Skip to content

AGENTS.md

Latest commit

 

History

History
44 lines (28 loc) · 1.95 KB

AGENTS.md

File metadata and controls

44 lines (28 loc) · 1.95 KB
Raw

@getpipher/vision

Capability-aware vision + paste extension for the pi coding agent. npm: @getpipher/vision. Replaces pi-vision-tool + pi-paster.

Org context: getpipher is the Pi coding-agent ecosystem. No GitLab mirror for getpipher.

What

Adds a describe_image tool to pi that is aware of the active primary model's input modalities — so image analysis is never wasteful.

  • Multimodal primary model (e.g. minimax-m3:cloud, qwen3.5:cloud) → the image passes through to the model natively. Zero delegation calls, zero extra tokens, zero extra latency. The describe_image tool is hidden from the model entirely, so the wasted call is structurally impossible — not just discouraged.
  • Text-only primary model → describe_image delegates to a configured vision model and returns the text description (the existing, working pattern).

The capability check is automatic and silent — no config, no opt-in, no flag.

Install

pi install npm:@getpipher/vision

@getpipher/vision replaces pi-vision-tool and pi-paster. Uninstall those first to avoid duplicate tools and double-attachment.

Structure

extensions/   # pi extension registering the describe_image tool + paste/paste-detection
lib/          # capability detection, delegation, marker rendering, globalThis shared state
tests/        # tsx --test suite (203 tests)

Common Commands

pnpm typecheck   # tsc --noEmit
pnpm test:run     # tsx --test

Notes

  • jiti dual-module-instance pitfall: shared state must live on globalThis, not module scope (memory pi-jiti-dual-module-instance.md).
  • Roadmap: v0.4.0 (batch + scale — 10+ images, parallel delegation) and v0.5.0 remain after v0.3.x shipped.