license	cc-by-4.0
language	en fr
size_categories	1K<n<10K
tags	canada government procurement contracts tenders rag llm-training
pretty_name	ProcureData — Canadian Government Procurement (Sample)
configs	config_name data_files default split path train procuredata_sample.jsonl

ProcureData — Canadian Government Procurement (Sample)

Sample of 4,978 Canadian federal and provincial procurement records. Mix of contracts, tenders, awards, ACAN notices, and standing offers. Last 3 months of data from the full ProcureData API.

Want the full dataset? 2.1M+ records, cursor-paginated REST API, daily refresh, query by department / vendor / category / value range. Get an API key →

What's in here

Field	Type	Description
source_id	string	Originating system (canadabuys, seao, ns, ab, yt)
entity_type	string	tender, contract, award, standing_offer, disclosure
record_id	string	Stable per-record identifier from the source
municipality	string	City/region if available
event_date	string (ISO date)	Award date for contracts, publish date for tenders
category	string	Procurement category (GSIN / UNSPSC / source-specific)
department	string	Awarding government department
vendor	string	Vendor as published
vendor_normalized	string	Cleaned vendor name (deduplicated across spellings)
contract_value	float	CAD amount, where disclosed
contract_end_date	string (ISO date)	Contract end where disclosed
data	object	Full raw record from the source (JSON)

Composition (sample)

entity_type	rows
tender	2,700
contract	1,950
award	225
standing_offer	100
disclosure	3

Use cases

• RAG / agents — ground LLM responses in actual Canadian procurement data
• Domain fine-tuning — train models on procurement vocabulary, vendor naming, GSIN codes
• Bid forecasting — historical patterns of department spend
• Vendor intelligence — who wins what, at what value
• Schema exploration before committing to API integration

Limitations of this sample

• Snapshot only — frozen from 2026-02-25 to 2026-05-25. New tenders close daily; this sample will not reflect them.
• 5K rows — full dataset is 2.1M+. Sample is enough to test schema, prompts, prototypes; not enough for production analytics.
• No querying — can't filter server-side, search by full-text, or paginate. Use the API for that.

Production access

If you need any of: fresh data, full coverage, search, filtering, or programmatic access, use the API:

• Docs: https://procuredata.ca/docs
• Pricing: https://procuredata.ca/pricing — free tier 25 req/day, Pro $49/mo 1K req/day
• Quickstart:

  curl "https://api.procuredata.ca/contract?q=cybersecurity&limit=10" \
    -H "x-api-key: YOUR_KEY"

License & attribution

Data sourced from public Canadian government procurement systems (CanadaBuys, SEAO Quebec, Nova Scotia, Alberta, Yukon). All source data is under open government licenses. Sample redistribution is permitted under CC-BY-4.0.

Contact

• API and enterprise inquiries: nicolas@builddata.ca
• Bug reports: https://github.com/NicolasPrimeau/Nimbus (or via the contact link on procuredata.ca)

Mirror locations

The same sample is published in three places — pick whichever you prefer:

• Hugging Face: https://huggingface.co/datasets/nimbusdata/procuredata-canadian-procurement-sample
• Kaggle: https://www.kaggle.com/datasets/niexon/procuredata-canadian-procurement-sample
• GitHub: https://github.com/nimbusdata/procuredata-sample