Perplexity SEC Query Patterns

Reference patterns for .claude/agents/perplexity-sec.md.

Core Rule

Use only:

python3 $CLAUDE_PROJECT_DIR/.claude/skills/earnings-orchestrator/scripts/pit_fetch.py --source perplexity --op search --search-mode sec ...

Return the wrapper JSON envelope directly.

PIT Mode (Historical)

python3 $CLAUDE_PROJECT_DIR/.claude/skills/earnings-orchestrator/scripts/pit_fetch.py \
  --source perplexity --op search --search-mode sec \
  --query "AAPL 10-K risk factors FY2024" \
  --pit 2025-02-01T00:00:00-05:00 \
  --max-results 10

Open Mode (Live)

python3 $CLAUDE_PROJECT_DIR/.claude/skills/earnings-orchestrator/scripts/pit_fetch.py \
  --source perplexity --op search --search-mode sec \
  --query "AAPL 10-K risk factors FY2024" \
  --max-results 10

Date Range

python3 $CLAUDE_PROJECT_DIR/.claude/skills/earnings-orchestrator/scripts/pit_fetch.py \
  --source perplexity --op search --search-mode sec \
  --query "AAPL 10-K risk factors FY2024" \
  --date-from 2024-01-01 --date-to 2024-12-31 \
  --max-results 10

Supported Filings

10-K (annual), 10-Q (quarterly), 8-K (current), S-1/S-4 (IPO/M&A)

Content Extraction

Each search result includes a snippet field with extracted content from EDGAR filings.

• Default extraction: ~4K tokens per result, ~10K total.
• Use --max-tokens <N> (up to 1000000) and --max-tokens-per-page <N> for deeper extraction.
• For full filing content, hand off URLs to neo4j-report (if the filing is in Neo4j) or use a dedicated content extraction agent.

Notes

• Uses --search-mode sec to target SEC EDGAR specifically.
• pit_fetch.py normalizes each result with:
  • available_at (date-only -> start-of-day NY tz)
  • available_at_source: "provider_metadata"
• Authentication is handled by pit_fetch.py via .env (PERPLEXITY_API_KEY).
• In PIT mode, date-only items from the PIT day or later are excluded (conservative).
• Response is JSON-only (data[], gaps[]) for deterministic hook validation.