brain-mcp

Instructions

    System-wide stats proving the prosthetic works.
    Shows everything that's preserved: conversations, domains, questions, decisions.
    The 'everything is okay' view.
    

Implementation Reference

• brain_mcp/server/tools_prosthetic.py:1084-1211 (handler)

  The implementation of the 'trust_dashboard' tool within the Cognitive Prosthetic toolset. It queries the summaries database to provide a status overview, or falls back to raw conversation data if the summaries database is unavailable.

  def trust_dashboard() -> str:
      """
      System-wide stats proving the prosthetic works.
      Shows everything that's preserved: conversations, domains, questions, decisions.
      The 'everything is okay' view.
      """
      db = get_summaries_db()
      if db:
          stats = db.execute("""
              SELECT COUNT(*) as total,
                     COUNT(DISTINCT domain_primary) as domains,
                     COUNT(CASE WHEN importance = 'breakthrough' THEN 1 END) as breakthroughs
              FROM summaries
          """).fetchone()
          total, domains, breakthroughs = stats
  
          rows = db.execute("SELECT open_questions, decisions FROM summaries").fetchall()
          total_oq, total_dec = 0, 0
          for oq_str, dec_str in rows:
              oqs = parse_json_field(oq_str)
              total_oq += sum(1 for q in oqs if q and q.lower() != "none identified")
              total_dec += len(parse_json_field(dec_str))
  
          sources = db.execute("""
              SELECT source, COUNT(*) as count FROM summaries GROUP BY source ORDER BY count DESC
          """).fetchall()
  
          domain_rows = db.execute("""
              SELECT domain_primary, COUNT(*) as count,
                     MAX(thinking_stage) as stage,
                     COUNT(CASE WHEN importance='breakthrough' THEN 1 END) as bt
              FROM summaries
              WHERE domain_primary != '' AND domain_primary IS NOT NULL
              GROUP BY domain_primary ORDER BY count DESC
          """).fetchall()
  
          oq_domains = set()
          for row in db.execute("SELECT domain_primary, open_questions FROM summaries").fetchall():
              for q in parse_json_field(row[1]):
                  if q and q.lower() != "none identified":
                      oq_domains.add(row[0])
                      break
  
          output = [f"## 🛡️ Trust Dashboard\n"]
          output.append(f"_Your cognitive safety net — proof that nothing is lost_\n")
          output.append(f"### 📊 Global Metrics")
          output.append(f"  **Conversations indexed:** {total:,}")
          output.append(f"  **Domains tracked:** {domains}")
          output.append(f"  **Open questions preserved:** {total_oq:,}")
          output.append(f"  **Decisions preserved:** {total_dec:,}")
          output.append(f"  **Breakthroughs captured:** {breakthroughs} 💎")
          output.append(f"  **Domains with active threads:** {len(oq_domains)}")
  
          output.append(f"\n### 📡 Sources")
          for src, count in sources:
              output.append(f"  {src}: {count:,}")
  
          output.append(f"\n### 🗺️ Domain Coverage (top 15)")
          for domain, count, stage, bt in domain_rows[:15]:
              bt_mark = f" 💎×{bt}" if bt else ""
              has_oq = " 🔴" if domain in oq_domains else " ✅"
              output.append(f"  {domain}: {count} convos ({stage or '?'}){bt_mark}{has_oq}")
  
          output.append(f"\n### 🔑 Safety Net Status")
          output.append(f"  {'🟢' if total > 5000 else '🟡'} Coverage: {total:,} conversations")
          output.append(f"  {'🟢' if domains > 15 else '🟡'} Breadth: {domains} domains")
          output.append(f"  {'🟢' if breakthroughs > 10 else '🟡'} Depth: {breakthroughs} breakthroughs captured")
          output.append(f"  {'🔴' if len(oq_domains) > 10 else '🟢'} Open threads: {len(oq_domains)} domains need attention")
  
          return "\n".join(output)
  
      # ── Fallback: stats from raw conversations + vectors ──
      try:
          con = get_conversations()
      except FileNotFoundError:
          return "No conversation data found. Run the ingest pipeline first."
  
      stats = con.execute("""
          SELECT COUNT(*) as total,
                 COUNT(DISTINCT conversation_title) as convos,
                 COUNT(DISTINCT source) as sources,
                 MIN(created) as first_msg,
                 MAX(created) as last_msg,
                 SUM(CASE WHEN has_question = 1 AND role = 'user' THEN 1 ELSE 0 END) as questions
          FROM conversations
      """).fetchone()
      total, convos, n_sources, first_msg, last_msg, total_qs = stats
  
      source_rows = con.execute("""
          SELECT source, COUNT(*) as count
          FROM conversations
          GROUP BY source
          ORDER BY count DESC
      """).fetchall()
  
      vec_count = lance_count()
  
      output = [f"## 🛡️ Trust Dashboard\n"]
      output.append(f"_Your cognitive safety net — what's preserved_\n")
  
      output.append(f"### 📊 Data Inventory")
      output.append(f"  **Total messages:** {total:,}")
      output.append(f"  **Conversations:** {convos:,}")
      output.append(f"  **Questions captured:** {total_qs:,}")
      output.append(f"  **Vectors indexed:** {vec_count:,}")
      output.append(f"  **Date range:** {str(first_msg)[:10]} to {str(last_msg)[:10]}")
  
      if source_rows:
          output.append(f"\n### 📡 Sources")
          for src, count in source_rows:
              output.append(f"  {src}: {count:,}")
  
      output.append(f"\n### 🔑 Pipeline Status")
      output.append(f"  🟢 Conversations: {total:,} messages ingested")
      output.append(f"  {'🟢' if vec_count > 0 else '🔴'} Vectors: {vec_count:,} embeddings indexed")
      output.append(f"  🔴 Summaries: not generated")
      output.append(f"\n### What's available without summaries")
      output.append(f"  - Keyword search across all conversations")
      output.append(f"  - Semantic (vector) search")
      output.append(f"  - Basic conversation stats and history")
      output.append(f"  - Approximate open threads and switching costs")
      output.append(f"\n### What summaries add")
      output.append(f"  - Thinking stages, cognitive patterns, emotional tones")
      output.append(f"  - Structured open questions and decisions")
      output.append(f"  - Domain classification and breakthrough detection")
      output.append(f"  - Key insights and quotable moments")
      output.append(_SUMMARIES_HINT)
      return "\n".join(output)

• brain_mcp/dashboard/routes/tools.py:85-91 (registration)

  Registration entry for 'trust_dashboard' in the dashboard tools registry, including metadata like description, category, and requirements.

      "name": "trust_dashboard",
      "description": "System-wide proof the safety net works",
      "category": "Cognitive Prosthetic",
      "requires": ["conversations", "summaries"],
      "params": {},
      "probe": {},
  },