Skip to content

Architecture

Jump to bottom
Your GitHub Name edited this page May 19, 2026 · 5 revisions

Architecture

PromptPilot is built around one idea:

A small language model can serve as the harness brain around a larger coding agent.

The SLM is not the coder. It is the control layer.

It decides:

  • What is the user asking?
  • Is the prompt clear enough?
  • Is this safe to rewrite?
  • Should the request be answered directly?
  • Should it be passed through unchanged?
  • Should noisy tool output be compressed?
  • What facts must be preserved?
  • Should the expensive coding agent be invoked?

The large coding agent remains responsible for:

  • Code understanding
  • Implementation
  • Debugging
  • Test repair
  • Architectural reasoning
Developer prompt / tool output
        ↓
PromptPilot
        ↓
SLM harness
  - intent
  - ambiguity
  - risk
  - preservation
  - routing
        ↓
Route
  - clarify
  - answer
  - passthrough
  - compress
  - invoke agent
        ↓
Codex / Claude-style coding agent
        ↓
Telemetry / replay / audit

Runtime layers

PromptPilot separates workflow control from coding execution:

  1. Input capture receives the developer prompt, session memory, and optional tool output.
  2. SLM harness classifies intent, extracts constraints, evaluates ambiguity, and estimates transformation risk.
  3. Route selection chooses clarify, answer, passthrough, compress, rewrite, or invoke-agent behavior.
  4. Frontier coding agent receives either the original prompt or a semantic-preserving harness output.
  5. Telemetry and replay record routing, token impact, compression events, and session handoff data.

This keeps the SLM central but bounded: it controls the workflow around the coding agent without replacing the model that reasons about and edits code.

Related pages

Clone this wiki locally