SRE Assessment

Market Data Freshness

Improve the provided Python CLI starter so it can evaluate freshness for three fictional feed consumers: `alpha`, `bravo`, and `charlie`. The system reads local JSON state, reports whether each feed is alive and safe to serve, blocks unsafe work when data is stale or unknown, and records a repeatable recovery path.

The input data is already provided in `starter/data/healthy.json`. `alpha`, `bravo`, and `charlie` are fixture feeds, not real systems. The task is to design the local reliability behavior around that fixture: classify freshness, fail closed, recover, and prove the result.

Assignment model

Stack: Python 3 CLI, standard library `argparse`, JSON state files, `unittest`
Tooling hint: `uv run python -m ...` is fine if you use uv locally
Starter path: `starter/freshness`, `starter/data/healthy.json`, `starter/tests`
Feeds: `alpha`, `bravo`, and `charlie` are fictional market-data consumers
Input data: `starter/data/healthy.json`; scenario commands write `/tmp/bravo-stale.json` and `/tmp/bravo-recovered.json`
Fixture schema: `max_age_seconds`, `feeds[].name`, `feeds[].alive`, `feeds[].last_tick_age_seconds`
Design target: A local reliability contract: status, safe-to-serve decisions, fail-closed blocking, recovery, tests, and CI
Implementation rules: Do not use real exchanges, broker APIs, streaming systems, cloud services, or external market data

Execution path

Run the starter, create stale data, recover, and submit evidence

01Open `starter/`; the codebase is a Python CLI package named `freshness`.
02Run the starter commands from `starter/`.
03Run `python3 -m unittest discover -s tests`.
04Run `python3 -m freshness smoke` against the bundled fixture data.
05Create stale `bravo` data with `python3 -m freshness scenario stale --feed bravo --output /tmp/bravo-stale.json`.
06Run `python3 -m freshness status --state /tmp/bravo-stale.json` and show only `bravo` is unsafe.
07Recover `bravo` with `python3 -m freshness recover --feed bravo --state /tmp/bravo-stale.json --output /tmp/bravo-recovered.json`.
08Add GitHub Actions that runs environment checks, unit tests, and smoke.
09Submit command output, the GitHub Actions run, the incident note, AI usage, and tradeoffs.

Acceptance gates

Fresh-checkout setup works.

Unit tests and smoke run locally.

GitHub Actions runs environment checks, tests, and smoke.

Stale feed behavior is detectable.

Unsafe work fails closed.

Recovery is reproducible.

Constraints are respected.

A short incident note is present.

Window

6 calendar days

6-8 focused hours total

Hard gates

acceptance gates

Checklist

6 evidence outputs

Source package

README.mdCANDIDATE_RUNBOOK.htmlHEALTH_MODEL_GUIDE.htmlEVIDENCE_AND_VALIDATION_GUIDE.htmlSUBMISSION_TEMPLATE.md

Starter runtime

Python CLI in starter/
starter/data/healthy.json
unittest coverage

Failure scenario

Write /tmp/bravo-stale.json
Block unsafe bravo actions
Keep alpha and charlie eligible

Evidence packet

Command output
GitHub Actions run
Incident response
Tradeoffs and AI usage

Instruction packet: README.md, CANDIDATE_RUNBOOK.html, HEALTH_MODEL_GUIDE.html, EVIDENCE_AND_VALIDATION_GUIDE.html, SUBMISSION_TEMPLATE.md. Produced files are evidence outputs, not source documents. This page does not track progress.

Assignment overview

Assignment checklist

README order

Hard gates for proceeding

Proceed-or-stop evidence from the public README.

SRE

What to showImplementersre-001

Fresh-checkout setup is the first hard gate for proceeding.

cd starterpython3 -m freshness status --state data/healthy.json

Setup commandfresh checkout
Show the setup command and expected output from a clean clone.
```
cd starter
python3 -m freshness status --state data/healthy.json
```
Hard gateproceeding check
Fresh-checkout setup must work without hidden services or prior local state.

Scored depth

Evidence quality, judgment, and tradeoffs from the report.

Follow-up interview

Discussion topics after the submitted work is reviewed.

Requirement	Gate	What to submit	Required proof
Hard gates for proceeding Proceed-or-stop evidence from the public README.
SRE-001Fresh checkout setup A fresh clone can be set up without hidden local state.	Hard gate	fresh-checkout setup evidence	Command proofREADME.md
SRE What to showImplementersre-001 Fresh-checkout setup is the first hard gate for proceeding. `cd starterpython3 -m freshness status --state data/healthy.json` Setup commandfresh checkout Show the setup command and expected output from a clean clone. `cd starter python3 -m freshness status --state data/healthy.json` Hard gateproceeding check Fresh-checkout setup must work without hidden services or prior local state. Evidence expectation Setup output is captured and can be repeated from a new clone. Verification note Evidence must be command-backed, local-safe, and short enough for a third party to verify without guessing what was run.
SRE-014Unit tests and smoke run locally Tests and a healthy smoke check pass on the local machine.	Hard gate	local test and smoke evidence	Command proofREADME.md
SRE-040GitHub Actions baseline checks GitHub Actions must run environment checks, tests, and smoke.	Hard gate	GitHub Actions workflow	Workflow proofREADME.md; SUBMISSION_TEMPLATE.md
SRE-022Stale bravo scenario Trigger stale data for bravo while alpha and charlie remain eligible.	Hard gate	stale-feed evidence in submission report	Scenario proofREADME.md; SUBMISSION_TEMPLATE.md
SRE-023Unsafe work fails closed Stale or unknown data blocks unsafe downstream work.	Hard gate	blocked unsafe-action evidence	Fail-closed proofREADME.md; HEALTH_MODEL_GUIDE.html
SRE-031Recovery workflow Recover the stale feed through a command or workflow another engineer can repeat.	Hard gate	recovery evidence in submission report	Recovery proofREADME.md; EVIDENCE_AND_VALIDATION_GUIDE.html
SRE-048Constraints are respected The submission avoids real secrets, paid infrastructure, cloud requirements, and broad platform buildout.	Hard gate	local-safe implementation and report notes	Scope proofREADME.md
SRE-052Incident first 30 minutes Respond to the 09:37 stale bravo incident with concrete operator decisions.	Hard gate	incident note in the final report or separate file	Incident noteREADME.md; SUBMISSION_TEMPLATE.md
Scored depth Evidence quality, judgment, and tradeoffs from the report.
SRE-060Health model depth Separate process liveness from safe-to-serve eligibility for all three feeds.	Evidence	status output and supporting explanation	Model fieldsREADME.md; HEALTH_MODEL_GUIDE.html
SRE-064Submission evidence packet Package command output, CI links, alert rationale, bad-config handling, and AI disclosure.	Evidence	final submission report sections	Evidence packetREADME.md; SUBMISSION_TEMPLATE.md
Follow-up interview Discussion topics after the submitted work is reviewed.
SRE-077Follow-up discussion Prepare to explain the model, alert thresholds, rollout safety, and production tradeoffs.	Follow-up interview	follow-up interview topics	Discussion prepREADME.md