cozymori
cozymori

Simpler, Easier, For Developers. Open-source frameworks for AI observability.

Products

  • VectorWave
  • VectorSurfer

Resources

  • Documentation
  • GitHub

© 2026 cozymori. All rights reserved.

Built with simplicity.

Overview

Getting Started

  • Introduction
  • Quick Start

VectorWave

  • VectorWave Overview
  • Installation
  • @vectorize Core
  • Semantic Caching
  • Self-Healing
  • Golden Dataset
  • Drift Detection
  • Replay Testing
  • RAG Search
  • Advanced Configuration
  • API Reference

VectorSurfer

  • VectorSurfer Overview
  • Getting Started
  • Usage Guide

Ecosystem

  • VectorCheck
  • VectorSurferSTL
  • Contributing

VectorSurfer

VectorSurfer Overview

AI observability dashboard — monitor, trace, and heal your AI functions.

What is VectorSurfer?

VectorSurfer is a real-time AI observability dashboard. It connects to your VectorWave data and provides visual tools for monitoring, tracing, and healing AI function calls.

Your App + @vectorize → Weaviate ← VectorSurfer Dashboard

VectorWave captures function executions, and VectorSurfer turns that data into actionable insights — all in one unified interface.

Key Features

Real-Time Dashboard

A responsive widget grid dashboard that shows all key metrics at a glance:

  • KPI Cards — Total calls, success rate, errors, cache hit rate, average duration
  • Execution Timeline — Function executions over time with success/error/cache breakdown
  • Token Usage — LLM token consumption by category
  • System Status — Weaviate connection status and registered function count

Trace Waterfall

Visualize distributed traces as interactive waterfall diagrams:

  • See the full call chain of decorated functions
  • Identify bottlenecks and slow spans instantly
  • Click into any span for detailed execution data
  • Color-coded by status (success, error, cached)

AI Healer

GPT-4 powered error diagnosis with a visual interface:

  • Browse recent errors with full context
  • One-click root cause analysis
  • View generated code fixes with diff highlighting
  • Create GitHub PRs directly from the dashboard
  • Batch healing for multiple errors at once

Replay Testing

Visual regression testing for AI functions:

  • Select functions and Golden Dataset entries
  • Run replays with real-time progress
  • Compare expected vs actual outputs side-by-side
  • Supports both exact match and semantic comparison

Ask AI

RAG-powered natural language Q&A over your VectorWave data:

  • Ask questions about your functions, errors, and execution patterns
  • AI retrieves relevant context from Weaviate and generates answers
  • Responses are automatically saved to AI History

AI History (Saved Responses)

Browse and manage all AI-generated responses:

  • Filter by source (Ask AI / Healer), bookmark status, or keyword
  • Bookmark important responses for quick access
  • Free plan: 5 AI calls/day, responses locked after 24 hours
  • Pro plan: unlimited calls and full history access

Semantic Search

Natural language search across all your AI function data:

  • Search functions, execution logs, and errors
  • Hybrid search (vector similarity + keyword filtering)
  • RAG-powered Q&A for exploring your codebase

Multi-Language UI

English, Japanese, and Korean UI out of the box — built for global teams.

How It Works

VectorSurfer reads from the same Weaviate instance that VectorWave writes to. There is no extra instrumentation needed — just use @vectorize in your Python code, and VectorSurfer shows everything.

VectorWave is not required to use VectorSurfer, but the SDK is what generates the data the dashboard displays.

Alternatives

DashboardBest For
VectorSurferProduction, teams, full-featured
VectorSurferSTLQuick setup, local dev, Python-only

Next Steps

  • Getting Started — Connect to VectorSurfer
  • Usage Guide — Detailed feature walkthrough
  • VectorWave Overview — The core SDK