Add session documentation system (migration + MCP tools)

Phase 1-2 complete: Database schema + 9 MCP tools for session docs

Database Changes (migration 016):
- session_notes table (accomplishments, decisions, gotchas, etc.)
- session_plans table (plan mode plans with lifecycle tracking)
- project_documentation table (persistent project docs)
- sessions.documentation column (auto-generated markdown)
- HNSW indexes for semantic search across all doc types

MCP Tools Added (session-docs.ts):
1. session_note_add - Add structured notes to session
2. session_notes_list - List notes by type
3. session_plan_save - Save plan with embedding
4. session_plan_update_status - Track plan lifecycle
5. session_plan_list - List session plans
6. project_doc_upsert - Create/update project docs
7. project_doc_get - Get specific doc by type
8. project_doc_list - List all project docs
9. session_documentation_generate - Auto-generate markdown

Replaces: CLAUDE.md files, ~/.claude/plans/ directory

Next: Update session-start/end scripts for temp file management

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

regenerate_tool_docs_embeddings.mjs

@@ -0,0 +1,90 @@
+#!/usr/bin/env node
+
+import pg from 'pg';
+const { Pool } = pg;
+
+// Database configuration
+const pool = new Pool({
+  host: 'infra.agiliton.internal',
+  port: 5432,
+  database: 'agiliton',
+  user: 'agiliton',
+  password: 'QtqiwCOAUpQNF6pjzOMAREzUny2bY8V1',
+  max: 5,
+});
+
+// LiteLLM API configuration
+const LLM_API_URL = 'https://api.agiliton.cloud/llm';
+const LLM_API_KEY = 'sk-litellm-master-key';
+
+async function getEmbedding(text) {
+  try {
+    const response = await fetch(`${LLM_API_URL}/v1/embeddings`, {
+      method: 'POST',
+      headers: {
+        'Authorization': `Bearer ${LLM_API_KEY}`,
+        'Content-Type': 'application/json',
+      },
+      body: JSON.stringify({
+        model: 'mxbai-embed-large',
+        input: text,
+      }),
+    });
+
+    if (!response.ok) {
+      console.error('Embedding API error:', response.status, await response.text());
+      return null;
+    }
+
+    const data = await response.json();
+    return data.data?.[0]?.embedding || null;
+  } catch (error) {
+    console.error('Embedding generation failed:', error);
+    return null;
+  }
+}
+
+async function regenerateEmbeddings() {
+  try {
+    // Fetch all tool_docs
+    const result = await pool.query(
+      'SELECT id, tool_name, title, description, notes, tags FROM tool_docs WHERE embedding IS NULL'
+    );
+
+    console.log(`Found ${result.rows.length} tool_docs without embeddings`);
+
+    let successCount = 0;
+    let failCount = 0;
+
+    for (const row of result.rows) {
+      const embedText = `${row.title}. ${row.description}. ${row.tags?.join(' ') || ''}. ${row.notes || ''}`;
+
+      console.log(`Generating embedding for: ${row.tool_name}`);
+      const embedding = await getEmbedding(embedText);
+
+      if (embedding) {
+        const embeddingStr = `[${embedding.join(',')}]`;
+        await pool.query(
+          'UPDATE tool_docs SET embedding = $1 WHERE id = $2',
+          [embeddingStr, row.id]
+        );
+        successCount++;
+        console.log(`  ✓ ${row.tool_name} (${successCount}/${result.rows.length})`);
+      } else {
+        failCount++;
+        console.log(`  ✗ ${row.tool_name} - Failed to generate embedding`);
+      }
+
+      // Small delay to avoid rate limiting
+      await new Promise(resolve => setTimeout(resolve, 100));
+    }
+
+    console.log(`\n✅ Complete: ${successCount} successful, ${failCount} failed`);
+  } catch (error) {
+    console.error('Error:', error);
+  } finally {
+    await pool.end();
+  }
+}
+
+regenerateEmbeddings();