AI/ML PyPI Package Tracker — Transformers, LangChain, Torch | $1/1K No Auth

Used by MLOps teams and AI-infrastructure/dev-tooling analysts tracking release cadence and dependency health across the model-serving and LLM-orchestration ecosystem, data scientists exploring the Python AI tooling landscape, and AI agents that need package metadata for tool selection mid-conversation.

Pricing: $1.00 per 1,000 packages. Download stats from pypistats.org are included at no extra cost.

Scrape Python package metadata and download statistics from PyPI — the official Python package index. Fetches version, author, license, homepage, project URLs, required Python version, and monthly download counts via the official PyPI JSON API. No proxy, no authentication, no HTML parsing. Pay per result.

Python is the language of AI — every ML framework, data tool, and model serving library lives on PyPI. Use this actor to track the core AI/ML Python stack. Note that packageNames still accepts an arbitrary list of packages; the AI/ML packages are a curated prefill, not hardcoded scraping logic.

What you get per package

Field	Example
package_name	transformers
version_latest	4.41.2
summary	State-of-the-art Machine Learning for JAX, PyTorch and TensorFlow
author	The Hugging Face team
license	Apache-2.0
home_page	https://github.com/huggingface/transformers
project_urls	{"Documentation": "...", "Source": "..."}
requires_python	>=3.8.0
keywords	language model, transformers, huggingface
downloads_last_month	1563278363
release_date	2026-05-14T19:25:26.443000Z
parse_confidence	1.0
warnings	[]
scraped_at	2026-05-31T12:00:00Z

AI/ML PyPI Package Tracker vs competitors

	This actor	epctex/pypi-scraper	trudax/pypi-packages
Data source	Official PyPI JSON API	HTML scraping	Unknown
Proxy required	No — $0	Yes (external proxy)	Unknown
Download stats	Yes (pypistats.org)	No	No
parse_confidence	Yes	No	No
Batch lookup	Yes	No	No
License field	SPDX + expression	Partial	No

How to use AI/ML PyPI Package Tracker

Package name lookup

Add package names to packageNames. Each name fetches full metadata from pypi.org/pypi/<name>/json.

Works with:

• AI/ML packages: transformers, torch, langchain, sentence-transformers, openai, anthropic
• Scoped/prefixed packages: any valid PyPI name

Note: transformers (https://pypi.org/project/transformers), sentence-transformers (https://pypi.org/project/sentence-transformers), anthropic (https://pypi.org/project/anthropic), torch, langchain, and openai are verified real, official, actively maintained AI/ML packages, checked 2026-07-15 via pypi.org.

A note on search

PyPI does not have an official search JSON API. The XML-RPC search endpoint was deprecated in 2023. This actor supports only direct name lookup — which is the most reliable and fastest approach. If you need to discover package names, use pypi.org/search first, then feed the names to this actor.

Download statistics

With includeDownloads=true (default), the actor fetches last-month download counts from pypistats.org — the official PyPI download statistics service. A 1.5-second delay is added between calls to respect the rate limit. Set includeDownloads=false to skip this step and speed up large batches.

Input configuration

{
  "packageNames": [
    "transformers",
    "torch",
    "langchain",
    "sentence-transformers",
    "openai",
    "anthropic"
  ],
  "maxItems": 50,
  "includeDownloads": true
}

Field	Type	Default	Description
packageNames	array	[]	PyPI package names to look up (curated AI/ML preset prefilled, but accepts any arbitrary list)
maxItems	integer	50	Max total records (0 = no limit)
includeDownloads	boolean	true	Fetch download stats from pypistats.org

How AI/ML PyPI Package Tracker works

1. For each package name, fetches https://pypi.org/pypi/<name>/json — the official PyPI JSON API.
2. Extracts and normalizes all metadata fields to a flat schema.
3. If includeDownloads=true, fetches https://pypistats.org/api/packages/<name>/recent for download counts.
4. Emits one record per package with a parse_confidence score (1.0 = all key fields present).

parse_confidence explained

Every record includes a parse_confidence score (0.0–1.0):

• 1.0 — all key fields (name, version, summary, release_date) are present.
• < 1.0 — one or more fields were missing; check warnings for details.

This field lets you filter out incomplete records and detect API changes before they silently corrupt your pipeline.

Pricing examples

Run	Items	Cost
10 packages (with download stats)	10	~$0.01
100 packages	100	~$0.10
1,000 packages	1,000	~$1.00
Daily requirements.txt audit (50 pkgs × 30 days)	1,500/mo	~$1.50/mo

Download stats from pypistats.org are included at no extra cost.

Frequently asked questions

Does this require a proxy?

No. PyPI and pypistats.org are fully public APIs with no access restrictions. Zero proxy cost for the buyer.

How fast does it run?

Without download stats: ~2–5 packages/second. With includeDownloads=true: ~0.5 packages/second (1.5s delay per pypistats call to avoid rate limiting).

What is license_expression?

Modern PyPI packages use SPDX expressions like BSD-3-Clause AND MIT instead of a single license name. This actor normalizes both: if license is null, it falls back to license_expression. The license output field contains whichever is available.

What happens if a package isn't found?

The actor logs a warning, skips the package, and continues. Missing packages are reported in the OUTPUT key-value store (failedDetails list).

Version monitoring

Schedule daily runs on your requirements.txt package list to catch upstream version changes before they break your environment. Especially valuable for MLOps teams tracking release cadence and dependency health across the model-serving and LLM-orchestration ecosystem — torch, transformers, and other foundational packages update frequently.

Use with AI agents (MCP)

This actor is callable as a tool by AI agents (Claude Desktop, Cursor, VS Code, n8n, LangGraph, CrewAI, or any MCP-compatible client) via Apify's hosted Model Context Protocol server. An agent uses it to look up PyPI package metadata mid-conversation — e.g. "what's the latest version of transformers and does it support Python 3.12?" or "compare the license of torch vs langchain".

Point your MCP client at this single tool:

{
  "mcpServers": {
    "apify": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.apify.com/?tools=bovi/ai-ml-python-package-tracker",
        "--header",
        "Authorization: Bearer <YOUR_APIFY_TOKEN>"
      ]
    }
  }
}

Minimal call an agent can make:

{ "packageNames": ["torch", "langchain", "transformers"] }

Reliability for agents: data comes from the official PyPI JSON API (no HTML scraping), so rows don't break on site redesigns. Every record includes a parse_confidence score and a warnings array for machine-readable quality signals. No API key needed inside the tool — auth is your Apify token in the client config above.

Integrations

Built for MLOps teams, dev-tooling analysts, and AI-infrastructure buyers auditing dependencies and tracking the AI tooling ecosystem — the JSON/dataset output drops into the tools you already run, no glue code:

• n8n / Make / Zapier — trigger a run or pipe every new dataset item into 500+ apps (Google Sheets, Airtable, Slack, HubSpot, your database) with no code: n8n, Make, Zapier.
• Webhooks — fire your own endpoint the moment a run finishes, to push results straight into your pipeline (docs).
• MCP server — expose this actor as a tool to Claude, Cursor, or any MCP client so an AI agent can pull this data mid-conversation (guide).
• API & SDKs — fetch the dataset as JSON, CSV, or Excel through the Apify REST API or the Python / JS SDKs.

See all Apify integrations.

Legal disclaimer

This actor uses only official, public APIs provided by the Python Software Foundation (pypi.org) and pypistats.org. It does not scrape HTML or access any authentication-gated or access-controlled data. All data accessed is publicly available without login. Not affiliated with the Python Software Foundation or pypistats.org.