OmniForge AI Workbench

Windows-first local AI orchestration platform that unifies multiple repos/tools into one CLI + REST API + dashboard.

It combines:

• Supermix_27 (local chat runtime)
• NeuroDSL-Infinity-Studio2 (NeuroDSL CLI + agent API bridge)
• nexusflow (workflow DSL execution)
• npu_easy + CIFAR-100/cifar100_model.onnx (local ONNX vision inference via DirectML/CPU)
• single-file-lab-studio (dashboard style/pattern)
• Windhawk (hotkey mod generation/restart integration)

This is an orchestration/workbench system, not a newly trained foundation model.

What It Does

OmniForge exposes a single local interface for:

• Chat + vision inference
• NeuroDSL tooling and hosted serve-agent-api proxying
• NexusFlow workflow execution
• Agentic planning/execution and NexusFlow export/run
• Benchmarks, comparisons, trends
• Async job queue + scheduler
• Search and mission packs
• Reports, alerts, remediations
• Notifications, health snapshots, and incident response workflows

Current Feature Set

Core runtime

• Flask API + local dashboard (/)
• CLI (python -m omniforge_ai ...)
• Generic operation router (ops-catalog, op-run, op-submit)
• Persistent history + artifact browser

AI / workflow integrations

• Supermix local chat adapter
• ONNX vision inference (npu_easy + DirectML/CPU fallback)
• NeuroDSL bridge + agent API subprocess manager/proxy
• NexusFlow example/file runner
• Windhawk mod generation + restart helpers
• Agentic planner/executor + NexusFlow pipeline export/run

Observability / automation

• Benchmark arena + analytics (list/compare/trends)
• Async jobs + persistent job records
• Scheduler (persisted recurring tasks)
• Reports (ops_digest, alert_watch, health_watch, incident_watch, etc.)
• Alerts + acknowledgements
• Auto-remediations (job submit / report / scheduler / generic op / notification / health snapshot)
• Notifications (console, history-event, file append, webhook)
• Health monitor (snapshots, scoring, compare)
• Incident Response Center (manual + auto-open from critical alerts, timeline, bundle capture)
• OpsGraph intelligence (snapshots, graph trace/path/blast-radius, incident explainers, watch runs, hotspots, component risk mapping)

Expected Workspace Layout

omniforge_ai expects sibling repos in the parent workspace directory (see omniforge_ai/omniforge_ai/config.py):

<workspace>\
  omniforge_ai\
  Supermix_27\
  NeuroDSL-Infinity-Studio2\
  nexusflow\
  npu_easy\
  CIFAR-100\
  single-file-lab-studio\

It also expects a local Windhawk install at:

• C:\Program Files\Windhawk (optional but recommended for hotkey integration)
• C:\ProgramData\Windhawk (for writable mods)

Requirements

• Windows (tested on local Windows environment)
• Python 3.10+ (3.13 has been used successfully)
• Local clones of the repos listed above
• Python dependencies:

cd omniforge_ai
python -m pip install -r requirements.txt

requirements.txt currently includes:

• flask
• numpy
• pillow
• torch
• onnxruntime

Quick Start

Start API + dashboard

cd omniforge_ai
python -m omniforge_ai serve --host 127.0.0.1 --port 8787

Open http://127.0.0.1:8787.

Start with Windhawk hotkey integration

python -m omniforge_ai serve-windhawk --host 127.0.0.1 --port 8787

This generates/copies the Windhawk hotkey mod and optionally restarts Windhawk before starting the server.

CLI Quick Examples

Core / adapters

python -m omniforge_ai status
python -m omniforge_ai smoke
python -m omniforge_ai chat "hello"
python -m omniforge_ai vision --image C:\path\to\image.jpg
python -m omniforge_ai neurodsl-devices
python -m omniforge_ai neurodsl-manifest
python -m omniforge_ai nexusflow-example npu_workflow_demo.nxf --pipeline build

Agentic / workflow

python -m omniforge_ai agentic-plan "benchmark npu workflow and explain the results"
python -m omniforge_ai agentic-execute "benchmark npu workflow and explain the results"
python -m omniforge_ai agentic-export-nxf "say hello and answer briefly"
python -m omniforge_ai agentic-run-nxf "say hello and answer briefly"

Benchmarks / jobs / scheduler

python -m omniforge_ai benchmark-profiles
python -m omniforge_ai benchmark-run --profile fast_local
python -m omniforge_ai benchmark-list --limit 10
python -m omniforge_ai benchmark-trends --suite vision_random --metric infer_ms_stats.avg
python -m omniforge_ai job-run benchmark.profiles
python -m omniforge_ai schedule-create --target-kind alerts.evaluate --params-json "{\"only_enabled\":true}" --interval-sec 300
python -m omniforge_ai scheduler-tick

Alerts / remediations / notifications / health / incidents

python -m omniforge_ai alert-templates
python -m omniforge_ai alerts-evaluate --force
python -m omniforge_ai remediation-templates
python -m omniforge_ai notifications-status
python -m omniforge_ai health-snapshot --profile full
python -m omniforge_ai incidents-status
python -m omniforge_ai incident-create "NPU throughput regression" --severity warning --tags perf,triage

OpsGraph analytics

python -m omniforge_ai opsgraph-snapshot --profile runtime
python -m omniforge_ai opsgraph-watch --profile runtime --compare-with latest
python -m omniforge_ai opsgraph-watch --profile runtime --auto-incident --incident-threshold 80
python -m omniforge_ai opsgraph-watch-list --limit 10
python -m omniforge_ai opsgraph-watch-trends --limit 20
python -m omniforge_ai opsgraph-watch-anomalies --baseline-limit 20
python -m omniforge_ai opsgraph-watch-forecast --limit 30 --horizon 5
python -m omniforge_ai opsgraph-hotspots --limit 10
python -m omniforge_ai opsgraph-components --limit 10 --order-by risk
python -m omniforge_ai opsgraph-path incident:inc_* alert:alert_* --max-depth 6
python -m omniforge_ai opsgraph-blast-radius alert:alert_* --depth 2 --direction out

Reports / artifacts / history

python -m omniforge_ai reports-profiles
python -m omniforge_ai report-generate --profile health_watch
python -m omniforge_ai history-list --limit 20
python -m omniforge_ai artifacts-roots
python -m omniforge_ai artifacts-list --root reports

REST API Overview

Selected endpoint groups (all local, JSON unless noted):

• /api/status
• /api/chat, /api/chat/load
• /api/vision/*
• /api/neurodsl/* and /api/neurodsl/agent-api/*
• /api/nexusflow/*
• /api/agentic/*
• /api/ops/* (generic op catalog/run/submit)
• /api/jobs*
• /api/benchmarks/*
• /api/search/*
• /api/missions*
• /api/scheduler/* and /api/schedules*
• /api/reports/*
• /api/alerts/*
• /api/remediations/*
• /api/notifications/*
• /api/health/*
• /api/incidents/*
• /api/opsgraph/*
• /api/history*
• /api/artifacts/*
• /api/stream/history (SSE)

Example:

Invoke-RestMethod http://127.0.0.1:8787/api/status
Invoke-RestMethod http://127.0.0.1:8787/api/health/status
Invoke-RestMethod http://127.0.0.1:8787/api/incidents/status

Incident / Alert Flow (AutoOps)

Current automation stack supports:

1. alert rules detect regressions/failures/event patterns.
2. remediations can auto-run actions (jobs, reports, scheduler ops, notifications, health snapshots).
3. notifications can fan out alert/health messages to console/history/file/webhook.
4. incidents can auto-open/link incidents from critical alerts and store timelines.
5. incident-capture-bundle links health snapshots and reports to incidents for triage.

Persistent Data (Under out/)

OmniForge stores runtime state locally in omniforge_ai/out/, including:

• history/
• jobs/
• benchmarks/
• schedules/
• reports/
• alerts/
• remediations/
• notifications/
• health/
• incidents/
• missions/
• agentic_nxf_runs/
• runtime/ (server stdout/stderr logs)

Windhawk Notes

• windhawk-generate-mod creates a .wh.cpp mod source and can copy it to Windhawk writable mods.
• serve-windhawk can auto-generate and restart Windhawk, but local Windhawk UI settings may still require manual compile/enable confirmation.

Known Notes / Limitations

• Supermix/vision/NeuroDSL/NexusFlow availability depends on your local repo state and models.
• Vision may use DirectML when available and fall back to CPU otherwise.
• NeuroDSL integration uses subprocess isolation for stability.
• Some features are Windows-specific (Windhawk, DirectML-focused paths).

Developer Notes

• Entry point: omniforge_ai/omniforge_ai/__main__.py
• Main CLI command table: omniforge_ai/omniforge_ai/cli.py
• Flask app/routes: omniforge_ai/omniforge_ai/server.py
• Core orchestration service + ops router: omniforge_ai/omniforge_ai/service.py

License / Usage

This README describes the local integration layer in this folder. Refer to each upstream repo for its own license and model/runtime usage terms.