Task 02c: subgraph yes_token fix, gamma resolved_at, historical backfill, batch mode

[MaksymDS]

Summary

• fix(subgraph): market filter now passes YES token decimal ID instead of condition ID hex — was causing 0 trades returned for all markets (Fix 1)
• fix(gamma): closedTime → resolved_at and outcomePrices → resolution_outcome now populated on upsert (Fix 2)
• feat(gamma): historical backfill via --closed --end-date-min/max; swept 2020–2026 → 865,789 markets (Fix 3)
• feat(cli): fflow collect subgraph --all-resolved batch mode with --min-volume, --max-volume, --categories, --limit; idempotency skip for already-collected markets (Fix 4)
• fix(gdelt): gdelt_unavailable warning fires only once per process
• test: regression tests for yes_token filter, enrichedOrderFilleds key, resolved_at extraction, _gamma_outcome parsing (28 tests, all green)

Results

See reports/TASK_02C_RESULTS.md for full before/after stats.

DB state after sweep: 911,237 total markets, 865,725 resolved, 852,602 with resolution_outcome.
Subgraph batch running in background: ~90 markets, 700K+ trades so far.

Test plan

• uv run pytest — 89 passed, 2 skipped
• Live subgraph test: 0x63a66ab... → 1,441 trades confirmed (Fix 1 working)
• Gamma sweep: 865,789 markets upserted across 76 month windows

🤖 Generated with Claude Code