Skillget

Hokage_Vision_Agent

agent
Guvenlik Denetimi
Uyari
Health Uyari
  • License — License: Apache-2.0
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 5 GitHub stars
Code Gecti
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

Agentic anime detection workbench: mock/Ultralytics backends, CLI/API/GUI, tool-calling agent.

README.md

Hokage Vision Agent

Agentic anime character detection workbench — YOLO backends, PySide6 desktop, FastAPI, Typer CLI, tool-calling agent.

English | 中文

CI
License
Python

Portfolio-grade CV workbench. Detection is performed by a vision backend (mock / Ultralytics / legacy YOLOv5). The agent does not invent labels; it only chooses safe project tools (detect, validate dataset, smoke train, evaluate, compare, registry updates).

Docs site (MkDocs): https://phoenix0531-sudo.github.io/Hokage_Vision_Agent/

Screenshots (real Qt grab)

Hokage home overview GUI
Home overview — real PySide6 window (MainWindow.grab()) 		Image detection with mock boxes and table
Image detection — mock boxes + results table (obito/naruto/gaara)
Mock backend evidence figure
Backend evidence figure — reproducible matplotlib card 		Architecture schematic
Architecture schematic — CLI / GUI / API → backends 
# real window grab + mock detect on demo fixture
PYTHONPATH=src python scripts/capture_real_shots.py
PYTHONPATH=src python scripts/generate_evidence.py

Default demo classes: obito, naruto, gaara with confidences 0.91 / 0.84 / 0.77 — same mock path CI uses. No private YOLO weights required.

Design boundaries

CLI / PySide6 GUI / FastAPI
        │
        ▼
 InferenceService  ◄── Agent tools (RuleBasedAgent + ToolRegistry)
        │
        ▼
 VisionBackend: MockBackend | UltralyticsBackend | YOLOv5LegacyBackend
        │
        ▼
 Dataset / Training / Model registry
  • Shared core types and services across CLI, API, GUI, Agent
  • Default backend is mock: deterministic boxes so CI and demos need no GPU or private weights
  • Destructive / real training paths use careful / dry-run style entrypoints
  • Legacy YOLOv5 stays behind a dedicated backend — do not copy legacy package guts into src/hokage_vision

Package map (src/hokage_vision)

Area Role
vision/ Inference service, backends factory, evaluation, compare
agents/ Orchestrator, tool registry, safety, rule / OpenAI / LangGraph providers
api/ FastAPI app + routes + schemas
cli.py Typer multi-command CLI (hokage-vision)
data/ YOLO dataset helpers, manifest, validation, split, annotation assist
training/ Trainer, smoke train, model registry
config/ YAML settings loader (configs/*.yaml)
reports/ Markdown report helpers

Console script: hokage-vision = hokage_vision.cli:main.

CLI surface

hokage-vision detect ...
hokage-vision dataset ...
hokage-vision dataset manifest ...
hokage-vision annotation ...
hokage-vision train ...
hokage-vision model ...
hokage-vision agent ...

Quickstart

One command runs the whole pipeline (detect, validate, smoke train, agent, report) with the deterministic mock backend — no GPU, no weights, no network:

python examples/quickstart.py

All artifacts land under runs/quickstart/. See docs/quickstart.md for the 60-second walkthrough.

Install

Python >= 3.12. Hatchling src layout.

git clone https://github.com/Phoenix0531-sudo/Hokage_Vision_Agent.git
cd Hokage_Vision_Agent
python -m pip install -e ".[dev,api]"
# optional extras: gui, train (ultralytics), llm, desktop-build, docs, all

Docker-first path:

docker compose build
docker compose run --rm test

Quick usage

python -c "import hokage_vision; print(hokage_vision.__version__)"
hokage-vision --help
hokage-vision detect --help

# CI default paths
pytest -q tests/unit tests/integration

# evidence figure
PYTHONPATH=src python scripts/generate_evidence.py

GUI and full training need corresponding extras and (for real weights) local files under models/ — see docs/usage.md and docs/data-and-models.md.

Config

  • configs/app.default.yaml — mock backend default
  • configs/model.default.yaml
  • configs/agent.default.yaml
  • configs/dataset.example.yaml, configs/training.example.yaml

Tests and CI

Layer Location
Unit tests/unit/ — mock backend, inference, registry, agent tools, dataset, rendering, …
Integration tests/integration/ — API health, CLI detect mock, CLI help
GUI tests/gui/ — separate workflow
Packaging tests/packaging/

Product CI workflow: Python 3.12, hard editable install .[dev,api], critical ruff, pytest unit + integration only (GUI has its own workflow).

Scope

  • In: anime-character detection workbench, multi-surface UX (CLI/API/GUI), agent tool layer, dataset/train scaffolding, reproducible mock evidence
  • Out: production content-moderation SaaS; guaranteed SOTA without your own training data; committing private YOLO weights

License

Apache-2.0. See LICENSE and THIRD_PARTY_NOTICES.md.

Yorumlar (0)

Sonuc bulunamadi