gdelt-doc-search（gdelt DOC 搜索）

GDELT DOC Search

Core Goal

• Execute atomic DOC API searches against https://api.gdeltproject.org/api/v2/doc/doc.
• Support topic/domain retrieval via explicit query syntax.
• Support both relative windows (timespan) and absolute UTC windows (STARTDATETIME, ENDDATETIME).
• Return structured JSON envelopes and optionally write raw response bytes.
• Keep runtime observable with structured logs and optional log file.

Required Environment

• Configure runtime by environment variables (see references/env.md).
• Start from assets/config.example.env.
• Load env values before running commands:

set -a
source assets/config.example.env
set +a

Workflow

1. Validate effective configuration.

python3 scripts/gdelt_doc_search.py check-config --pretty

1. Run a relative-window DOC search.

python3 scripts/gdelt_doc_search.py search \
  --query '("climate change" OR pollution)' \
  --mode artlist \
  --format json \
  --timespan 1day \
  --max-records 50 \
  --pretty

1. Run an absolute-window timeline query.

python3 scripts/gdelt_doc_search.py search \
  --query '("climate change" OR pollution) sourcecountry:us' \
  --mode timelinevolraw \
  --format json \
  --start-datetime 20260301000000 \
  --end-datetime 20260308000000 \
  --timeline-smooth 5 \
  --pretty

1. Persist raw API payload to a file for downstream tools.

python3 scripts/gdelt_doc_search.py search \
  --query '(wildfire OR drought)' \
  --mode artlist \
  --format json \
  --timespan 1week \
  --output ./data/gdelt-doc/wildfire.json \
  --pretty

Built-in Robustness

• Apply retry with exponential backoff on transient HTTP/network failures.
• Respect Retry-After when present on retriable responses.
• Throttle request frequency with a minimum interval between requests.
• Validate query/time parameter combinations before remote calls.
• Validate DOC constraints (MAXRECORDS<=250, TIMELINESMOOTH<=30).
• Emit JSON results while writing operational logs to stderr and optional log file.

Scope Decision

• Keep only DOC API retrieval in this skill.
• Keep atomic operations only; do not add internal scheduler/polling loops.

References

• references/env.md
• references/gdelt-data-sources.md
• references/gdelt-doc-search.md
• references/gdelt-limitations.md
• references/openclaw-chaining-templates.md

Script

• scripts/gdelt_doc_search.py

OpenClaw Invocation Compatibility

• Keep skill trigger metadata in name, description, and agents/openai.yaml.
• Invoke in prompts with $gdelt-doc-search.
• Keep the skill atomic: one query execution per invocation.
• Use script parameters for retrieval conditions (--query, --mode, --format, --timespan or --start-datetime/--end-datetime).
• If you need polling, let OpenClaw agent orchestrate repeated invocations externally (scheduler/loop), not inside this skill.

OpenClaw Prompt Templates

Use these templates directly in OpenClaw and only replace bracketed placeholders.

1. Recon (config and endpoint check)

Use $gdelt-doc-search.
Run:
python3 scripts/gdelt_doc_search.py check-config --pretty
Return only the JSON result.

1. Search (relative window)

Use $gdelt-doc-search.
Run:
python3 scripts/gdelt_doc_search.py search \
  --query '[QUERY_EXPRESSION]' \
  --mode [MODE] \
  --format json \
  --timespan [TIMESPAN] \
  --max-records [N] \
  --pretty
Return only the JSON result.

1. Validate (absolute window and output persistence)

Use $gdelt-doc-search.
Run:
python3 scripts/gdelt_doc_search.py search \
  --query '[QUERY_EXPRESSION]' \
  --mode [MODE] \
  --format json \
  --start-datetime [YYYYMMDDHHMMSS] \
  --end-datetime [YYYYMMDDHHMMSS] \
  --output [OUTPUT_FILE] \
  --pretty
Check command exit code and bytes_written > 0.
Return JSON plus one-line pass/fail verdict.