Skip to content

Local LLM Tiering ($0.00 First)

gosec implements a 3-tier cognitive LLM hierarchy engineered to maximize cost efficiency without sacrificing intelligence.

The 3-Tier Cognitive Architecture

flowchart TD
    Prompt["Developer / Agent Request"] --> Check{"Local Ollama Available?"}
    Check -->|Yes| Tier0["Tier 0/1: Local Ollama ($0.00)\nqwen3-coder:30b / deepseek-r1:14b"]
    Check -->|No / Busy| Tier2["Tier 2: Fast Cloud Default\nGemini 3.8 Flash"]
    Tier0 --> Complex{"Complex Architecture or ADR?"}
    Tier2 --> Complex
    Complex -->|Yes| Tier3["Tier 3: Deep Cognitive Tier (Restricted)\nGemini 3.8 Pro"]
    Complex -->|No| Output["Completed Code & Artifacts"]
    Tier3 --> Output
Hold "Alt" / "Option" to enable pan & zoom

1. Tier 0/1: Local Ollama ($0.00)

  • Target Host: http://127.0.0.1:11434
  • Recommended Models:
  • qwen3-coder:30b: Primary coding and refactoring engine.
  • deepseek-r1:14b: Deep reasoning, logic checks, and AST analysis.
  • llama3.3:70b: High-context comprehension.
  • Cost: $0.00 (runs entirely on local GPU/Metal).
  • Graceful Fallback: If Ollama is offline or times out, gosec seamlessly routes requests to Tier 2 with zero hard errors.

2. Tier 2: Fast Cloud Default

  • Target Model: Gemini 3.8 Flash.
  • Role: High-throughput code synthesis, test generation, MCP tool calls, and CI triage.
  • Speed: Sub-second latency.

3. Tier 3: Deep Cognitive Architecture (Restricted)

  • Target Model: Gemini 3.8 Pro.
  • Role: Reserved strictly for complex system design, architectural decision records (ADRs), and critical security threat modeling.

Configuration (config.yaml)

llm:
  provider: "ollama"
  endpoint: "http://127.0.0.1:11434"
  default_model: "qwen3-coder:30b"
  fallback_model: "deepseek-r1:14b"
  timeout_seconds: 60
  temperature: 0.0
  auto_fallback: true