Mark/webui_AWM
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
AWM pipe function for openwebui, includes python runner for direct execution
22 commits 1 branch 0 tags 710 KiB
Find a file
Exact
hermes 39e3a61934
All checks were successful
CI / ci (push) Successful in 1m4s
Details
ci: re-run via runner pickup verification
2026-09-26 20:07:05 -07:00
.forgejo/workflows fix(ci): self-contained workflow — drop unreachable JS actions 2026-09-16 08:55:29 -07:00
docs feat: implement multi-level recursive artifact generation, persistence, and CLI auto-mode 2026-01-08 19:53:12 -08:00
src Update: Analysis of section artifact writing functionality 2026-01-09 17:24:54 -08:00
tests Update run.py with file attachment support and cleanup unused files 2026-01-14 19:01:03 -08:00
.gitignore chore: ignore out/ directory and remove previously committed artifacts from index 2026-01-08 19:53:46 -08:00
.pre-commit-config.yaml chore: add basic pre-commit hooks (trailing-whitespace, eof-fixer, ruff) 2026-06-03 17:29:39 -07:00
AGENTS.md feat: implement multi-level recursive artifact generation, persistence, and CLI auto-mode 2026-01-08 19:53:12 -08:00
awm.py Update: Analysis of section artifact writing functionality 2026-01-09 17:24:54 -08:00
build.py feat: implement multi-level recursive artifact generation, persistence, and CLI auto-mode 2026-01-08 19:53:12 -08:00
generic_pipe_cli.py Add OpenWebUI pipe templates extracted from AWM workflow 2026-01-09 20:52:53 -08:00
PIPE_DEVELOPMENT_GUIDE.md Add documentation for OpenWebUI pipe development 2026-01-09 21:49:50 -08:00
PIPE_TEMPLATE_README.md Add OpenWebUI pipe templates extracted from AWM workflow 2026-01-09 20:52:53 -08:00
QWEN.md Update: Analysis of section artifact writing functionality 2026-01-09 17:24:54 -08:00
README.md Add README for AWM implementation 2026-01-09 23:59:54 -08:00
run.py Update run.py with file attachment support and cleanup unused files 2026-01-14 19:01:03 -08:00

README.md

AWM (Artifact Writing Machine)

Overview

AWM (Artifact Writing Machine) is an AI-powered document generation system that creates structured documents with multiple sections using a multi-agent pipeline approach.

Architecture

The system uses four main agents in a pipeline:

  • Intent Agent: Classifies requests and determines scope
  • Planner Agent: Creates execution plans
  • Executor Agent: Generates content
  • Reviewer Agent: Validates output quality

Features

  • Hierarchical Document Structure: Creates parent-child relationships between documents and sections
  • Section Artifacts: Individual files for each document section (e.g., artifact_01_section_01.md)
  • Automatic Compilation: Combines sections into complete documents
  • Follow-up Commands: Auto-generates next steps for document completion
  • Multi-format Support: HTML, Markdown, JavaScript, CSS, SVG

File Structure

  • awm.py: Main system implementation (built from src/ modules)
  • run.py: CLI runner for executing AWM workflows
  • build.py: Builds awm.py from modular source files
  • src/: Modular source code
  • out/: Output directory with timestamped project folders
  • tests/: Test files

Usage

python run.py "Create a document with multiple sections" --auto

Generated files are saved in out/YYYYMMDD_HHMMSS_<slug>/ with:

  • Individual section files: artifact_XX_section_YY.ext
  • Compiled project: compiled_project.ext
  • Registry: registry.json

Key Concepts

  • NORMAL Scope: Creates atomic content artifacts (no TODOs)
  • LARGE Scope: Creates manifest artifacts (with TODOs for sections)
  • Artifacts: Named as artifact_XX (main) and artifact_XX_section_YY (sections)
  • Persistence: Uses code blocks in conversation history to maintain state