Add v2 artifact metadata schema

[Pigbibi]

Summary

• accept signal schema v2 with required model_version and scoring_version
• keep schema v1 signals readable with unchanged default validator behavior
• emit versioned metadata and expiry for generated theme momentum snapshots

Validation

• python3 -m pytest -q (42 passed)
• python3 scripts/validate_latest_signal.py examples/latest_signal.example.json
• git diff --check

Research-only scope; no execution, portfolio, account, or production dependency changes.

[github-actions]

🤖 Codex PR Review

✅ Merge allowed: blocking findings were cleared by independent Codex arbitration

⚖️ Codex Review Arbitration

✅ clear: The prior blocker is demonstrably fixed by the cumulative diff. src/research_signal_context_pipelines/theme_momentum.py still emits schema v2, but the PR now adds a public theme-momentum compatibility validator, validate_theme_momentum_snapshot(), that explicitly accepts both schema_version "1" and "2", requires the new metadata only for v2, and is re-exported from src/research_signal_context_pipelines/__init__.py. The contract is then exercised in tests/test_theme_momentum.py, which validates a newly built v2 snapshot and a legacy v1 snapshot derived from it after removing expires_at, model_version, and scoring_version. That directly disproves the earlier claim that there were no accompanying reader/validator updates or compatibility tests for theme-momentum artifacts.

ℹ️ Other Findings

1. 🟡 [MEDIUM] Logic in src/research_signal_context_pipelines/theme_momentum.py

validate_theme_momentum_snapshot() only checks that mode and artifact_type keys exist; it never verifies that they are the expected theme momentum values. A different artifact payload with those fields present can therefore pass validation and be misclassified as a valid theme momentum snapshot, leading downstream readers to consume incompatible data. (line 273)

Suggestion: Validate mode == "theme_momentum_snapshot" and artifact_type == THEME_MOMENTUM_ARTIFACT_TYPE inside validate_theme_momentum_snapshot() before accepting the payload.

2. 🟡 [MEDIUM] Reliability in src/research_signal_context_pipelines/theme_momentum.py

generated_at is accepted as any non-empty string instead of a real ISO-8601 timestamp. This means corrupted metadata such as "generated_at": "yesterday" will be treated as valid, defeating the validator's role as a guardrail for stale/invalid snapshot metadata. (line 282)

Suggestion: Parse generated_at with the same datetime validation used elsewhere (for example, the ISO datetime helper used by validate_signal) and reject malformed timestamps.

---

Review by Codex PR Review bot • PR