DeprecationWatch

Declare the APIs your project uses. Get alerted before they break.

Open-source API deprecation monitoring for developers who consume third-party APIs — not the teams that publish them.

Dependabot bumps package versions. It doesn't tell you that OpenAI is removing gpt-4-turbo in October or that Stripe deprecated an endpoint your checkout flow still calls. DeprecationWatch closes that gap.

Why this exists

• Dependabot / Renovate — package versions, not API-level deprecations
• Visualping — visual page diffs, not structured developer alerts
• API management platforms — built for API producers, enterprise-priced
• Optic — closest prior art; repository archived January 2026

DeprecationWatch monitors official changelogs, SDK releases, and HTTP Deprecation / Sunset headers — then alerts you at 90, 30, and 7 days before removal.

Dashboard

Local web UI — stat cards, provider health, urgency-sorted timeline:

pip install "deprecationwatch[dashboard]"
dep-watch dashboard
# → http://127.0.0.1:8080

Or from a clone:

git clone https://github.com/arunmm8335/DeprecationWatch && cd DeprecationWatch
pip install -e ".[dashboard]"
dep-watch dashboard

Toggle Show expired to include past removal dates:

Quick start

pip install deprecationwatch

# optional extras
pip install "deprecationwatch[alerts]"      # Slack / email
pip install "deprecationwatch[dashboard]"   # local web UI
pip install "deprecationwatch[all]"         # everything

dep-watch init          # creates depwatch.yml
dep-watch check         # scan live sources

Install from source (latest main):

pip install git+https://github.com/arunmm8335/DeprecationWatch.git

Example output

✓ Stripe - no deprecations found
⚠ OpenAI gpt-4-turbo - DEPRECATED | removal: 2026-10-23 | 116 days left
  → Guide: https://platform.openai.com/docs/deprecations
✓ GitHub - no deprecations found

Configuration

Copy depwatch.yml.example or run dep-watch init:

version: 1
project: my-app

apis:
  - name: OpenAI
    type: openai
    track:
      - gpt-4
      - gpt-4-turbo

  - name: Stripe
    type: stripe
    track:
      - checkout.sessions.create

  - name: Twilio
    type: generic_changelog
    changelog_url: https://www.twilio.com/en-us/changelog
    track_keywords:
      - SMS

alerts:
  slack_webhook: https://hooks.slack.com/services/XXX/YYY/ZZZ
  days_warning: [90, 30, 7]

Supported adapters

Type	Source	LLM needed
openai	OpenAI deprecations page	No
stripe	Stripe changelog RSS	Optional
github_rest	GitHub changelog RSS	Optional
generic_changelog	Any changelog URL	Yes (for diffs)
http_headers	Live endpoint Deprecation / Sunset headers	No

Add more providers — see CONTRIBUTING.md.

HTTP header probes

Many APIs send standard Deprecation and Sunset headers on live responses. Probe endpoints you call in production:

  - name: Example API
    type: http_headers
    track:
      - https://api.example.com/v1/users
    request_headers:
      Authorization: "Bearer ${YOUR_TOKEN}"

Daemon mode

Run checks on a schedule locally (cron alternative):

dep-watch watch --interval 24h
dep-watch watch --interval 30m --once   # single run

Pairs with alerts — get Slack/email when thresholds are crossed without setting up cron.

Alerts

Email and Slack via Apprise. Deduplicated — you won't get the same 30-day warning twice.

pip install "deprecationwatch[alerts]"

dep-watch check --alert-dry-run   # test without sending
dep-watch check                   # sends when thresholds hit
dep-watch check --no-alerts       # skip (for CI)

CI / GitHub Action

Fail the build when a tracked API is within 30 days of sunset:

# .github/workflows/dep-watch.yml
name: API Deprecation Check

on:
  schedule:
    - cron: "0 9 * * 1"
  pull_request:
  workflow_dispatch:

jobs:
  check:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: arunmm8335/DeprecationWatch/action@v1
        with:
          config: depwatch.yml
          fail-on: breaking
          days-threshold: "30"

Or run locally in CI mode:

dep-watch check --fail-on breaking --days-threshold 30

Emits GitHub workflow annotations when running in Actions.

LLM classification

Unstructured changelogs use LiteLLM — bring your own model:

export DEPWATCH_LLM_MODEL=ollama/llama3              # free, local
export DEPWATCH_LLM_MODEL=anthropic/claude-haiku-4-5 # ~$0.001/run
export DEPWATCH_LLM_MODEL=none                         # skip LLM (OpenAI adapter still works)

What we can't detect (yet)

Be honest about limits — PRs welcome:

• Providers that don't publish deprecations anywhere
• Changes buried in blog posts with no changelog
• Undocumented breaking changes with no headers
• False positives from LLM classification on generic changelogs

We optimize for precision over recall. A missed deprecation is bad; a false alarm every week is worse.

Development

git clone https://github.com/arunmm8335/DeprecationWatch
cd DeprecationWatch
python3 -m venv .venv && source .venv/bin/activate
pip install -e ".[dev,alerts,dashboard]"

pytest
dep-watch check --config depwatch.yml.example --fail-on none

Project layout

dep-watch check      CLI — scan & alert
dep-watch watch      Daemon — scheduled checks
dep-watch dashboard  Local web UI
dep-watch report     JSON/table export
action/              GitHub Action composite
src/dep_watch/
  adapters/          One module per API provider
  alerts.py          Apprise + deduplication
  classifier.py      LiteLLM fallback

Roadmap

• Config parser, SQLite storage, OpenAI adapter
• Stripe / GitHub / generic changelog adapters
• Apprise alerts with deduplication
• GitHub Action, local dashboard, expired filter
• HTTP Deprecation / Sunset header monitor
• dep-watch watch daemon mode
• PyPI packaging (pip install deprecationwatch) — see PUBLISHING.md
• Community adapters (AWS, Azure, Razorpay, …)

Contributing

Contributions are welcome — especially new adapters. Read CONTRIBUTING.md for the adapter pattern and PR guidelines.

License

MIT — see LICENSE.