Visualize a graph & share a run report

Problem

You want to see the experience you authored before running it — and after a run, you want to hand a teammate (or pin to a PR) a single artifact that shows what actually happened: which nodes ran, which failed, how long each took, and how many tokens it cost. Reading raw JSONL is not it.

Solution

Two commands, both pure transforms — no API key, nothing to deploy.

bash

# 1. Render the static DAG as a Mermaid flowchart on stdout.
#    Paste it straight into a GitHub README, issue, or PR — Mermaid renders natively.
oe graph examples/review-branch

# Left-to-right layout, written to a file:
oe graph examples/review-branch --lr -o review-branch.mmd

# Or a self-contained HTML page (open in any browser, no server):
oe graph examples/review-branch --html -o review-branch.html

bash

# 2. After a run, turn the event log into a shareable HTML run report.
#    The same DAG, but each node is coloured by its status (success / failed / skipped),
#    plus an events timeline and per-node duration & token counts.
oe run examples/review-branch --input pr_url=https://github.com/org/repo/pull/42
# → run-id: run_abc123

oe inspect run_abc123 --experience examples/review-branch --html -o run_abc123.html

Walkthrough

oe graph loads experience.yaml, validates it, and renders the node graph as a Mermaid flowchart. Nodes are grouped into phase subgraphs, each node kind gets its own shape and colour (tool, agent, skill, dataset, experience, cli-agent), and edges produced by for_each fan-out and conditional when: dependencies are labelled. It never calls an LLM. Default output is stdout (top-down); --lr flips it left-to-right, -o <file> writes to disk, and --html wraps it in a standalone page.

oe inspect <run-id> --html reads the run's .openexpertise/runs/<run-id>.jsonl event log and produces the same DAG — but now each node is coloured by its run status, alongside an events timeline and a per-node table of duration and token usage. Because it is a single self-contained HTML file, you can attach it to a PR, drop it in a ticket, or archive it next to the run directory. Use --experience <path> so the command can locate the run directory when you are not in the experience folder, and --lr for a wide graph.

The two views compose: oe graph is the design-time picture (what the experience can do), and oe inspect --html is the run-time picture (what one execution actually did).

Variations

Pin a report to a PR. In CI, run the experience, then oe inspect <run-id> --html -o report.html and upload report.html as a build artifact (or commit it to a gh-pages-style branch) so reviewers can open the run report directly from the pull request.
Keep diagrams in your README current. Commit the oe graph ... -o experience.mmd output (or a fenced Mermaid block) next to the experience; regenerate it whenever the graph changes so the picture never drifts from the YAML.
Plain stdout for chat. Omit --html / -o and paste the Mermaid block into a GitHub comment or any Mermaid-aware tool.

Visualize a graph & share a run report ​

Problem ​

Solution ​

Walkthrough ​

Variations ​

See also ​

Visualize a graph & share a run report

Problem

Solution

Walkthrough

Variations

See also