DNCP Paraguay Technology Leads

Turn Paraguay public procurement notices into a clean stream of auditable technology opportunities.

This Actor queries the official DNCP open-data API, prefilters notices, downloads only relevant tender details, applies deterministic scoring, and delivers one structured dataset item per accepted opportunity.

Why use it

Raw procurement feeds are broad. This Actor focuses on opportunities involving:

• software and licenses;
• cybersecurity and network equipment;
• cloud, servers, and data centers;
• IT services and digitalization;
• artificial intelligence and databases;
• telecommunications;
• CCTV and video surveillance.

Every result includes the official notice URL, DNCP API URL, scoring signals, UNSPSC codes, buyer, amount, deadline, and source documents. The classifier is deterministic and auditable: an LLM does not decide which rows you pay for.

Current scope

Version 0.1 is deliberately inexpensive:

• HTTP-only; no browser;
• no proxy;
• no OCR;
• no attachment downloads;
• no LLM cost;
• one sequential DNCP request stream;
• hard input limits for pages and results.

llmEnrichment is reserved for a future Hermes gateway release and must remain disabled in v0.1.

Input

Field	Default	Description
lookbackDays	7	Search window when dateFrom is omitted.
dateFrom / dateTo	current window	Explicit dates in YYYY-MM-DD.
onlyOpen	true	Return open DNCP notices only.
keywords	technology list	Case- and accent-insensitive signals.
unspscPrefixes	["43"]	Technology classification prefixes.
minimumTechScore	40	Minimum accepted score.
maxResults	100	Hard cap on delivered paid rows.
maxPages	20	Hard source pagination cap.
pageSize	50	DNCP search records per page.

Example:

{
  "lookbackDays": 7,
  "onlyOpen": true,
  "minimumTechScore": 40,
  "maxPages": 2,
  "maxResults": 25,
  "keywords": [
    "software",
    "licencia",
    "ciberseguridad",
    "firewall",
    "inteligencia artificial"
  ]
}

Output

{
  "schemaVersion": "1.0",
  "source": "DNCP Paraguay",
  "sourceLicense": "CC-BY-4.0",
  "ocid": "ocds-03ad3f-493234-1",
  "tenderId": "1f1b6719-ea2a-6972-aa82-3de7bcb442e4",
  "title": "ADQUISICIÓN DE LICENCIAS PARA EL RECTORADO DE LA UNA",
  "statusDetails": "En Convocatoria (Abierta)",
  "buyerName": "Rectorado / Universidad Nacional de Asunción",
  "amount": 37594000,
  "currency": "PYG",
  "tenderEndDate": "2026-10-19T08:00:00-04:00",
  "techScore": 85,
  "techSegments": ["software"],
  "matchedSignals": ["unspsc:43", "keyword:licencia"],
  "opportunityUrl": "https://www.contrataciones.gov.py/licitaciones/convocatoria/...",
  "apiUrl": "https://www.contrataciones.gov.py/datos/api/v3/doc/tender/...",
  "enrichmentStatus": "disabled"
}

The STATS key-value-store record reports scanned rows, candidates, detail requests, emitted results, source errors, and truncation.

Pay-Per-Event

The code emits the custom event basic-lead exactly once for each accepted row written to the default dataset. Configure this event in Apify Console before Store publication.

Suggested beta price: $0.01 per delivered lead. With the default maxResults=100, the maximum explicit basic-lead charge for one run would be $1.00, before Apify synthetic platform events and compute costs. Users can set a lower result cap or platform spending limit.

No event is intentionally emitted for:

• search rows rejected by the prefilter;
• duplicate tender IDs;
• closed notices when onlyOpen is enabled;
• details below minimumTechScore;
• DNCP detail requests that fail.

Apify may also record synthetic platform events. Store pricing and spending limits remain controlled by Apify's charging manager.

Scoring

Current deterministic signals:

Signal	Score
Configured UNSPSC prefix	+45
Configured keyword in title	+40

Scores are capped at 100. matchedSignals explains every awarded point. Segments are inferred from the matching keyword and specific UNSPSC family where possible.

Local development

Requirements:

• Python 3.13;
• uv;
• Apify CLI;
• Docker/Colima for container verification.

uv sync --all-groups
uv run pytest -q --disable-socket --allow-unix-socket
uv run ruff check .
uv run ruff format --check .
uv run mypy src
apify validate-schema

Run against DNCP using local Apify storage:

mkdir -p storage/key_value_stores/default
cp tests/fixtures/input_smoke.json storage/key_value_stores/default/INPUT.json

ACTOR_TEST_PAY_PER_EVENT=true \
DNCP_DELAY_SECONDS=4.2 \
APIFY_LOCAL_STORAGE_DIR="$PWD/storage" \
uv run python -m dncp_tech_leads

Build the production image:

$docker build -t dncp-tech-leads:local .

Responsible source usage

The Actor uses the official DNCP API and defaults to a 4.2-second delay between requests in anonymous mode. Keep short date windows and conservative page limits. Do not remove the delay for scheduled or public production runs without authenticated DNCP access and an approved rate policy.

This Actor is a lead-discovery tool, not legal, procurement, or bid advice. Always verify deadlines, requirements, amounts, amendments, and eligibility in the official DNCP notice and documents.

Roadmap

1. authenticated DNCP client and documented retry policy;
2. item-description and category scoring;
3. optional Hermes LLM summaries after deterministic acceptance;
4. attachment requirement extraction;
5. saved searches and webhook alerts.

Data source

• Dirección Nacional de Contrataciones Públicas, Paraguay;
• official DNCP API v3;
• open-contracting records exposed by DNCP.