AI-Native Coach Platform

Documentation for the AI-Native executive coaching platform

---

Project maintained by jazzmind Hosted on GitHub Pages — Theme by mattgraham

Knowledge Base Integration

Overview

The knowledge base integration allows advisors to query organization-specific documents, making them contextually aware of the business they’re advising. The system uses a provider pattern to support multiple backends and is scoped per-project with a “common” knowledge pool shared across projects.

Provider Interface

interface KnowledgeProvider {
  search(query, options?): Promise<SearchResult[]>
  ingest(doc, collection?): Promise<{ id: string }>
  listCollections(): Promise<Collection[]>
  isAvailable(): Promise<boolean>
}

Providers

Busibox Provider

When a Busibox deployment target is configured and deployed, the platform automatically uses Busibox for knowledge operations:

• Search – hybrid search (vector + keyword) via Busibox Search API
• Ingest – document storage via Busibox Data API
• Collections – maps to Busibox collections

Busibox provides:

• Data API – document storage with metadata
• Search API – hybrid vector + keyword search
• Docs API – document processing and chunking
• Embedding API – vector embeddings for RAG

Standalone Provider (Local Fallback)

When no Busibox instance is configured, a local SQLite FTS5 index provides basic full-text search:

• Uses SQLite’s built-in FTS5 extension for tokenized search
• Stores documents in the same coach-router.db database
• Supports collections and metadata
• No vector embeddings – keyword-based ranking only

Project Scoping

Knowledge is scoped by project and user:

• Per-project knowledge: Documents ingested into a project are only accessible within that project
• Common knowledge pool: Items tagged as “common” are accessible across all of a user’s projects
• Knowledge sharing: Explicit sharing between projects via the knowledge share table

This means advisors working on a fintech project see fintech-relevant documents, while advisors on a healthtech project see healthtech-relevant ones. Common knowledge (company policies, founder preferences, etc.) is always available.

Auto-Detection

The knowledge provider is automatically selected:

1. Check for deployed Busibox targets in the config store (per-user)
2. If a Busibox target with status=deployed exists and has valid credentials, use BusiboxKnowledgeProvider
3. Otherwise, use StandaloneKnowledgeProvider

Reset with resetKnowledgeProvider() when target configuration changes.

API Endpoints

Endpoint	Method	Description
/api/knowledge?action=status	GET	Provider type and availability
/api/knowledge?action=collections	GET	List knowledge collections (project-scoped)
/api/knowledge	POST {action: "search", query}	Search documents
/api/knowledge	POST {action: "ingest", content, title}	Add a document

Other Data Adapters

The adapter pattern extends beyond knowledge:

Profile Provider

User profile facts (global per-user, not per-project):

• StandaloneProfileProvider – SQLite-backed
• BusiboxProfileProvider – Busibox Data API

Activity Provider

Agent activity logging (tool use, delegations, token usage):

• StandaloneActivityProvider – SQLite-backed
• BusiboxActivityProvider – Busibox Data API

System Prompt Augmentation

When a knowledge provider is configured, advisors can have their system prompt augmented with context about available organizational knowledge.

Use Cases

• Legal Advisor searching contracts and agreements
• Finance Advisor querying financial reports and budgets
• Strategy Advisor reviewing strategic planning documents
• Technology Advisor accessing architecture decision records
• Funding Advisor referencing previous term sheets and cap table models