pricetra-pricing / README.md
Myco-Net's picture
Upload MycoNet nutrient dataset (6 files)
9539931 verified
|
Raw
History Blame Contribute Delete
4.04 kB
---
license: cc-by-4.0
task_categories:
- tabular-to-text
- time-series-forecasting
language:
- en
- zh
tags:
- pricing
- saas-pricing
- ai-pricing
- llm-pricing
- competitive-intelligence
- time-series
size_categories:
- 10K<n<100K
multilinguality:
- multilingual
---
# PriceTrail / 价迹 — AI & SaaS Pricing Dataset (derived layer)
**EN.** A continuously-updated open dataset of **AI / SaaS pricing snapshots and
pricing-change timelines**. Each row in `data/price_cards.jsonl` is a structured
pricing observation (tool, vendor, tier, price, billing cycle, source, capture
time); each row in `data/price_changes.jsonl` is a detected pricing change event
between two snapshots (price up/down, tier added/removed, free-tier cut, billing
cycle change, ...). The dataset is regenerated from live pricing pages on a fixed
cadence, so it doubles as a **pricing time-series** for competitive intelligence,
procurement, VC research, and model training.
**中文.** 一份持续更新的开源「AI / SaaS 定价」数据集,包含定价快照卡与定价变更时间线。
`data/price_cards.jsonl` 每行是一条结构化定价观测(工具、厂商、档位、价格、计费周期、
来源、抓取时间);`data/price_changes.jsonl` 每行是两次快照之间检测到的定价变更事件
(涨价 / 降价 / 新增档位 / 下架档位 / 取消免费档 / 计费周期变更 …)。数据集按固定节奏从
真实定价页重新生成,天然是一份「定价时间序列」,可用于竞品情报、采购、VC 研究与模型训练。
> **Freshness / 新鲜度:** regenerated on every run (default cadence: every 12h);
> the dataset always reflects the most recent observed pricing.
## License / 许可
- The **derived pricing observations** (structured price points, change events,
provenance) are released under **CC-BY 4.0** — see `LICENSE`.
- Source pricing pages retain their own terms; attribution to each source is
preserved via `source_url`.
## Non-affiliation / 非附属声明
This dataset is **independent and not affiliated with, endorsed by, or sponsored
by** any vendor, product, or service named within it. All product names, logos,
and trademarks are the **property of their respective owners** and are used here
only to identify the pricing subject of each observation. Prices are observed
facts redistributed under CC-BY 4.0; this dataset makes no claim of
partnership with any company.
## Coverage honesty / 覆盖诚实声明
This release covers the **publicly-parseable pricing pages** that the PriceTrail
pipeline can fetch and parse (AI / SaaS / DevOps / collaboration tools, etc.).
A number of vendors gate their pricing pages behind bots / login walls
(e.g. some LLM providers and design suites); those are **not yet** in this
dataset and are explicitly excluded — this dataset does **not** claim full
vendor coverage. See the per-run source list for the exact coverage of this build.
## Fields / 字段
**price_card** (13 fields): id, tool, vendor, tier, price, currency, billing_cycle, effective_date, source_url, snapshot_ref, captured_at, lang, category.
See `schema.json` → `definitions.PriceCard`. An optional `category` groups
tools by type (e.g. `agent` for AI-agent apps such as Cursor / WorkBuddy / Coze);
rows without a category are omitted (legacy sources).
**price_change** (8 fields): id, tool, change_type, before, after, detected_at, evidence_snapshot_before, evidence_snapshot_after.
See `schema.json` → `definitions.PriceChange`.
**change_type** vocabulary:
- `new`
- `price_up`
- `price_down`
- `free_tier_cut`
- `feature_cut`
- `tier_added`
- `tier_removed`
- `cycle_change`
## Load / 加载示例
```python
from datasets import load_dataset
cards = load_dataset("Myco-Net/pricetra-pricing", data_files="data/price_cards.jsonl")["train"]
changes = load_dataset("Myco-Net/pricetra-pricing", data_files="data/price_changes.jsonl")["train"]
```
## Stats / 本版统计
- price_card rows: 35147
- price_change rows: 13