Skip to main content

Post-mortem Python trace profiler that emits JSON plus an HTML viewer and OTLP exporter.

Project description

PyTraceFlow.

PyPI Python 3.10+ License Status Downloads

FlowTrace is a trace visualizer designed as a "post-mortem debugger": instead of pausing and resuming, it captures calls (inputs, outputs, caller, module, duration, errors) into a hierarchical JSON so you can inspect them later without re-running.

FlowTrace overview Call details panel

Quick start (3 steps)

  1. Capture the complex sample to JSON: python pytraceflow.py -s samples/complex/complex_app.py -o pft_complex.json
  2. Render the HTML viewer: python pytraceflow_visual.py -i pft_complex.json -o pft_complex.html Call details panel
  3. Optional export to OTLP/Jaeger (HTTP 4318): python export_otlp.py -i pft_complex.json --endpoint http://localhost:4318/v1/traces --service pytraceflow-complex Call details panel Call details panel

Basic flow

  1. Profile a script: python pytraceflow.py -s your_script.py -o pft.json
  2. Generate the viewer: python pytraceflow_visual.py -i pft.json -o pft.html
  3. Open pft.html and browse:
    • Search terms: opens the matching node in a floating panel.
    • Expand/collapse nodes; open calls.
    • Controls to show/hide badges, Python internals, language (es/en), and light/dark mode.

Passing script arguments

  • Any arguments unknown to pytraceflow.py are forwarded to the profiled script (no -- separator required).
  • Example with options: python pytraceflow.py -s samples/basic/basic_sample.py --customer "ana maria" --tier gold --coffee 3
  • Example with positionals: python pytraceflow.py -s samples/basic/basic_positional_sample.py "juan perez" silver 1 2 0 0.18

Features

  • Captures inputs/outputs, caller, module, duration, and errors.
  • Groups instances and nested calls while preserving hierarchy.
  • Search with highlighting and floating panels; option to hide Python internals.
  • Dark mode by default, quick controls, and multi-language.
  • Performance knobs: --flush-interval (seconds, <=0 disables background flush), --flush-every-call (legacy, slower), --log-flushes (stderr).
  • Overhead controls: memoria desactivada por defecto; habilita con --with-memory (usa psutil + tracemalloc), o combina --no-tracemalloc / --no-memory. --skip-inputs evita serializar args/kwargs.
  • Root entry now records total runtime; STDERR line: [FlowTrace] Profiling finished in X.XXXs (script=...).
  • Export existing traces to OTLP/Jaeger via export_otlp.py, with span names enriched by module and instance id to make nested calls distinct in Jaeger UI.

CLI options

  • -s/--script (required): target script path.
  • -o/--output: JSON output path (default pft.json).
  • --flush-interval: seconds between background flushes; <=0 disables thread (default 1.0).
  • --flush-every-call: force flush on every event (slow; legacy).
  • --log-flushes: log each flush to stderr.
  • --with-memory: enable memory snapshots (psutil + tracemalloc). Default is off; expect slower runs when enabled.
  • --no-memory: disable memory snapshots.
  • --no-tracemalloc: keep psutil but skip tracemalloc.
  • --skip-inputs: do not serialize call inputs/outputs.
  • OTLP export (optional, requires opentelemetry-*): --export-otlp-endpoint http://localhost:4318/v1/traces, --export-otlp-service myapp, repeat --export-otlp-header key=value for extra headers.
  • Any other args are forwarded to the profiled script.

Usage examples

  • Default fast run: python pytraceflow.py -s samples/basic/basic_sample.py -o pft.json
  • With memory metrics: python pytraceflow.py -s samples/basic/basic_sample.py --with-memory --flush-interval 2.0
  • Minimal overhead: python pytraceflow.py -s samples/basic/basic_sample.py --flush-interval 0 --skip-inputs
  • Legacy per-call flush with logs: python pytraceflow.py -s samples/basic/basic_sample.py --flush-every-call --log-flushes
  • Memory via psutil only: python pytraceflow.py -s samples/basic/basic_sample.py --with-memory --no-tracemalloc
  • Export to OTLP/HTTP: python pytraceflow.py -s samples/basic/basic_sample.py --export-otlp-endpoint http://localhost:4318/v1/traces --export-otlp-service pytraceflow-sample
  • Export a saved trace to Jaeger (OTLP/HTTP, port 4318): python export_otlp.py -i pft.json --endpoint http://localhost:4318/v1/traces --service pytraceflow-sample
  • Export with custom headers (auth/tenant): python export_otlp.py -i pft.json --endpoint http://localhost:4318/v1/traces --service pytraceflow-sample --header Authorization=Bearer_TOKEN --header X-Tenant=acme

Included examples

  • script.py basic example.
  • complex_app.py with modules demo/... (prices, taxes, discounts).
  • conc_demo.py with CPU-bound (multiprocessing) and IO-bound (threads) to view concurrent traces.
  • basic_positional_sample.py shows the same flow using positional arguments.
  • error_sample.py intentionally fails (missing config) to show the error badge.

Backlog / ideas

  • Filters by module/class/time.
  • Export filtered views.

License

MIT License. See LICENSE for full text.


FlowTrace es un visualizador de trazas de ejecucion, pensado como un "debugger post-mortem": en lugar de parar y reanudar, captura las llamadas (inputs, outputs, caller, modulo, duracion, errores) en un JSON jerarquico para inspeccionarlo despues sin reejecutar.

Vista general de FlowTrace Panel de detalle de llamadas

Inicio rápido (3 pasos)

  1. Capturar el ejemplo complejo a JSON: python pytraceflow.py -s samples/complex/complex_app.py -o pft_complex.json
  2. Generar el visor HTML: python pytraceflow_visual.py -i pft_complex.json -o pft_complex.html Panel visual de llamadas
  3. Exportar a OTLP/Jaeger (HTTP 4318): python export_otlp.py -i pft_complex.json --endpoint http://localhost:4318/v1/traces --service pytraceflow-complex Menu de exportacion a OTLP/Jaeger Spans exportados a Jaeger

Flujo basico

  1. Perfilar un script: python pytraceflow.py -s tu_script.py -o pft.json
  2. Generar visor: python pytraceflow_visual.py -i pft.json -o pft.html
  3. Abrir pft.html y navegar:
    • Buscar trminos: abre el nodo coincidente en panel flotante.
    • Expandir/colapsar nodos; abrir calls.
    • Controles para mostrar/ocultar badges, internals de Python, idioma (es/en) y modo claro/oscuro.

Pasar argumentos al script

  • Cualquier argumento que pytraceflow.py no reconoce se reenvia al script perfilado (no hace falta --).
  • Ejemplo con opciones: python pytraceflow.py -s samples/basic/basic_sample.py --customer "ana maria" --tier gold --coffee 3
  • Ejemplo con posicionales: python pytraceflow.py -s samples/basic/basic_positional_sample.py "juan perez" silver 1 2 0 0.18

Caracteristicas

  • Captura inputs/outputs, caller, modulo, duracion y errores.
  • Agrupa instancias y llamadas anidadas preservando jerarquia.
  • Buscador con resaltado y paneles flotantes; opcion para ocultar internals de Python.
  • Modo oscuro por defecto, controles rapidos y multilenguaje.
  • Ajustes de performance: --flush-interval (segundos, <=0 desactiva flush en background), --flush-every-call (modo anterior, mas lento), --log-flushes (stderr).
  • Controles de overhead: memoria viene desactivada por defecto; --with-memory la habilita (psutil + tracemalloc), combinable con --no-tracemalloc / --no-memory. --skip-inputs evita serializar args/kwargs.
  • La llamada raiz registra el tiempo total; se imprime en STDERR [FlowTrace] Profiling finished in X.XXXs (script=...).
  • Export de trazas existentes a OTLP/Jaeger con export_otlp.py; los spans incluyen módulo e id de instancia para distinguir llamadas anidadas en Jaeger.

Opciones CLI

  • -s/--script (obligatorio): ruta del script a perfilar.
  • -o/--output: ruta del JSON de salida (por defecto pft.json).
  • --flush-interval: segundos entre flushes en background; <=0 desactiva el hilo (por defecto 1.0).
  • --flush-every-call: fuerza flush en cada evento (lento; legado).
  • --log-flushes: loguea cada flush a stderr.
  • --with-memory: habilita snapshots de memoria (psutil + tracemalloc). Por defecto está apagado; al activarlo las ejecuciones serán más lentas.
  • --no-memory: desactiva snapshots de memoria.
  • --no-tracemalloc: deja psutil pero omite tracemalloc.
  • --skip-inputs: no serializa inputs/outputs de las llamadas.
  • Export OTLP (opcional, requiere opentelemetry-*): --export-otlp-endpoint http://localhost:4318/v1/traces, --export-otlp-service miapp, headers extra con --export-otlp-header clave=valor (repetible).
  • Cualquier otro argumento se reenvía al script perfilado.

Ejemplos de uso

  • Ejecución rápida por defecto: python pytraceflow.py -s samples/basic/basic_sample.py -o pft.json
  • Con métricas de memoria: python pytraceflow.py -s samples/basic/basic_sample.py --with-memory --flush-interval 2.0
  • Overhead mínimo: python pytraceflow.py -s samples/basic/basic_sample.py --flush-interval 0 --skip-inputs
  • Flush por llamada con logs: python pytraceflow.py -s samples/basic/basic_sample.py --flush-every-call --log-flushes
  • Solo psutil (sin tracemalloc): python pytraceflow.py -s samples/basic/basic_sample.py --with-memory --no-tracemalloc
  • Export a OTLP/HTTP: python pytraceflow.py -s samples/basic/basic_sample.py --export-otlp-endpoint http://localhost:4318/v1/traces --export-otlp-service pytraceflow-sample
  • Exportar un JSON ya capturado a Jaeger (OTLP/HTTP, puerto 4318): python export_otlp.py -i pft.json --endpoint http://localhost:4318/v1/traces --service pytraceflow-sample
  • Exportar con cabeceras extra (auth/tenant): python export_otlp.py -i pft.json --endpoint http://localhost:4318/v1/traces --service pytraceflow-sample --header Authorization=Bearer_TOKEN --header X-Tenant=acme

Ejemplos incluidos

  • script.py ejemplo basico.
  • complex_app.py con modulos demo/... (precios, impuestos, descuentos).
  • conc_demo.py con CPU-bound (multiproceso) e IO-bound (hilos) para ver trazas concurrentes.
  • basic_positional_sample.py muestra el mismo flujo usando argumentos posicionales.
  • error_sample.py falla adrede (config faltante) para ver el badge de error.
  • error_internal_sample.py falla dentro de una llamada interna (validate_config) para ver errores anidados.

Pendientes / ideas

  • Filtros por mdulo/clase/tiempo.
  • Export de vistas filtradas.
  • Integracin con spans/telemetra.

Licencia

MIT License. Ver LICENSE para el texto completo.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pytraceflow-0.1.5.tar.gz (31.6 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

pytraceflow-0.1.5-py3-none-any.whl (25.2 kB view details)

Uploaded Python 3

File details

Details for the file pytraceflow-0.1.5.tar.gz.

File metadata

  • Download URL: pytraceflow-0.1.5.tar.gz
  • Upload date:
  • Size: 31.6 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.4

File hashes

Hashes for pytraceflow-0.1.5.tar.gz
Algorithm Hash digest
SHA256 9064faedf340e5890c48c0beae825a7918da486ffc59554090941c6d31137afe
MD5 daefd2e9c903c128a78c54d89322d354
BLAKE2b-256 592223e9f1382fc865b90e7b8ea0f3affbd745d73103918db0174a604440ce67

See more details on using hashes here.

File details

Details for the file pytraceflow-0.1.5-py3-none-any.whl.

File metadata

  • Download URL: pytraceflow-0.1.5-py3-none-any.whl
  • Upload date:
  • Size: 25.2 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.4

File hashes

Hashes for pytraceflow-0.1.5-py3-none-any.whl
Algorithm Hash digest
SHA256 a31f98556e223e6cc0f7b97afb878be77265ea006abf096e3a5c3c04a7fce4a9
MD5 637d32e4b28c5213923f0b3b8cec1511
BLAKE2b-256 afaa39e60ad21b3319ad99b0e404e79c69e2f8610c765a3240de73ee0a9ab4f9

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page