Knowledge Pipeline

The Knowledge Pipeline is Whilst’s core feature that automatically extracts, analyzes, and organizes knowledge from Slack conversations.

How It Works

When someone mentions @whilst in Slack, the pipeline runs in the background:

Slack Mention (@whilst)
  ↓
Bot Responds (immediate)
  ↓
[Background] Knowledge Pipeline
  ├─ Thread Ingestion (fetch + store messages)
  ├─ AI Analysis (topics, decisions, action items)
  ├─ Auto-Doc Generation (structured documentation)
  └─ Relationship Extraction (collaboration patterns)

Pipeline Stages

1. Thread Ingestion

Captures the full Slack thread and stores it in the database.

Fetches all messages via the Slack API
Maps Slack user IDs to internal actors
Stores conversation metadata and individual messages

2. AI Analysis

Uses OpenAI to extract structured information from the thread:

interface ThreadAnalysis {
  topics: string[];
  keyPoints: string[];
  decisions: Array<{
    decision: string;
    reasoning: string;
    participants: string[];
  }>;
  actionItems: Array<{
    task: string;
    assignee?: string;
    dueDate?: string;
  }>;
  sentiment: 'positive' | 'neutral' | 'negative' | 'mixed';
  urgency: 'high' | 'medium' | 'low';
  workContext: string;
  summary: string;
}

3. Auto-Doc Generation

Generates a structured document from the analysis:

AI-generated title and content
Automatic tags from extracted topics
Links back to the source Slack thread
Visibility set based on urgency and content type

4. Relationship Extraction

Analyzes collaboration patterns between thread participants:

Shared topics and expertise areas
Communication style (formal, casual, technical)
Relationship type (peer, mentor, partner)
Stored in the relationship graph for future context

Code Location

Component	Package
Thread Ingestion	`packages/knowledge-pipeline/src/thread-ingestion.ts`
Thread Analyzer	`packages/knowledge-pipeline/src/thread-analyzer.ts`
Auto-Doc Generator	`packages/knowledge-pipeline/src/auto-doc-generator.ts`
Relationship Analyzer	`packages/knowledge-pipeline/src/relationship-analyzer.ts`
Context Retriever	`packages/knowledge-pipeline/src/context-retriever.ts`

Getting Started

Features

Guides

Development

Knowledge Pipeline

How It Works

Pipeline Stages

1. Thread Ingestion

2. AI Analysis

3. Auto-Doc Generation

4. Relationship Extraction

Code Location

​How It Works

​Pipeline Stages

​1. Thread Ingestion

​2. AI Analysis

​3. Auto-Doc Generation

​4. Relationship Extraction

​Code Location

How It Works

Pipeline Stages

1. Thread Ingestion

2. AI Analysis

3. Auto-Doc Generation

4. Relationship Extraction

Code Location