What (One-liner)

Canvas sentiment atomic node: blends news NLP + TV rating + CNN F&G → sentiment score [-1, +1] + direction.

How to use

sentiment_factor_v1 (NVDA, blended) → bear_agent_v1 (sentiment weighting)
sentiment_factor_v1 ('', fg) → risk_management_v1 (market-wide F&G as overlay)

Core formulas

See frontmatter. Core: multi-source average, missing sources drop out.

Assumptions & applicability

Assumptions: scrapers stable + LM dict ~80% accuracy + US-only coverage.

Applies to: sentiment-weighted auxiliary signals, event studies.

Does not apply to: real-time seconds-level, high-precision NLP, non-US, deep narrative (institution_agent).

Known limitations

See frontmatter. Six items: scraper fragility / lag / accuracy / US-only / F&G market-wide only / low source_count noise.

References

• Loughran-McDonald (2011) JF — financial-text dictionaries
• CNN Fear & Greed Index

Golden Test

N/A — not byte-deterministic.

Changelog

• 1.0.0 (2026-04-21) — First Active