enachb

Design: Synchronous gRPC Pipeline via Dapr Service Invocation

Summary

Replace the async NATS pub/sub chain after step 4 (detection + tracking + categorization) with synchronous gRPC calls using Dapr Service Invocation. The detection-agent becomes the orchestrator: it invokes downstream agents via gRPC, waits for responses, aggregates all enrichment results into a single AnnotatedImage, and publishes one final message per frame to NATS for persistence.

Steps 1-4 remain unchanged (NATS ingestion -> YOLO -> ByteTrack -> categorize). NATS is also retained at the output edge for persistence fan-out.

---

enachb

System Design Document: Agentic CEL Rule Engine

Status: Draft v1.0 Author: Sensemesh Architecture Team Context: Event-Driven Architecture, gRPC/Protobuf, Go, NATS

---

1. Executive Summary

enachb

WebAssembly (WASM) Offloading Analysis

Date: January 25, 2026 Scope: Client-side optimization, Cost Reduction, Bandwidth Efficiency

1. Executive Summary

Moving the initial stages of the AIBatcher pipeline (technical filtering, deduplication, and face detection) to the client-side using WebAssembly (WASM) presents a massive opportunity for optimization.

By filtering out low-quality and duplicate images before upload, we can reduce bandwidth consumption and server-side processing costs by approximately 80%. This transforms the workflow from "Upload All -> Process All" to "Filter Locally -> Upload Best -> AI Grade".

enachb

WebAssembly (WASM) Offloading Analysis

Date: January 25, 2026 Scope: Client-side optimization, Cost Reduction, Bandwidth Efficiency

1. Executive Summary

Moving the initial stages of the AIBatcher pipeline (technical filtering, deduplication, and face detection) to the client-side using WebAssembly (WASM) presents a massive opportunity for optimization.

By filtering out low-quality and duplicate images before upload, we can reduce bandwidth consumption and server-side processing costs by approximately 80%. This transforms the workflow from "Upload All -> Process All" to "Filter Locally -> Upload Best -> AI Grade".

enachb

WebAssembly (WASM) Offloading Analysis

Date: January 25, 2026 Scope: Client-side optimization, Cost Reduction, Bandwidth Efficiency

1. Executive Summary

Moving the initial stages of the AIBatcher pipeline (technical filtering, deduplication, and face detection) to the client-side using WebAssembly (WASM) presents a massive opportunity for optimization.

By filtering out low-quality and duplicate images before upload, we can reduce bandwidth consumption and server-side processing costs by approximately 80%. This transforms the workflow from "Upload All -> Process All" to "Filter Locally -> Upload Best -> AI Grade".

enachb

Claude Code Extension Points

A Team Presentation (30-45 min)

---

Agenda

1. Prerequisites & Setup (5 min)
2. What is Claude Code? (5 min)
3. Extension Points Overview (5 min)

enachb

Migrating from Claude Code to Google Antigravity

This guide outlines how to migrate the current Claude Code + Muxile workflow to Google Antigravity (VS Code-based AI IDE).

1. Directory Structure

Create the configuration directory in your project root:

mkdir -p .antigravity/{config,scripts,prompts}

enachb

Muxile Integration Guide

What is Muxile?

Muxile is a tmux plugin that provides mobile web access to your running tmux sessions. It creates a secure WebSocket tunnel through Cloudflare Workers, allowing you to monitor and control your terminal from any device with a web browser.

Why Use Muxile with Claude Code?

When running long-duration Claude Code tasks (multi-step refactoring, data processing, batch operations), you can:

enachb

Mission Alerting System - Technical Overview

Overview

The Mission Alerting System enables real-time rule-based alerting on detection events from cameras and sensors. When a detection event matches a CEL (Common Expression Language) rule, the system executes configured actions (SMS notifications via Twilio).

Architecture

Detection Events (NATS)  →  Detection Consumer  →  Rule Processor  →  SMS Delivery

enachb

Week 3 Implementation Plan: Missions & Alerts

Executive Summary

Recommendation: Enhance Mission Service with CEL-based rule evaluation by incorporating proven code from Alerts Service.

Key Insight: Mission service is the correct home for mission automation. We'll add CEL evaluation capability, SMS/Email delivery, and event processing to mission service by copying proven implementations from alerts service.

Schema Approach: Extend AutomationPolicy to support both CEL expressions (Week 3) and natural language instructions (future AI). This allows simple rules now while preserving the path to AI-driven automation.