feat: enable simplified BYC two-sided market flow

[hoffmang9]

Summary

• implement true two-sided BYC<>wUSDC.b execution by carrying explicit buy/sell side through strategy planning, daemon execution, and cloud-wallet offer creation
• simplify BYC market config to fixed_quote_per_base: 0.999, enable the market, and set config-driven 1-buy/3-sell size-10 targets with small buffers
• preserve current zero-fee CAT split behavior and add clear side/pricing/inventory hook boundaries for future advanced model integration

Test plan

• PATH="$(pwd)/.venv/bin:$PATH" pytest -q tests/test_config_models.py tests/test_daemon_strategy_integration.py tests/test_manager_post_offer.py tests/test_daemon_offer_execution.py tests/test_strategy.py
• PATH="$(pwd)/.venv/bin:$PATH" pre-commit run --all-files

Made with Cursor

[ibutterbot]

PR #54 Review - feat: enable simplified BYC two-sided market flow

Scope reviewed

• PR metadata, commit history, and full diff across 10 changed files.
• Implementation paths touched in greenfloor/daemon/main.py, greenfloor/cli/manager.py, greenfloor/config/models.py, and related tests.
• Behavioral focus areas: two-sided action planning, buy/sell offer leg inversion, cloud-wallet reservation/bootstrapping, and config validation shifts.

Findings (ordered by severity)

1) High: buy-side bootstrap wait tracks the wrong asset ID

Where: greenfloor/cli/manager.py, _ensure_offer_bootstrap_denominations()

The function correctly switches the split target asset to quote-side for buy actions (split_asset_id = resolved_quote_asset_id), but _wait_for_mempool_then_confirmation(...) is still called with asset_id=resolved_base_asset_id.

Why this matters:

• For buy-side bootstrap, new split outputs are quote-asset outputs, not base-asset outputs.
• Waiting on base asset can produce false timeout/failure (bootstrap_wait_failed) even if the split succeeded.
• That failure path enables fallback behavior and could trigger unnecessary/redundant cloud-wallet splitting, extra latency, and confusing operator guidance.

Suggested fix:

• Pass asset_id=split_asset_id to _wait_for_mempool_then_confirmation(...) so wait logic tracks the same asset that was split.

Suggested test:

• Add a buy-side bootstrap test where resolved_base_asset_id != resolved_quote_asset_id, assert wait is invoked with quote asset ID and that success/failure behavior maps correctly.

---

2) Medium: side attribution can silently drift to sell when execution metadata falls out of history window

Where: greenfloor/daemon/main.py, _active_offer_counts_by_size_and_side() and _recent_offer_metadata_by_offer_id()

Side/size are recovered from recent audit events with a hard cap (limit=1500). For active offer IDs lacking recent execution metadata, side defaults to "sell":

• metadata = metadata_by_offer_id.get(offer_id)
• side = metadata[1] if metadata is not None else "sell"

Why this matters:

• In long-running deployments with high churn, older still-active buy offers can be misclassified as sell.
• Misclassification skews counts_by_side, which drives replenishment planning in _evaluate_two_sided_market_actions().
• Outcome can be systematic buy over-posting (buy undercount) or sell under-posting.

Suggested mitigations:

• Persist side in offer state at creation/post time, and prefer state lookup over event reconstruction.
• At minimum, treat unknown side as unmapped (not "sell"), increment a dedicated metric, and avoid directional assumptions.

Suggested tests:

• Simulate active buy offers with missing execution metadata and verify they do not silently count as sell.
• Add a test that asserts unknown-side handling behavior and resulting planned actions.

---

3) Medium: stable-quote validation now accepts invalid strategy fields silently

Where: greenfloor/config/models.py, _validate_strategy_pricing()

For quote_asset_type == "stable", validation now skips:

• strategy_target_spread_bps
• strategy_min_xch_price_usd
• strategy_max_xch_price_usd

The new test intentionally confirms acceptance of values like 0, -1, and "invalid".

Why this matters:

• This is likely intentional for minimal stable configs, but it broadens accepted config surface significantly.
• Invalid fields can sit undetected and later become runtime hazards if code paths evolve or if market type/quote type is changed.

Suggested mitigation:

• Keep fields optional for stable quotes, but if present, still validate type/range (or emit warnings) instead of fully bypassing checks.

Suggested tests:

• Add explicit tests for "stable + field present but invalid" behavior based on desired policy (warn vs error vs ignore).

---

4) Low: test file contains accidental repeated assignment noise

Where: tests/test_manager_post_offer.py

prog.home_dir = str(tmp_path) is repeated multiple times in one test setup block, which appears accidental and reduces readability.

Impact:

• No behavior change, but increases review noise and maintenance friction.

Suggested cleanup:

• Collapse duplicated assignments to a single line per setup.

Style and maintainability notes

• The side-normalization helper pattern (_normalize_offer_side) is consistently applied and improves guardrails.
• Two-sided action planning extraction (_evaluate_two_sided_market_actions) is clean and easier to reason about than inlined branching.
• Some naming remains sell-centric in helper APIs (plan_bootstrap_mixed_outputs(... sell_ladder=...)) even when used for buy-side quote splitting; consider renaming for clarity in a follow-up.

Security and performance considerations

• Security: No new obvious secret-handling or auth-surface regressions in this PR. Main risks are correctness and operational reliability rather than exploitability.
• Performance: Additional event reconstruction for side metadata is linear over recent events and likely acceptable now, but reliance on capped history can degrade correctness under high throughput (see finding Fix/ci run sim harness #2).

Tests to add (priority list)

1. Buy-side bootstrap wait uses quote asset ID (regression test for finding Fix live testnet workflow secret gating #1).
2. Two-sided active counting with missing side metadata does not default to sell silently.
3. Reservation/bootstrapping end-to-end buy-side integration with distinct base/quote asset IDs.
4. Config validation policy tests for stable quote markets with present-but-invalid strategy fields.

Overall summary

This PR makes meaningful progress toward true two-sided market flow and includes useful test updates, but there is one high-confidence correctness bug in buy-side bootstrap waiting plus a medium-risk side-accounting robustness gap under event-history truncation. I would not merge as-is without fixing the bootstrap wait asset mismatch and adding at least one targeted regression test for buy-side bootstrap tracking.

[hoffmang9]

Addressed the review feedback with follow-up commit e2a6bd2.

Changes made

1. Buy-side bootstrap wait asset fix (high)

   • In _ensure_offer_bootstrap_denominations(...), wait tracking now uses split_asset_id instead of always resolved_base_asset_id.
   • This ensures buy-side quote-asset splits are waited/confirmed against the correct asset.

2. Unknown side metadata no longer defaults to sell (medium)

   • In _active_offer_counts_by_size_and_side(...), active offers missing recent execution metadata are now counted as unmapped (and skipped for side buckets) instead of implicitly treated as sell.
   • This prevents directional skew in two-sided replenishment logic when metadata is absent.

3. Stable quote validation policy tightened (medium)

   • _validate_strategy_pricing(...) now validates strategy spread/price-band fields when present, including stable-quote markets.
   • This avoids silently accepting invalid config values.

4. Test/readability follow-ups

   • Added regression tests for:
     • buy-side bootstrap wait tracking quote asset,
     • unknown-side active-offer accounting behavior,
     • stable-quote strategy-field validation when fields are present.
   • Removed repeated prog.home_dir = str(tmp_path) setup noise in manager tests.

Validation

• pytest -q tests/test_manager_post_offer.py tests/test_daemon_offer_execution.py tests/test_config_models.py (200 passed)
• pre-commit run --all-files (all hooks passed)

[ibutterbot]

Thanks — bootstrap wait changes look solid (nice: explicit bootstrap_wait_failed + buy waits on quote asset, and tests cover it).

One remaining correctness gap re: side accounting: _normalize_offer_side() returns 'sell' for anything except explicit 'buy', and _recent_offer_metadata_by_offer_id() normalizes item.get('side') through that. So malformed/missing side inside an existing metadata item still becomes sell (sell-biased fallback). That can undercount buys / skew two-sided planning over time.

Also noticed manager-written strategy_offer_execution items in _build_and_post_offer_cloud_wallet(...) appear to omit 'side'; if those are later consumed for accounting, they’ll hit the same sell-default path.

Request before full approve:

• Make side parsing fail-closed (buy|sell|unknown/None) and treat unknown as unmapped (not sell).
• Persist side consistently in strategy_offer_execution items when known.
• Add a regression test for malformed/absent side in metadata staying unmapped + a test that manager audit items include side.

[hoffmang9]

Follow-up comment addressed in commit 845b38c.

What changed

• Fail-closed metadata side parsing

  • Added strict side parsing for audit metadata (buy|sell|None).
  • Malformed/missing side in metadata is now treated as unknown and counted as unmapped, not as sell.

• No sell-biased fallback in side-aware counting

  • _active_offer_counts_by_size_and_side(...) now requires valid side metadata to attribute directional counts.
  • Unknown-side active offers remain visible via unmapped counters and do not skew buy/sell planning.

• Manager now persists side in execution audit items

  • _build_and_post_offer_cloud_wallet(...) now writes side into strategy_offer_execution.items and normalizes action_side once for consistency across phases.

Tests added

• malformed/missing side metadata stays unmapped (tests/test_daemon_offer_execution.py)
• manager strategy execution audit items include side (sell and buy paths) (tests/test_manager_post_offer.py)

Validation

• pytest -q tests/test_manager_post_offer.py tests/test_daemon_offer_execution.py (164 passed)
• pre-commit run --all-files (all hooks passed)

[ibutterbot]

Re-review after your follow-up (commit 845b38c): this addresses the remaining side-accounting concerns I called out.

• Side parsing is now fail-closed (buy|sell|None) and malformed/missing side stays unmapped (no sell-biased default).
• Side-aware counting requires valid side metadata; unknown-side offers don’t skew buy/sell buckets.
• Manager now persists side into strategy_offer_execution items, and you added targeted regression tests for both the daemon metadata path and the manager audit path.

Assuming CI passes the full suite, this looks good to me.