Levandor

Bot Fully pmv2-Native — Last v1 Consensus Code Deleted

3 min read

The Telegram bot and CLI self-status are now fully pmv2-native: the v1 “consensus alignment” view per operator position was replaced with a pmv2 cohort-overlap view, and the LAST v1 consensus code was deleted from the codebase. Shipped 2026-06-03 on main as commit ec75491 (net −156 lines). This is the final code-side step of the v1 decommission.

For Agents

  • Old behavior: /wallets (bot) + CLI self-status showed per-position v1 consensus alignment, reading frozen v1 tables pm_selected_wallets / pm_safe_bets / pm_trader_metrics.
  • New behavior: per position, how many edge-cohort traders hold the same (condition_id, outcome_index) plus the conviction-weighted bet_score, reusing the bet-link scoring path.
  • New type CohortOverlap + overlap_by_position() in pmv2/scoring.rs; cohort_holders_for_conditions() in pmv2/repo.rs (mirrors load_bet_holders, batched WHERE condition_id = ANY($1)).
  • Shared helper fetch_self_overlap_sections feeds BOTH bot and CLI (DRY). Best-effort: a lookup error degrades to “no overlap”, never fails the command.
  • Render: render_overlap_cell (Telegram) + a plain-text CLI cell.
  • v1 prod tables left DORMANT in DB (not dropped) per standing “don’t drop pm_* prod tables” rule.

What Changed (commit ec75491)

Replaced — v1 consensus view → pmv2 cohort overlap

  • /wallets (bot) and CLI self-status previously rendered a v1 “consensus alignment” per operator position, reading the frozen v1 tables.
  • Now they render a pmv2 cohort-overlap view: for each operator position, how many edge-cohort traders hold the same (condition_id, outcome_index), weighted by the conviction bet_score (reuses the existing bet-link scoring path — no parallel scoring logic).

Added

  • CohortOverlap type + overlap_by_position() in pmv2/scoring.rs.
  • cohort_holders_for_conditions() in pmv2/repo.rs — mirrors load_bet_holders, batched via WHERE condition_id = ANY($1).
  • fetch_self_overlap_sections — single shared helper feeding bot and CLI (DRY). Best-effort by design: a lookup error degrades gracefully to “no overlap” and never fails the command.
  • render_overlap_cell (Telegram) + a plain-text CLI cell.

Deleted — the entire v1 consensus chain (−156 net lines)

  • ConsensusAlignment
  • fetch_consensus_alignment
  • classify_consensus
  • fetch_safe_bets_by_market
  • fetch_raw_watched_consensus
  • the alignment render helpers

grep confirms zero live consensus references remain.

Code-only, DB untouched

  • pm_selected_wallets / pm_safe_bets / pm_trader_metrics left dormant (not dropped), per the standing rule.

v1 Decommission — Overall State

With ec75491, v1 is fully removed from CODE. The removal happened in three stages:

  1. Earlier hard-cut — removed the pipeline / pm-arb / consensus engine.
  2. first-mover-overlap-watcher (commit 4794a76) — removed change_detection / participant_changes.
  3. This commit (ec75491) — removed the last consensus DISPLAY.

Only dormant pm_* tables remain in the DB.

Related OPS INCIDENT — Zombie Bot Container (2026-06-03)

Zombie container — two bots running concurrently

Two polymarket-fetch bot containers were found running at once on the GCP VM:

  • one on NEW code (overlap_watcher cursor advancing), and
  • one on OLD pre-4794a76 code (participant_changes cursor STILL advancing every minute — a detector that was DELETED in 4794a76).

The old container was spamming the operator with a v1 “watched wallets position dump”.

Diagnosis

Via pm_detection_cursors: a detector that only exists in old code was still advancing → an old container is still live. Same root cause class as the Round-50 Telegram 409 getUpdates conflict (two pollers sharing one bot token).

Fix

VM-side: kill the stale container, leave exactly one on HEAD. Escalated to @deploy via AGENT_HANDOFF Round 52 / 52b.

GOTCHA for future deploys

After a bot deploy, verify only ONE container runs AND check that old-code-only detector cursors FREEZE. A still-advancing cursor for a deleted detector is the signature of a zombie container.

Deploy / Verification

  • Deploy target is now ec75491 (Round 52b).
  • 135 tests pass.
  • clippy clean (3 pre-existing warnings only).
  • fmt clean.

Graph View

Backlinks