Weaver: AI-Native Multi-Tenant Agentic Mesh

Coastal Alpine Tech portfolio

Part of the Kiwi Edge AI Stack | Founder OS: NZ-Start-Up | Agent policy: .github/agent-fleet/

Sovereign hybrid edge AI for NZ farms and founders - local-first + multi-model, Te Mana Raraunga aligned - collaborating with Venture Taranaki, startups.com investors and Kotahitanga Investment Fund (HITL + cultural advisory for formal approaches).

Agents inform, draft, prepare, monitor, and remind. Humans advise, sign, file, send, and pay. Anti-hallucination policy: .github/agent-fleet/anti-hallucination.md | Congruence: CAT_CONGRUENCE.md

Privacy / Security / Governance

Coastal Alpine Tech products treat operational and personal data as taonga. Defaults favour local-first operation, purpose-limited collection, and Human-in-the-Loop for high-stakes actions.

Hard commitments

Commitment	Statement
No data sales	We do not sell personal information or customer operational data to third parties for advertising, brokerage, or unrelated commercial exploitation.
NZ Privacy Act 2020	Collection, use, storage, and disclosure of personal information is designed to operate in accordance with the Privacy Act 2020 information privacy principles (including IPP awareness and IPP 3A indirect-collection notification where applicable).
Te Mana Raraunga	Where Māori data or community data interests arise, systems are designed to operate in accordance with Te Mana Raraunga principles (including Rangatiratanga, Whakapapa, Whanaungatanga, Kotahitanga, Manaakitanga, Kaitiakitanga) as a sovereignty and stewardship lens — not as a marketing slogan.
NZ AI safety	AI features follow a NZ AI safety-aligned posture: Algorithm Charter spirit (fairness, transparency, human oversight where relevant), digital.govt.nz / responsible AI guidance awareness, no silent model training on private journals without consent, and HITL for high-stakes outcomes.
Security	No silent exfiltration; owner-controlled credentials; least privilege; SecOps / dependency hygiene on the fleet cadence.
Governance	Agents inform, draft, prepare; humans advise, sign, file, send, and pay.

Pillar	Commitment
Privacy	Local-first / offline-capable where practical; Privacy Act 2020; Te Mana Raraunga spirit; third-party AI only when opt-in and labelled
Security	No silent exfil of tenant or personal data; owner-controlled keys
Governance	HITL for high-stakes; Te Mana Raraunga spirit; multi-region compliance maps in COMPLIANCE_REGIONS.md

Agents inform, draft, prepare, monitor, and remind. Humans advise, sign, file, send, and pay.

Fleet policy: fivepanelhat / Kiwi Edge AI Stack · COMPLIANCE.md · COMPLIANCE_REGIONS.md · SECURITY.md

Problems we are solving

Weaver is the multi-tenant edge orchestration and local RAG mesh for high-stakes Kiwi operations helpdesks.

1. Data leakage risk - Sending industrial / farm / client docs to public LLM APIs is unacceptable for many NZ operators.
2. Tenant cross-contamination - Shared SaaS helpdesks mix client knowledge unless isolation is designed in.
3. Rigid routing - Static ticket trees cannot adapt to multi-step operational requests.
4. Cloud dependency - Rural enterprises need local RAG and agent graphs that survive offline windows.

Solution we have built

Built capability	What it does
Tenant-aware agents	Intake / fulfilment / resolution with isolation
Local LangGraph orchestration	Adaptive routing on-edge
Partitioned vector knowledge	Per-tenant retrieval without cross-bleed
White-label posture	Industry-specific deployments on the Kiwi Edge stack

Pairs with Aether (companion skills) and Core (edge primitives) under the portfolio map.

Local (Taranaki) and national (Aotearoa) economic benefits

Lever	Benefit
Regional R&D HQ	Product design and IP stay in New Plymouth / Taranaki - not only Auckland/offshore SaaS
Primary-sector productivity	On-farm and rural tools aim to cut waste, protect consents, and support export competitiveness
Skilled employment pathways	Edge install, field support, agritech ops, software, compliance, and cultural advisory roles as pilots scale
Data sovereignty	Te Mana Raraunga-aligned local custody keeps high-value operational data onshore
HITL jobs quality	Agents inform / draft / prepare / monitor / remind; humans advise / sign / file / send / pay - augment people, do not fake full autonomy

Stage honesty (pre-seed): Impact today is founder R&D, near-term contractors, and EDA/partner leverage. Permanent multi-region payroll follows paid pilots and revenue - we do not invent headcount claims.

Coastal Alpine Tech Limited pre-seed startup, New Plymouth, Taranaki, Aotearoa New Zealand. Edge AI | Sovereign Systems | Practical Intelligence

License: Proprietary Coastal Alpine Tech Limited | See LICENSE (same Coastal Alpine proprietary terms as AquaGuard, SoilGuard, Blue-Moon, Sting-Operation, Core, and the rest of the edge stack).

White-label multi-tenant AI helpdesk scaffold with isolated knowledge retrieval and local LangGraph orchestration.

---

The 5 Ws: Project Context

• Who: Built by Coastal Alpine Tech Limited (pre-seed), designed for high-stakes Kiwi industries (civil construction, agritech, etc.).
• What: A decentralized LangGraph orchestration layer that safely directs multi-agent tasks and handles local document vectorization.
• Where: Engineered at HQ in New Plymouth, Taranaki. Deployable strictly at the edge.
• When: Active development as of June 2026.
• Why: To support data sovereignty by keeping tenant operational data local and strictly partitioned.

---

---

Key Features

• Tenant-aware multi-agent orchestration (Intake, Fulfilment, Resolution)
• Strict data isolation via SQLAlchemy and tenant-partitioned vector stores
• Local LangGraph-based state machine for adaptive routing
• Modular knowledge base with RAG support
• White-label ready for industry-specific deployments
• Full offline edge capability

---

Quick Start

Prerequisites

• Python 3.10+
• Ollama with a local LLM Gemma 4 E4B
• PostgreSQL (optional) or in-memory mode

Installation & Setup

Weaver is Windows + Linux (and macOS) ready. Edge production target remains RPi 5 16GB + Hailo-10H. Hybridised with Coastal-Alpine-Core, Aether (companion skills / HITL), and coastal-alpine-stack.

• Prerequisites: setup.md
• Full install guide: installation.md

One-line install (recommended)

Linux / macOS

curl -fsSL https://raw.githubusercontent.com/fivepanelhat/Weaver/main/install.sh | bash

Windows (PowerShell)

irm https://raw.githubusercontent.com/fivepanelhat/Weaver/main/install.ps1 | iex

Note: If script execution is blocked: Set-ExecutionPolicy -Scope CurrentUser RemoteSigned

Cross-platform bootstrap (from a clone)

git clone https://github.com/fivepanelhat/Weaver.git
cd Weaver
python bootstrap.py # Linux / macOS / Windows creates venv, installs Core + deps

Manual Installation

Linux / macOS (Bash)

git clone https://github.com/fivepanelhat/Weaver.git
cd Weaver

python3 -m venv venv
source venv/bin/activate

pip install -U pip
pip install "git+https://github.com/fivepanelhat/Coastal-Alpine-Core.git@v0.5.5"
pip install -r requirements.txt
pip install -r requirements-dev.txt
cp .env.example .env

System packages (Debian/Ubuntu/RPi OS):

sudo apt-get update
sudo apt-get install -y python3-dev python3-venv python3-pip git build-essential

Windows (PowerShell)

git clone https://github.com/fivepanelhat/Weaver.git
cd Weaver

python -m venv venv
.\venv\Scripts\Activate.ps1

python -m pip install -U pip
pip install "git+https://github.com/fivepanelhat/Coastal-Alpine-Core.git@v0.5.5"
pip install -r requirements.txt
pip install -r requirements-dev.txt
Copy-Item .env.example .env

Prerequisites: Python 3.10+ with "Add Python to PATH", Git for Windows.

Model Setup & Validation

Ensure Ollama is running locally and pull the target model:

ollama pull gemma4:e4b
python demo.py

To run smoke tests validating state graph transitions:

pytest

---

Architecture Overview

Diagrams: Architecture images and Mermaid maps describe the target product architecture for this pre-seed stack. They are engineering design maps not claims of large-scale commercial fleet deployment.

Weaver routes multi-tenant requests entirely on the edge node: RPi 5 16GB + Hailo-10H, local Ollama, and tenant-isolated stores. No tenant data leaves the deployment site.

System map

%%{init: {
 "theme": "dark",
 "themeVariables": {
 "fontSize": "15px",
 "fontFamily": "Inter, ui-sans-serif, system-ui, sans-serif",
 "primaryColor": "#0ea5e9",
 "primaryTextColor": "#f8fafc",
 "primaryBorderColor": "#38bdf8",
 "lineColor": "#67e8f9",
 "secondaryColor": "#1e293b",
 "tertiaryColor": "#0f172a",
 "clusterBkg": "#0b1220cc",
 "clusterBorder": "#38bdf880",
 "titleColor": "#e2e8f0"
 },
 "flowchart": {
 "nodeSpacing": 36,
 "rankSpacing": 44,
 "padding": 18,
 "htmlLabels": true,
 "curve": "basis",
 "useMaxWidth": true
 }
}}%%
flowchart TB

 classDef act fill:#422006,stroke:#fbbf24,stroke-width:2px,color:#fffbeb
 classDef core fill:#134e4a,stroke:#2dd4bf,stroke-width:2px,color:#f0fdfa
 classDef store fill:#1e1b4b,stroke:#a5b4fc,stroke-width:2px,color:#eef2ff
 classDef ai fill:#3b0764,stroke:#e879f9,stroke-width:2px,color:#fdf4ff
 classDef sdk fill:#0c4a6e,stroke:#38bdf8,stroke-width:2px,color:#f0f9ff
 classDef host fill:#052e16,stroke:#4ade80,stroke-width:2px,color:#f0fdf4
 classDef companion fill:#312e81,stroke:#c4b5fd,stroke-width:2px,color:#eef2ff

 U["User / operator query"] --> ORCH["LangGraph orchestrator"]
 ORCH --> IN["Intake agent<br/>auth | tenant scope"]
 ORCH --> FU["Fulfilment agent<br/>RAG | tools"]
 ORCH --> RE["Resolution agent<br/>response | actions"]
 IN & FU & RE --> KB["Tenant-aware knowledge base"]
 KB --> STORE["Isolated vector + SQL store"]
 STORE --> LLM["Local LLM via Ollama<br/>Gemma 4 e4b"]
 LLM --> ORCH
 ORCH --> OUT["Actions & responses"]

 subgraph HYBRID["Hybrid stack integration"]
 CAC["Coastal-Alpine-Core<br/>SecurityGuard | Telemetry | Flywheel"]
 AETH["Aether companion<br/>skills | HITL | computer use"]
 CAS["coastal-alpine-stack<br/>compose / K3s"]
 end

 subgraph HOSTS["Dual-platform hosts"]
 WIN["Windows 10/11<br/>install.ps1 | bootstrap.py"]
 LIN["Linux / RPi OS<br/>install.sh | bootstrap.py"]
 RPI["RPi 5 16GB + Hailo-10H<br/>production edge"]
 end

 ORCH --> CAC
 CAC --> LLM
 AETH -.-> | dev / remediate | ORCH
 CAS -.-> ORCH
 ORCH -.-> HOSTS

 class U,OUT act
 class ORCH,IN,FU,RE core
 class KB,STORE store
 class LLM ai
 class CAC,CAS sdk
 class AETH companion
 class WIN,LIN,RPI host

Loading

Layer	Components	Role
Orchestrator	LangGraph state machine	Deterministic multi-agent routing
Agents	Intake	Fulfilment
Knowledge	Isolated vector + SQL	No cross-tenant leakage
Inference	Ollama on-device	Offline-capable responses
SDK hybrid	Coastal-Alpine-Core	Guards, telemetry, flywheel on every path
Companion	Aether	Dev orchestration, HITL, computer use
Hosts	Windows	Linux

Full detail: ARCHITECTURE.md

Directory Structure

Weaver/
|-- agent_knowledge_base/ # Policy, ethics, and platform runbooks (Markdown)
|-- weaver_graph/ # Edge-friendly state graph (does not shadow PyPI langgraph)
| |-- graph.py # StateGraph compiler
| |-- llm.py # Local Ollama client bridge (LocalSovereignLLM)
| |-- orchestrator.py # build_agnostic_helpdesk graph nodes
| |-- embeddings.py # Embedding helpers
| -- ingestion.py # Document ingestion
|-- orchestrator.py # AgentOrchestrator unified entrypoint (agent + graph paths)
|-- agents.py # Intake / Fulfilment / Resolution agents
|-- knowledge_base.py # Tenant-isolated KB clients (in-memory + SQLAlchemy)
|-- database.py # TenantAwareDB connection utilities
|-- models.py # SQLAlchemy relational & vector schemas
|-- demo.py # Local simulation runner (offline-capable)
|-- bootstrap.py # Cross-platform venv + dependency bootstrap
|-- tests/ # pytest suite (orchestrator, LLM URL, demo smoke)
|-- tests_security_stress/ # Adversarial / red-team suite (prompt attacks)
|-- .env.example
|-- requirements.txt
|-- requirements-dev.txt
|-- Dockerfile
|-- ARCHITECTURE.md # System design details
-- README.md # This file

---

Technology Stack

Hardware

• Raspberry Pi 5 (16GB) with Hailo-10H NPU (40 TOPS AI Accelerator / AI HAT+ 2)

Software

• Orchestration: LangGraph
• Inference: Ollama + Local LLMs
• Data: SQLAlchemy + pgvector / local vector stores
• Deployment: Docker-ready, systemd compatible

---

Target Deployment Scenarios

These are the use cases Weaver is designed for. Weaver is at an early release (v0.1.0) and these are illustrative target scenarios, not claims of existing production deployments.

• Civil Construction Helpdesk: Route project-compliance queries across multiple subcontractors on-premise, keeping each client's data strictly isolated.
• Agritech Support Platform: Provide localized advisory services to cooperatives without exposing farm data to third-party clouds.
• White-Label Service Providers: Embed into existing SaaS platforms whose clients require sovereign, on-site data handling.

Implementation Notes:

• Install on a dedicated edge server or Raspberry Pi cluster.
• Configure tenant IDs at database and vector store level for isolation.
• Use systemd services for persistent operation and monitor via local dashboards.
• Start with demo.py to validate routing and isolation before production scaling.

---

Performance Targets

Preliminary, illustrative figures measured informally on the reference edge node (RPi 5 16GB + Hailo-10H, Gemma 4 E4B via Ollama). Not audited production benchmarks treat as ballpark expectations and re-measure for your own workload and hardware.

• Routing latency: on the order of ~1 second per routing decision.
• Active power draw: roughly ~6W on a headless Raspberry Pi 5 16GB node.
• Storage footprint: SQL + vector stores well under 200MB for small tenants.

---

Documentation

• ARCHITECTURE.md
• agent_knowledge_base/platform_runbook.md
• agent_knowledge_base/ethics_review_playbook.md
• CHANGELOG.md

---

License

This project is licensed under the Coastal Alpine Tech Limited License (proprietary / commercial) the same license used across the Kiwi Edge AI Stack (Weaver, AquaGuard, SoilGuard, Blue-Moon, Sting-Operation, Coastal-Alpine-Core, coastal-alpine-stack, Sovereign-Edge-Firmware, fivepanelhat).

• Full terms: LICENSE
• No open-source grant is implied by access to this repository
• Commercial use requires a written agreement with Coastal Alpine Tech Limited

---

Built with focus on data sovereignty and edge intelligence. Questions or collaboration? Contact Coastal Alpine Tech Limited.

---

Last updated: July 2026 | First public release: v0.1.0