AI_ENGINEERING2 Multi-file Components

Local Ollama & Open-Weights Model Orchestrator

Guide for running offline AI agents against local GPUs and Apple Silicon with Ollama and LangChain.

Author: TRIHEX Architecture Lab

•

License: MIT

•

Version: 2.1.0

Inert Code Verified · Heuristic Static Security ScannerHeuristic Scanner: LOW Risk

Heuristic static security scanner audited 2 files (2520 bytes). External code. Review before execution.

v2.1.0

Skill Files (2)

Compatible Runtimes

AntigravityClaude CodeCursor

SKILL.md·markdown

# Local Ollama & Open-Weights Model Orchestrator

Deploying and querying local Llama 3.3, DeepSeek-R1, and Qwen2.5 with zero API cost and private data custody.

## Core Invariants & Rules
1. Point base URL to http://localhost:11434/v1 with temperature: 0.1 for deterministic logic.
2. Verify input validation and security boundaries on all executions.
3. Keep configurations modular, typed, and auditable.

## Operational Workflow
1. Analyze requirements against existing codebase patterns.
2. Apply standardized implementation templates.
3. Execute verification suites before deployment.