Elaway Knowledge Base API

What It Is

Search engine for Elaway's internal knowledge - documents, support history, Slack conversations, and deal data. Automatically syncs daily from 8+ sources and provides semantic search via 32 specialized tools.

Who Uses It

Sales: Find similar deals, pricing history, customer issues
Support: Search past ticket resolutions, common problems, FAQs
Product: Discover feature requests, bugs, usage patterns
Leadership: Access company docs, meeting notes, decisions

Typical Use Cases

Quick answers: "What's our policy on subscription transfers?"
Troubleshooting: "How do we fix charging error code E503?"
Customer context: "Show me all tickets from BRF Solsiden"
Deal research: "Compare installation costs for similar buildings in Oslo"
Pattern analysis: "What are common reasons for unstable charging?"
Expert finder: "Who participates most in #help-me-platform threads?"
Data freshness: "When was the last Notion sync?"

How It Works

Daily syncs pull latest data from Notion (05:00), HubSpot (06:00), Slack (06:00)
Embeddings generated for all text using OpenAI (1536-dim vectors)
Stored in pgvector with HNSW indexes for fast similarity search
Query via JSON-RPC using semantic search or table filters
Results ranked by cosine similarity and returned with full context

9,588

Documents

15,095

Tickets

893

Slack

WoD Deals

127

Notion

Quick Start

POST /functions/v1/mcp-readonly

Base URL: https://<PROJECT_REF>.supabase.co

Example Request

curl -X POST "https://<PROJECT_REF>.supabase.co/functions/v1/mcp-readonly" \
  -H "Authorization: Bearer $SUPABASE_ANON_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/call",
    "params": {
      "name": "kb_search",
      "arguments": {
        "query": "charging error codes",
        "match_count": 5
      }
    }
  }'

Available Tools

Search & Discovery

kb_search

Primary semantic search across all sources (documents, tickets, Slack, WoD)

search

OpenAI Deep Research compatible search (flat results)

fetch

Get full content by ID after search

Data Browsing

query_table

SQL-like queries with filters (up to 500 rows)

count_rows

Aggregations, statistics, group by

list_tables

Discover all 14 tables and schemas

db_table_stats

Quick row counts for all tables

Relationship Traversal

fetch_thread_messages

Complete Slack thread (up to 200 messages)

fetch_conversation_messages

Complete HubSpot ticket thread

fetch_document_full

Full document content (up to 200K chars)

fetch_wod_deal_full

Complete WoD deal with circuits, costs, offers

fetch_notion_children

Navigate Notion page hierarchy

Analytics & Insights

analyze_slack_thread_network

Identify experts by participation

analyze_training_resolution_patterns

Common issues and resolutions

find_similar_documents

Find docs by metadata similarity

compare_wod_deals

Side-by-side deal comparison (2-5 deals)

aggregate_costs_by_category

WoD cost analytics

get_data_freshness_report

Check sync status for all sources

Embedding Tools

get_embeddings

Raw 1536-dimensional vectors

generate_embedding

Convert text to vector

compute_similarity

Compare text against records

Database Schema

Table	Records	Description
`documents`	9,588	PDFs, Word docs, FAQs, specifications, guides
`training_conversations`	15,095	HubSpot support tickets with embeddings
`training_messages`	~50K	Individual messages within conversations
`slack_threads`	450+	Slack thread root messages
`slack_messages`	893	Slack messages with embeddings
`wod_deals`	57	Wheel of Deal pricing (42 columns)
`notion_pages`	127	Platform documentation

Data Freshness

Source	Sync Schedule
Slack messages	Daily at 06:00 UTC
HubSpot tickets	Daily at 06:00 UTC
Notion pages	Daily at 05:00 UTC
Documents	Manual ingestion

Response Format

{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "content": [
      {
        "type": "text",
        "text": "{\"results\": [...], \"total\": 5}"
      }
    ]
  }
}