Playbook: Standing Up a New SIE Instance

This playbook covers the end-to-end process of deploying a new Strategic Intelligence Engine instance on a WordPress site. Follow the phases in order. Each phase includes verification criteria — do not proceed to the next phase until verification passes.

Prerequisites

Before beginning deployment, confirm the following are in place:

• WordPress 6.4+ installed and accessible via SFTP or SSH
• PHP 8.2+ running on the server
• WordPress REST API enabled and accessible
• A domain with SSL configured
• API keys obtained for: OpenAI or Anthropic (LLM provider), Pinecone (vector database)
• Git repository initialized for the site’s Knowledge Core content
• Obsidian vault configured with the KB directory structure

Phase 1: Plugin Installation

Step 1.1 — Install the mu-plugin. Upload sie-cpt.php to wp-content/mu-plugins/. This registers the four custom post types (knowledge_base, sie_faq, sie_insight, sie_guide) and their taxonomies (knowledge_topic, sie_topic). The mu-plugin ensures content types persist independently of the main SIE plugin.

Step 1.2 — Install the SIE plugin. Upload the sie-wp-plugin directory to wp-content/plugins/ and activate it from the WordPress admin.

Step 1.3 — Configure wp-config.php constants. Add API keys and sensitive credentials as PHP constants in wp-config.php rather than storing them in the database:

define( 'SIE_OPENAI_API_KEY',    'sk-...' );
define( 'SIE_ANTHROPIC_API_KEY', 'sk-ant-...' );
define( 'SIE_PINECONE_API_KEY',  'pcsk_...' );
define( 'SIE_PINECONE_HOST',     'https://your-index.svc.pinecone.io' );
define( 'SIE_PINECONE_INDEX',    'your-index-name' );

Verification: Navigate to the SIE settings page in WordPress admin. Confirm all API keys show as “Locked (defined in wp-config.php).” Confirm the four custom post types appear in the admin sidebar. Confirm the Knowledge Topics and SIE Topics taxonomy pages are accessible.

Phase 2: Knowledge Core Initialization

Step 2.1 — Create the taxonomy structure. In WordPress admin, create the top-level knowledge_topic terms that map to your KB directory structure. For each term, populate the _sie_path_pattern field with the corresponding folder path (e.g., /AI/, /SEO/, /CORE/). Create child terms for subcategories.

Step 2.2 — Configure site labels and slugs. In SIE settings, configure the public-facing labels and URL slugs for the knowledge base and triad CPTs (FAQ, Insight, Guide). Set the KB archive slug (default: kb).

Step 2.3 — Flush permalinks. Navigate to Settings > Permalinks and click Save. This registers the new rewrite rules for the custom post types and taxonomies.

Verification: Visit yoursite.com/kb/ and confirm the archive page loads (it will be empty). Visit a knowledge topic URL and confirm it resolves. Check that the REST API returns valid responses at /wp-json/sie/v1/topics.

Phase 3: Pinecone Index Setup

Step 3.1 — Create the Pinecone index. If not already created, create a new Pinecone index with the appropriate dimensions for your embedding model (e.g., 1536 for OpenAI text-embedding-3-small, 3072 for text-embedding-3-large). Use cosine similarity as the distance metric.

Step 3.2 — Verify connectivity. From the SIE settings page or via the Site Health panel, run the Pinecone connection test. Confirm the plugin can reach the index and reports the correct dimension count and metric.

Verification: The SIE Site Health panel shows a green status for Pinecone connectivity. The index is empty (vector count = 0), which is expected before the first sync.

Phase 4: Knowledge Pipeline Configuration

Step 4.1 — Configure the sync mapping. The kb_sync_mapping.json file maps Obsidian file paths to WordPress taxonomy term IDs. Generate this mapping by running the topic discovery endpoint (/wp-json/sie/v1/topics) and matching your KB folder paths to the term IDs created in Phase 2.

Step 4.2 — Run the initial sync. Execute the Knowledge Pipeline sync for the first time. This will process all markdown files in the KB directory, generate embeddings, index them in Pinecone, and create corresponding WordPress posts.

Step 4.3 — Monitor the sync. Watch the sync logs for errors: failed frontmatter validation, unmapped taxonomy paths, embedding failures, and WordPress API errors. Address any failures before proceeding.

Verification: Check the Pinecone index — vector count should match the number of synced content chunks. Check WordPress — knowledge base posts should exist with correct taxonomy assignments. Spot-check several posts to confirm content and metadata integrity.

Phase 5: Agent Fleet Bootstrap

Step 5.1 — Configure the base agent. Set up the first agent (typically the Editor Agent) with its role definition, tool access (WordPress REST API wrapper), and Knowledge Core access boundaries. Define the agent’s system prompt with references to the relevant governance protocols.

Step 5.2 — Test with a controlled task. Assign the agent a simple, verifiable task: draft a knowledge base article on a topic where the correct answer is already in the Knowledge Core. Review the output’s Verification Ledger to confirm the Iron Word protocol is functioning.

Step 5.3 — Deploy additional agents. Once the base agent is verified, deploy additional agents following the Agent Onboarding playbook. Each agent should be tested individually before being added to the active fleet.

Verification: The test agent produces a draft with a complete Verification Ledger. The confidence score, reasoning, and source citations are present and accurate. The draft appears in WordPress with correct custom meta fields.

Post-Deployment Checklist

After all phases are complete, confirm the following:

• [ ] mu-plugin and SIE plugin installed and active
• [ ] API keys configured via wp-config.php constants
• [ ] Taxonomy structure created with path pattern mappings
• [ ] Permalinks flushed and KB URLs resolving
• [ ] Pinecone index populated and connected
• [ ] Knowledge Pipeline sync completing without errors
• [ ] WordPress posts created with correct taxonomy and metadata
• [ ] At least one agent tested and producing valid Verification Ledgers
• [ ] Site Health panel showing all green statuses
• [ ] Audit logging active and capturing agent actions