renderflow 2026.8

Generic workflow app runtime and Streamlit renderer with provider plugins

renderflow

Workflow runtime and rendering API for:

• Streamlit UI
• CLI execution
• HTML report export
• Individual figure export

Core Idea

renderflow owns the interface contract and rendering behavior. Provider packages should mostly define workflows (run_workflow + params), not custom UI/CLI plumbing.

Demo App

Live Streamlit demo:

https://demo-for-renderflow.streamlit.app/

CLI

List installed providers:

renderflow list-providers

List provider workflows:

renderflow list-workflows --provider crsd-inspector

Show interpreted workflow parameters:

renderflow show-params --provider crsd-inspector --workflow signal_analysis

Provider-scoped CLI (no provider cli.py needed):

[project.scripts]
crsd-inspector = "renderflow.cli:main"

With that entrypoint:

• crsd-inspector -h shows available workflows.
• crsd-inspector range_doppler_processing -h shows only that workflow's parameters/defaults/help.

Execute a workflow in terminal mode:

renderflow execute \
  --provider crsd-inspector \
  --workflow signal_analysis \
  --param crsd_directory=examples \
  --param prf_hz=1000 \
  --output terminal

Execute and export both:

• one combined report file (--html)
• per-figure files (--save-figures-dir + --figure-format)

renderflow execute \
  --provider crsd-inspector \
  --workflow signal_analysis \
  --param crsd_directory=examples \
  --html output/report.html \
  --save-figures-dir output/figures \
  --figure-format html

Add per-figure JSON in the same run:

renderflow execute \
  --provider crsd-inspector \
  --workflow signal_analysis \
  --param crsd_directory=examples \
  --html output/report.html \
  --save-figures-dir output/figures \
  --figure-format html \
  --figure-format json

Export multiple figure formats in a single run:

renderflow execute \
  --provider crsd-inspector \
  --workflow signal_analysis \
  --param crsd_directory=examples \
  --save-figures-dir output/figures \
  --figure-format html \
  --figure-format json

Comma-separated format lists are also accepted:

renderflow execute \
  --provider crsd-inspector \
  --workflow signal_analysis \
  --param crsd_directory=examples \
  --html output/report.html \
  --save-figures-dir output/figures \
  --figure-format html,json

If --figure-format is omitted, per-figure export defaults to html. Image formats (png, jpg, jpeg, svg, pdf) require Kaleido. renderflow includes kaleido as a dependency.

Launch Streamlit:

renderflow run --provider crsd-inspector

Shell Completion

Tab completion is supported via argcomplete for both:

• renderflow ...
• provider-scoped commands (for example crsd-inspector ...)

Activate in Bash for current shell:

eval "$(register-python-argcomplete renderflow)"
eval "$(register-python-argcomplete crsd-inspector)"

After activation, workflow subcommands and options complete with Tab.

Workflow Result Contract

Use renderflow.workflow.Workflow inside provider workflows:

from renderflow.workflow import Workflow

workflow = Workflow(name="My Workflow", description="...")
workflow.params = {
    "threshold": {
        "type": "number",
        "default": 0.5,
        "label": "Threshold",
        "description": "Minimum score to keep",
    },
}

workflow.add_text("Summary text")
workflow.add_table("Metrics", {"name": ["a"], "value": [1]})
workflow.add_plot(fig, title="Spectrum", figure_id="spectrum", save=True)
workflow.add_code("print('debug')", language="python")
return workflow.build()

add_plot(..., save=False) marks a plot as not exportable when using figure-save operations.

Minimum return contract from run_workflow(...):

• must return a dict
• either:
  • modern shape: {"results": [ ... ]}
  • legacy shape: {"text": [...], "tables": [...], "plots": [...]}
• for modern shape, each item in results must be a dict with:
  • type in text | table | plot | code
  • if type == "plot", item must include figure

Provider Contract Options

1) Explicit AppSpec (fully explicit)

Entry point:

[project.entry-points."renderflow.providers"]
my-provider = "my_provider.app_definition:get_app_spec"

get_app_spec() returns renderflow.contracts.AppSpec.

2) Auto-Defined Provider (minimal)

Entry point (no local app_definition.py or renderflow.py shim required):

[project.entry-points."renderflow.providers"]
my-provider = "renderflow.autodefine:auto_build_app_spec"

If no app_definition exists, renderflow auto-builds from:

• <provider>.workflows.* modules with run_workflow(...)
• optional <provider>.renderflow module:
  • APP_NAME = "..."
  • WORKFLOWS_PACKAGE = "provider.custom_workflows" (optional)
  • optional custom metadata constants for provider setup

Workflow parameters are pulled from:

1. workflow.params if a workflow object exists in the module
2. PARAMS module global
3. inferred function signature defaults

This lets packages like crsd-inspector keep only workflow definitions and optional init logic, while renderflow handles CLI + Streamlit parameter interpretation and rendering.