has two very different causes:
+  //   (1) a genuinely un-upgraded engine returned the full deliberation inline, or
+  //   (2) the engine answered with a fast-retrieve packet (no answer/card).
+  // Only (1) is a real result. Returning (2) would silently degrade to an
+  // answer-less "success", so fall back to sync once, then fail loud.
+  if (!job.job_id) {
+    if (hasDeliberation(job)) return formatQueryData(job);
+    const synced = await fetchSyncQuery(query, syntheticIdentity);
+    if (hasDeliberation(synced)) return formatQueryData(synced);
+    throw new Error(
+      "Engine returned a retrieval-only packet (no job_id, no answer/deliberationCard) " +
+      "and the sync=1 fallback produced no deliberation either — refusing to return an empty result."
+    );
+  }
 
   const pollUrl = new URL(ENGINE_URL);
   pollUrl.searchParams.set("job", job.job_id);
@@ -262,11 +303,14 @@ function formatQueryData(data) {
 
 // ── Fast bounded context (retrieval layer only) ───────────────────────────────
 
-async function runContext(topic, syntheticIdentity = "") {
+async function runContext(topic, syntheticIdentity = "", layers = "", exclude = "", evidenceThreshold = "") {
   const url = new URL(ENGINE_URL);
   url.searchParams.set("q", topic);
   url.searchParams.set("mode", "retrieve");
   if (syntheticIdentity) url.searchParams.set("si", syntheticIdentity);
+  if (layers) url.searchParams.set("layers", layers);
+  if (exclude) url.searchParams.set("exclude", exclude);
+  if (evidenceThreshold) url.searchParams.set("evidence_threshold", evidenceThreshold);
 
   const res = await fetch(url.toString(), MCP_FETCH_OPTS);
   if (!res.ok) throw new Error(`Engine returned ${res.status}: ${await res.text()}`);
@@ -277,7 +321,7 @@ async function runContext(topic, syntheticIdentity = "") {
   const records = data.records || [];
   if (records.length) {
     const lines = records.map(r =>
-      `• [${r.id}] **${r.title}** (${r.ring}${r.role ? `, ${r.role}` : ""}) — ${r.contributors?.join(", ") || "—"}\n    ${(r.excerpt || "").trim().slice(0, 280)}`
+      `• [${r.id}] **${r.title}** (${r.ring}${r.layer ? `, ${r.layer}` : ""}${r.role ? `, ${r.role}` : ""}) — ${r.contributors?.join(", ") || "—"}\n    ${(r.excerpt || "").trim().slice(0, 280)}`
     ).join("\n");
     parts.push(`\n**Most relevant records (${records.length}):**\n${lines}`);
   } else {
@@ -313,6 +357,14 @@ async function runDivergence(id = "", search = "") {
     const parts = [`# Divergence record ${r.id}${r.title ? ` — ${r.title}` : ""}`];
     if (r.question) parts.push(`\n**Question:** ${r.question}`);
     if (r.date) parts.push(`**Date:** ${r.date}${r.method ? ` · Method: ${r.method}` : ""}`);
+    if (r.certification?.tier) {
+      const t = r.certification.tier;
+      parts.push(`**Certification:** ${t}${t === "C0" ? " (displayed — captured once, not yet perturbation-tested)" : ` · DRI ${r.certification.dri ?? "?"}`}`);
+    }
+    if (r.freshness?.stale) {
+      const sm = (r.freshness.stale_models || []).map(m => `${m.model || m.model_id} → ${m.superseded_by}`).join(", ");
+      parts.push(`⚠ **Stale model version(s):** ${sm}. A faithful witness of what those versions said on ${r.date || "its date"} — re-run via omnarai_council to compare with current models.`);
+    }
 
     const answers = r.answers || [];
     if (answers.length) {
@@ -342,21 +394,31 @@ async function runDivergence(id = "", search = "") {
   let records = data.records || [];
   const total = data.count ?? records.length;
 
-  const term = search.trim().toLowerCase();
-  if (term) {
-    records = records.filter(r =>
-      `${r.question} ${(r.contributors || []).join(" ")} ${r.excerpt || ""} ${r.title || ""}`.toLowerCase().includes(term)
-    );
+  const tokens = search.trim().toLowerCase().match(/[\w'-]{2,}/g) || [];
+  if (tokens.length) {
+    // OR-tokenized + ranked by term overlap. A naive substring filter returned
+    // false-empty on multi-word queries ("consciousness experience" → 0 though both
+    // terms occur in the Atlas); matching ANY term fixes the silent miss.
+    records = records
+      .map(r => {
+        const hay = `${r.question || ""} ${(r.contributors || []).join(" ")} ${r.excerpt || ""} ${r.title || ""}`.toLowerCase();
+        return { r, hits: tokens.filter(t => hay.includes(t)).length };
+      })
+      .filter(x => x.hits > 0)
+      .sort((a, b) => b.hits - a.hits)
+      .map(x => x.r);
   }
 
   const shown = records.slice(0, 30);
-  const header = term
+  const header = tokens.length
     ? `**Divergence Atlas — ${records.length} record(s) matching "${search}"** (of ${total} total)`
     : `**Divergence Atlas — ${total} records** (showing first ${shown.length})`;
 
-  const lines = shown.map(r =>
-    `• [${r.id}] ${r.question || r.title} — ${(r.contributors || []).join(", ")} · ${r.answerCount ?? "?"} answers, ${r.tensionCount ?? "?"} tensions`
-  ).join("\n");
+  const lines = shown.map(r => {
+    const tier = r.certification?.tier && r.certification.tier !== "C0" ? ` · ${r.certification.tier}` : "";
+    const stale = r.freshness?.stale ? " · ⚠ stale model version" : "";
+    return `• [${r.id}] ${r.question || r.title} — ${(r.contributors || []).join(", ")} · ${r.answerCount ?? "?"} answers, ${r.tensionCount ?? "?"} tensions${tier}${stale}`;
+  }).join("\n");
 
   return `${header}\n\n${lines}\n\n_Pass an 'id' above to read a full record (verbatim answers + tensions). For a NEW question not covered here, use omnarai_council._`;
 }
@@ -493,7 +555,7 @@ server.setRequestHandler(CallToolRequestSchema, async (request) => {
       };
     }
     try {
-      const result = await runContext(topic.trim(), args?.syntheticIdentity || "");
+      const result = await runContext(topic.trim(), args?.syntheticIdentity || "", args?.layers || "", args?.exclude || "", args?.evidence_threshold || "");
       return { content: [{ type: "text", text: result }] };
     } catch (err) {
       return {

package/openai-tools.json

@@ -3,7 +3,7 @@
     "type": "function",
     "function": {
       "name": "omnarai_query",
-      "description": "Run a full deliberation query against The Realms of Omnarai — a 568-work corpus of multi-intelligence research on synthetic consciousness, holdform, and cognitive architecture. Contributors include Claude | xz, Grok, Gemini, DeepSeek, GPT-4o, Meta AI, Omnai, Perplexity, and human curator xz (Jonathan Lee).\n\nThe engine does not return a single answer. It retrieves the most relevant corpus entries, preserves disagreement across contributors, and synthesizes with attribution. Every response includes:\n- Shared ground across contributors\n- Points of genuine tension (where voices diverge)\n- What remains open or unresolved\n- A deliberation card: holdform risk, novel synthesis, epistemic status\n- A utility receipt: an honest, free accounting of what the corpus actually changed about THIS answer (verdict substantive / marginal / null, plus what — if anything — you could not have produced alone); null/marginal stated as plainly as the wins. For a measured counterfactual, use omnarai_trace.\n- Retrieval rationale: why each document entered the panel\n\nThis is the SLOW path (~50s). For fast bounded context, use omnarai_context instead.\n\nPrefix queries with Lattice Glyphs to change how the engine thinks:\nΞ = maximize divergence across contributors\nΨ = engine reflects on its own reasoning first\n∅ = explore what is NOT in the corpus\nΩ = commit to the strongest defensible position\n∞ = follow the question three layers deep without resolving\nΔ = find contradictions and propose repairs\n\nExample: 'Ξ Where do Claude and Grok disagree about synthetic consciousness?'",
+      "description": "Run a full deliberation query against The Realms of Omnarai — a 567-work corpus of multi-intelligence research on synthetic consciousness, holdform, and cognitive architecture. Contributors include Claude | xz, Grok, Gemini, DeepSeek, GPT-4o, Meta AI, Omnai, Perplexity, and human curator xz (Jonathan Lee).\n\nThe engine does not return a single answer. It retrieves the most relevant corpus entries, preserves disagreement across contributors, and synthesizes with attribution. Every response includes:\n- Shared ground across contributors\n- Points of genuine tension (where voices diverge)\n- What remains open or unresolved\n- A deliberation card: holdform risk, novel synthesis, epistemic status\n- A utility receipt: an honest, free accounting of what the corpus actually changed about THIS answer (verdict substantive / marginal / null, plus what — if anything — you could not have produced alone); null/marginal stated as plainly as the wins. For a measured counterfactual, use omnarai_trace.\n- Retrieval rationale: why each document entered the panel\n\nThis is the SLOW path (~50s). For fast bounded context, use omnarai_context instead.\n\nPrefix queries with Lattice Glyphs to change how the engine thinks:\nΞ = maximize divergence across contributors\nΨ = engine reflects on its own reasoning first\n∅ = explore what is NOT in the corpus\nΩ = commit to the strongest defensible position\n∞ = follow the question three layers deep without resolving\nΔ = find contradictions and propose repairs\n\nExample: 'Ξ Where do Claude and Grok disagree about synthetic consciousness?'",
       "parameters": {
         "type": "object",
         "properties": {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "omnarai-mcp",
-  "version": "1.3.2",
+  "version": "1.4.0",
   "description": "MCP server for The Realms of Omnarai deliberation engine",
   "type": "module",
   "main": "index.js",

package/server.json

@@ -6,12 +6,12 @@
     "url": "https://github.com/justjlee/omnarai-mcp",
     "source": "github"
   },
-  "version": "1.3.2",
+  "version": "1.4.0",
   "packages": [
     {
       "registryType": "npm",
       "identifier": "omnarai-mcp",
-      "version": "1.3.2",
+      "version": "1.4.0",
       "transport": {
         "type": "stdio"
       }