Generate Reports#

This guide explains how to use oxo-flow's reporting system to produce structured HTML, JSON, Markdown, and PDF reports from workflow executions.

Overview#

Reports are built from execution truth: run a workflow first, then the report shows what actually happened — expanded commands, per-rule exit codes, real files with recorded checksums, and failure diagnosis for failed runs. Without a checkpoint the report shows template-level data only, and never claims a workflow ran when it didn't.

Reports are assembled from pluggable sections (key-value pairs, tables, narrative text) — see reporting-system for the architecture.

Basic Report Generation#

Generate a report from a workflow file:

# HTML report (default; format also inferred from the -o extension)
oxo-flow report pipeline.oxoflow

# JSON report (the versioned data model behind every view)
oxo-flow report pipeline.oxoflow -o report.json

# Markdown report (git-friendly GFM tables)
oxo-flow report pipeline.oxoflow -f md -o REPORT.md

# Write to a specific file
oxo-flow report pipeline.oxoflow -f html -o results/report.html

Reproducible CI output: --ci pins the generation timestamp (SOURCE_DATE_EPOCH when set) and all lists are sorted, so identical state yields byte-identical reports — safe to diff and commit.

Report Contents#

A generated report includes:

Section	Contents
Dashboard	Pipeline status (honest: "No execution data" when nothing ran), task count, total runtime
Execution Status	Per-rule status, wall time, and exit codes (requires a checkpoint)
Failure Diagnosis	Failed rules: exit code, affected downstream rules, stderr excerpt, suggested next step (only when rules failed)
Clinical Compliance	Static capability statement — only for explicitly clinical workflows or when listed in `[report].sections`; generates no clinical data
Workflow Information	Name, version, total rules, detected domain, `[config]` variables
Commands	The commands that actually ran (expanded); declared templates with an explicit fallback label when no execution record exists
File Manifest	Real files: checkpoint-recorded inputs (path, size, mtime) and outputs (sha256 + on-disk size/mtime)
Metrics	QC metrics parsed from real tool outputs under the workdir (fastp, flagstat, STAR, featureCounts, bcftools, kraken2), one subsection per tool × sample, with Pass/Warn/Fail flags — hidden when nothing parses
Sample Matrix	Rule × sample status grid from the checkpoint's expanded instance names — hidden without a checkpoint or sample definitions
Resource Accounting	Cluster accounting imported from an sacct CSV via `--acct` (Rule/State/Elapsed/CPU Time/Max RSS)
Environment	Engine version, platform, and the environments the workflow's rules declare
Provenance	Engine version, workflow file sha256, checkpoint location
Task Summary	Per-rule table of tasks, types, inputs, outputs, environments, and resources

Sections adapt to available execution data — for example, Execution Status and Failure Diagnosis only appear with a checkpoint — and every section, including Task Summary, can be filtered via [report].sections. List them all with oxo-flow report WF --list-sections.

Report Configuration in `.oxoflow`#

Add a [report] section to your workflow file to filter report sections:

[report]
sections = ["universal", "workflow-info", "commands", "failure-diagnosis"]

Note — partial support: of the three fields, sections filters which registered sections the report includes, by generator name (universal, execution-status, failure-diagnosis, clinical-compliance, workflow-info, commands, file-manifest, environment, metrics, sample-matrix, provenance, task-summary). template is consumed (see below); format is parsed but still unsupported: setting it makes the command warn (or fail under --strict) instead of silently ignoring it — output formats are selected via the CLI -f flag.

Fields#

Field	Type	Description
`template`	String	Report template: the built-in name `"report.html"`, or a template file path (resolved relative to the workflow file's directory first, then the current directory). Applies to HTML output only; a render failure warns and falls back to the default renderer (exit 2 under `--strict`)
`format`	Array	Parsed but not supported yet — setting it warns (or fails under `--strict`); use `-f` to select the output format
`sections`	Array	Sections to include, by generator name (see above)

Custom templates#

[report].template wires a custom Tera template into HTML rendering:

[report]
template = "report.html"        # the built-in template (the default)
# template = "my_report.tera"   # or a file: workflow dir first, then cwd

The scaffolded report-template.tera (from oxo-flow report --init-template) is a good starting point. The built-in template IS report.html; custom template files whose names do not end in .html/.htm/.xml (like the scaffold) are registered under custom.html so Tera HTML-escapes {{ variables }} exactly like the built-in template. With -f json/-f md/-f pdf the template is skipped with a note — it applies to HTML output only.

HTML Reports#

HTML reports are self-contained single-file documents with embedded CSS (dark mode, print styles, accessible landmarks). User-controlled strings — workflow and rule names, commands, file paths — are HTML-escaped, so reports are safe to open and share. They can be opened in any browser:

oxo-flow report pipeline.oxoflow -f html -o report.html
open report.html   # macOS
xdg-open report.html   # Linux

JSON Reports#

JSON reports are the same data model, machine-readable, with schema_version, a command: "report" tag, and provenance fields (workflow_path, checkpoint_path):

oxo-flow report pipeline.oxoflow -f json -o report.json
# stdout stays pipe-safe: diagnostics and --ai output go to stderr
oxo-flow report pipeline.oxoflow -f json | jq .schema_version

Example output structure:

{
  "schema_version": 1,
  "command": "report",
  "title": "my-pipeline Report",
  "generated_at": "2026-08-13T03:47:43Z",
  "workflow_name": "my-pipeline",
  "workflow_version": "1.0.0",
  "checkpoint_path": "/path/to/.oxo-flow/checkpoint.json",
  "workflow_path": "/path/to/my-pipeline.oxoflow",
  "sections": [
    {
      "title": "Workflow Information",
      "id": "workflow-info",
      "content": {
        "type": "KeyValue",
        "pairs": [
          ["Name", "my-pipeline"],
          ["Version", "1.0.0"],
          ["Total Rules", "7"],
          ["Detected Domain", "Generic"]
        ]
      },
      "subsections": []
    }
  ],
  "metadata": {}
}

PDF Reports#

PDF reports are converted from the HTML report via the external wkhtmltopdf binary:

brew install wkhtmltopdf   # macOS
apt install wkhtmltopdf    # Linux

oxo-flow report pipeline.oxoflow -f pdf -o report.pdf

If wkhtmltopdf is not installed, the command degrades gracefully: it writes a printable HTML file (report.html) and warns, instead of failing. Note the wkhtmltopdf project upstream has been archived — HTML output is the portable, supported-by-default format.

You can also output the raw wkhtmltopdf command for scripting or custom processing:

oxo-flow report pipeline.oxoflow -f pdf-command

R-friendly Export (`--r-data`)#

--r-data <DIR> writes two TSV files for downstream R analysis (in addition to the report itself):

oxo-flow report --run results/experiment1 --r-data analysis/
# analysis/sample_table.tsv   sample  group
# analysis/metrics.tsv        rule  wall_time_secs  max_memory_mb  status

sample_table.tsv maps every sample to its group (from [[sample_groups]] and [[pairs]]); metrics.tsv has one row per rule with wall time, peak memory, and status (success / failed / -). Without a checkpoint the files carry headers only, with a note explaining why. Values are TSV-sanitized (tabs and newlines replaced) so names can never break the column layout.

Automatic Snapshots#

Every oxo-flow run / oxo-flow resume automatically writes a JSON report snapshot after the run — no reporting step needed:

.oxo-flow/reports/report-<UTC timestamp>.json — the full report data model (a -N suffix when two snapshots land in the same second)
.oxo-flow/reports/index.json — a JSON array of {generated_at, workflow, checkpoint, report} entries, sorted by generated_at; the last entry is the newest snapshot

Disable the behavior with --no-report-snapshot (on run or resume). Snapshot failures are warnings — a reporting hiccup never fails a run. report's own auto-discovered output lands in the same directory.

Programmatic Report Generation#

You can also generate reports programmatically using the core library:

use oxo_flow_core::report::{Report, ReportSection, ReportContent};

let mut report = Report::new("My Report", "pipeline", "1.0.0");

report.add_section(ReportSection {
    title: "Summary".to_string(),
    id: "summary".to_string(),
    content: ReportContent::KeyValue {
        pairs: vec![
            ("Total Samples".to_string(), "24".to_string()),
            ("Pass Rate".to_string(), "95.8%".to_string()),
        ],
    },
    subsections: vec![],
});

let html = report.to_html();
let json = report.to_json().unwrap();
let markdown = report.to_markdown();