Skip to content

presentation-checklist.md

Latest commit

 

History

History
338 lines (243 loc) · 8.16 KB

presentation-checklist.md

File metadata and controls

338 lines (243 loc) · 8.16 KB

Blacknode Presentation Checklist

Use this as the fast live-demo order. It keeps the flow focused on proof: local runtime, visual editor, run replay, NVIDIA workflows, MCP agent control, and self-hosted deployment.

Opening Position

Blacknode is the visual workflow editor for agent stacks. Agents are strong at intent and reasoning; workflow construction needs typed structure, validation, visible state, run history, and exportable artifacts.

Core identity:

Blacknode turns agent intent into typed, visible, runnable workflows.

NVIDIA identity:

Blacknode is the visual workflow editor for the NVIDIA agent stack.

Quick Demo Order

Step Action Proof
1 Open README Short intro, 2x2 image grid, playable demos, walkthrough link.
2 Run blacknode doctor Colored [OK], [WARN], [NOT OK] environment status.
3 Run blacknode demo Local runtime returns Hello World.
4 Start start.bat or ./start.sh Editor opens at http://localhost:3000.
5 Open Templates -> Text Pipeline -> Cook Visible graph returns Hello World.
6 Build Text -> Text -> Concat -> Output by hand Typed handles connect and validate visually.
7 Open Runs after cooking Run history, event timeline, replay, result, errors.
8 Save in Workflows Workflow becomes a reusable JSON artifact.
9 Open NVIDIA Video Intelligence Mission Control No-key video stack plan: input, Cosmos/VLM, NeMo Retriever, NIM/Nemotron QA, deployment, report.
10 Open NVIDIA AI Mission Control No-key NVIDIA workflow plan and readiness surface.
11 Open NVIDIA Local NIM Launch Docker command and local endpoint wiring.
12 Open NVIDIA NIM with NVIDIA_API_KEY Hosted NIM model result and model-call run events.
13 Open NVIDIA NIM Benchmark Text, latency, metrics, and raw samples.
14 Run blacknode mcp Agent can list nodes, inspect schemas, build, validate, run, export.
15 Run streamable HTTP MCP AI-Q, NeMo Agent Toolkit, and HTTP MCP clients can connect.
16 Run Docker Compose Self-hosted editor, backend, run store, and HTTP MCP endpoint.

Terminal Checks

Local Runtime

blacknode doctor
blacknode demo

Expected:

  • Doctor prints colored status labels.
  • Demo prints Blacknode demo OK.
  • Demo result is Hello World.

Validate, Run, Export

Windows:

blacknode validate templates\text-pipeline.json
blacknode run templates\text-pipeline.json
blacknode export-python templates\text-pipeline.json --output workflow.py
python workflow.py

macOS/Linux:

blacknode validate templates/text-pipeline.json
blacknode run templates/text-pipeline.json
blacknode export-python templates/text-pipeline.json --output workflow.py
python workflow.py

Expected:

  • Validation returns "ok": true.
  • Run returns "value": "Hello World".
  • Exported Python prints Hello World.

Editor Checks

Template Run

  1. Start the editor.
  2. Open Templates.
  3. Click Text Pipeline.
  4. Click the Output node.
  5. Click Cook.

Expected:

  • Output node shows Hello World.
  • Runs tab records the execution.
  • Results on unrelated nodes stay visible until those nodes cook again.

Manual Graph Build

  1. Click Clear.
  2. Add Text, Text, Concat, Output.
  3. Set Text values to Hello and World.
  4. Connect Text value -> Concat a.
  5. Connect Text value -> Concat b.
  6. Connect Concat value -> Output value.
  7. Click Organize.
  8. Click Cook on Output.

Expected:

  • Graph lays out cleanly.
  • Typed handles match by color/type.
  • Output shows Hello World.

Workflow Save

  1. Open Workflows.
  2. Name the workflow.
  3. Click Save.
  4. Reopen it from the saved workflow list.
  5. Right-click the row for Insert, Rename, Duplicate, Delete.

Expected:

  • Saved workflow appears under workflows/*.json.
  • Workflow can be reopened or inserted.

Run Replay

  1. Cook any workflow.
  2. Open Runs.
  3. Open the newest run.
  4. Use step/play/scrub controls.
  5. Use Open workflow when a run snapshot is available.

Expected:

  • Node highlights follow the execution timeline.
  • Model calls and tool calls are counted.
  • Result or error is visible in the run record.

NVIDIA Checks

No-Key Planning

Windows:

blacknode run templates\nvidia-ai-mission-control.json
blacknode run templates\nvidia-video-intelligence-mission-control.json
blacknode run templates\nvidia-local-nim-launch.json

macOS/Linux:

blacknode run templates/nvidia-ai-mission-control.json
blacknode run templates/nvidia-video-intelligence-mission-control.json
blacknode run templates/nvidia-local-nim-launch.json

Expected:

  • Mission Control returns an NVIDIA stack plan.
  • Video Intelligence Mission Control returns the folder input, Cosmos/VLM, NeMo Retriever, NIM/Nemotron QA, deployment, and report plan.
  • Local NIM Launch returns Docker command text and endpoint output.

Hosted NIM

Windows:

$env:NVIDIA_API_KEY="your-key"
start.bat

macOS/Linux:

export NVIDIA_API_KEY="your-key"
./start.sh

Editor path:

  1. Open Templates.
  2. Click NVIDIA NIM or NVIDIA NIM MCP Demo.
  3. Confirm the Model node uses a nim: model.
  4. Click Cook on Output.
  5. Open Runs and inspect model-call timing.

Expected:

  • Output shows the NIM model response.
  • Runs panel records model-call events.

Local NIM

  1. Open Templates.
  2. Click NVIDIA Local NIM Launch.
  3. Click Cook.
  4. Copy the generated Docker command.
  5. Start the NIM container in a separate terminal.
  6. Point NIMHealthCheck, NIMAgent, or NIMBenchmark at the local endpoint.

Expected:

  • Local command and endpoint are generated in the graph.
  • Same workflow surface can route to hosted or local NIM.

MCP Checks

Stdio MCP

blacknode mcp

Agent prompt:

Using the blacknode MCP tools, list the available node types, show the schema
for Text, Concat, and Output, then create a workflow that concatenates "Hello"
and " World", validates it, runs it, and exports it as Python.

Expected:

  • Agent lists node types.
  • Agent creates a typed workflow.
  • Validation passes.
  • Run returns Hello World.
  • Export returns Python.

Streamable HTTP MCP

blacknode mcp --transport streamable-http --host 127.0.0.1 --port 9901 --path /mcp

Endpoint:

http://127.0.0.1:9901/mcp

Expected:

  • Same Blacknode MCP tool surface over HTTP.
  • AI-Q, NeMo Agent Toolkit, or another streamable HTTP MCP client can connect.

Custom Node Check

  1. Open Script.
  2. Paste:
from blacknode.node import node

@node(inputs=["text:Text", "n:Int"], outputs=["result:Text"])
def FirstNWords(ctx: dict) -> dict:
    words = ctx.get("text", "").split()
    n = int(ctx.get("n", 10))
    return {"result": " ".join(words[:n])}
  1. Click Run or press Ctrl+Enter.
  2. Add the new Custom node to the canvas.
  3. Connect and cook it.

Expected:

  • Custom node appears without server restart.
  • Typed ports work like built-in nodes.

Tool Workflow Checks

Templates:

  • Python Tool Agent
  • Visual Tool Agent
  • Subnet Tool Call
  • Subnet Tool Agent

Expected:

  • PythonFn exposes Python code as a callable tool.
  • SubnetAsTool turns a visual subgraph into a tool.
  • ToolBox collects tools.
  • ToolCall tests a tool directly.
  • AgentLoop can call tools through the graph.

Docker Check

docker compose up --build

Open:

http://127.0.0.1:3000

Services:

Service Port Purpose
editor 3000 Browser editor.
editor-server 7777 Backend, workflow store, run store, cook API.
blacknode-mcp 9901 Streamable HTTP MCP server at /mcp.

Expected:

  • Self-hosted editor works.
  • Run history persists through the backend.
  • HTTP MCP endpoint is available at /mcp.

Closing Proof

Blacknode provides one visible workflow surface across:

  • local CLI checks
  • browser graph editing
  • typed validation
  • run replay
  • model routing
  • NVIDIA hosted and local NIM flows
  • MCP agent control
  • streamable HTTP MCP
  • Docker Compose deployment
  • Python export