fcode-python

Factorial Code — Python

Guidelines for writing Python that runs on Factorial Code. Runtime is Python 3.13. For the platform model (processes, modules, datastore) see fcode-core-concepts.

Aspect	Guideline
Runtime	Python 3.13
Process entry file	main.py
Entry point	def main()
Parameters	fcode.context.parameters
Variables	os.getenv("X") or fcode.env.X
Import a module	fcode.import_module("module-slug")

Gotchas

• Define main() as the entry point, but never call it yourself — Factorial Code invokes it.
• fcode.import_module() names must be hardcoded string literals, never variables: fcode.import_module("shopify-client") ✅, fcode.import_module(name) ❌.
• Datastore stores only strings/numbers — json.dumps objects before set, json.loads after get.
• Use snake_case, not camelCase; follow PEP 8; add type hints where helpful.
• Wrap the main flow in try/except, log the caught error with context via fcode-logs (see Logging), and raise actionable errors. Don't rely on global variables for state — pass it through parameters or return values.
• Never hardcode or log secrets — read them from os.getenv.

Process template

def main():
    parameters = fcode.context.parameters

    # Your code here

    return { "message": "Success!" }

Helpers

import os

# Execution / process / schedule metadata
execution_id = fcode.execution.id
process_id = fcode.execution.process.id
schedule_id = fcode.execution.schedule.id   # when run from a schedule
timezone = fcode.execution.timezone

# Environment variables (secrets/config)
api_key = os.getenv("API_KEY")  # or fcode.env.API_KEY

# Import a Factorial Code module (hardcoded name only)
client = fcode.import_module("module-name")
client_v1 = fcode.import_module("module-name", "v1.0")  # pinned version

# Run another process
fcode.processes.run("process-identifier", options)

Logging

Log through the shared fcode-logs module — level-gated logging inherited by every workspace. It reads the LOG_LEVEL team variable (debug | info | warn | error, default info) and forwards to print (stdout for debug/info, stderr for warn/error), so call sites read like a plain print:

log = fcode.import_module("fcode-logs")

log.info("sync started", process_slug)  # stdout when LOG_LEVEL ≤ info
log.debug(request_payload)               # stdout only when LOG_LEVEL=debug
log.warn("token missing — skipping")      # stderr when LOG_LEVEL ≤ warn
log.error("sync failed", str(err))         # stderr — always emitted

Be verbose. Log the start/end of the main flow, every external call, and each major decision at info; dump payloads and intermediate state at debug (free in production — gated off unless LOG_LEVEL=debug). Always log inside except: record the error with context — what operation, which inputs — at error before re-raising, or at warn for a recovered/skipped path. error is always emitted.

Set LOG_LEVEL=debug in a local or dev workspace to trace a full run; production stays at info. Never log secrets.

Dependencies

External pip packages install automatically — just import them. When the package name differs from the import name, declare it with @add-package:

# @add-package requests
import requests

Datastore & storage

import json, os

# Datastore (strings/numbers only)
fcode.datastore.set("key", "value")
fcode.datastore.set("key", json.dumps({ "name": "John", "age": 30 }))
value = fcode.datastore.get("key")
fcode.datastore.delete("key")

# Storage (files)
local_path = os.path.join(os.environ.get("TMP_DATA_DIR"), "localfile.txt")

with open(local_path, "rb") as f:
    obj = fcode.storage.upload("path/myfile.txt", f)

objects = fcode.storage.list()

content = fcode.storage.download("path/myfile.txt")
with open(local_path, "wb") as f:
    f.write(content)

fcode.storage.delete("path/myfile.txt")

Local disk: write temp files under os.environ.get("TMP_DATA_DIR").

Form file uploads: a form file field ("ui:widget": "file") is uploaded to Storage before execution and arrives as an fcode.storage://… reference (a list if multiple files). Strip the prefix to download:

parameters = fcode.context.parameters
uploaded_path = parameters.get("inputFile")
content = fcode.storage.download(uploaded_path.replace("fcode.storage://", ""))

Variables & schedules

Read/write team variables and manage process schedules at runtime — scoped to your own team, no API token needed (like datastore/storage):

# Team variables (config/secrets)
fcode.variables.set("API_KEY", "secret", sensitive=True)
v = fcode.variables.get("API_KEY")  # TeamVariable or None
all_vars = fcode.variables.list()
fcode.variables.delete("API_KEY")

# Schedules (cron or one-off date_time) for a process
schedule = fcode.schedule.create(
    "my-process",
    cron="0 0 6 * * SUN",  # or: date_time="2026-04-24T12:30:00.000"
    parameters={"foo": "bar"},
)
schedules = fcode.schedule.list(process_id=fcode.execution.process.id)
fcode.schedule.pause(schedule["id"])
fcode.schedule.resume(schedule["id"])
fcode.schedule.delete(schedule["id"])
# delete every schedule for a process (pass the process UUID)
fcode.schedule.delete_for_process(fcode.execution.process.id)

fcode.variables.set/delete only persist server-side; they are not reflected in fcode.env within the same run (fcode.env is a snapshot taken at start).

Sending email

Send email with the built-in fcode.send_mail — no SMTP setup required. The mail server and credentials live in the executor manager, never in your process.

info = fcode.send_mail(
    to="user@example.com",   # string or list[str]
    subject="Report ready",
    text="Plain-text body",  # provide text, html, or both
    html="<b>HTML body</b>",
)
# info => { "messageId", "accepted", "rejected" }

• The From address is fixed by the platform; a from you pass is ignored.
• Each execution can send up to 3 emails by default; once the limit is reached, further calls throw.
• Locally (fcode run) there is no manager, so the email is logged, not sent.

Return values

# Standard
return { "message": "Success!" }

# Custom HTTP status (webhooks)
return { "status": 404, "body": { "message": "Not found" }, "headers": { "Content-Type": "application/json" } }

# Transient (not persisted in execution results)
return { "transient": True, "data": sensitive_data }