fix: promote _STRATEGY_PARAMS / _COMMON_PARAMS to module level

[keitaj]

Summary

• PR fix: wire JSON risk namespace through to Config (close PR #148 gap) #149 left _STRATEGY_PARAMS and _COMMON_PARAMS as locals inside main(), so validation.strategy_validator.known_market_making_keys() raised ImportError whenever the JSON-config typo detector ran (i.e. on every --config invocation).
• This PR promotes both lists to module scope (right after _RISK_PARAMS) and adds a regression test class that pins the contract.

Why this matters

Without this fix the migration tool added in the farming repo (scripts/env_to_json.py) would generate a working JSON config that crashes the bot the first time it tries to enumerate known MM keys. Only the typo-warning path was affected; without --config (i.e. the existing CLI/env path) the bug was latent.

Changes

• bot.py — move _COMMON_PARAMS and _STRATEGY_PARAMS out of main() to module level. main() now carries a short comment pointing at the module-level definitions; call-sites for _collect_params / argparse.Namespace are unchanged.
• tests/test_strategy_validator.py — new TestKnownMarketMakingKeys class with 6 regression tests that exercise the function and assert the three name lists are importable from bot at module scope.

Test plan

• flake8 . --max-line-length=120 passes
• pytest tests/ -q passes (1050 / 1050)
• End-to-end round-trip verified: scripts/env_to_json.py .env.farming | bot's load_json_configs returns 68 keys with no ImportError and no spurious typo warnings
• No CLI/env regression — existing CLI invocation path is byte-identical

🤖 Generated with Claude Code

[keitaj]

Review: promote _STRATEGY_PARAMS / _COMMON_PARAMS to module level

Tight, surgical fix for the latent ImportError in known_market_making_keys that PR #149 left behind. Pure code-motion (delete locals, add module-level definitions) plus a regression-test class that pins the contract from both sides — direct import and through the public function. End-to-end round-trip with the migration tool was verified before opening the PR. No risk-guardrail or order-logic changes.

---

1. Comment wording in main() ─ slight mismatch (nit)

bot.py L1388-1393:

The comment says "the local references below preserve the existing call-site shape of _collect_params" but after the move the call sites just dereference module globals — there are no "local references" anymore. Minor wording inconsistency that could mislead a future reader into thinking shadowing is still in play.

Suggestion:

# ``_COMMON_PARAMS`` and ``_STRATEGY_PARAMS`` are defined at module
# level (top of this file) so the JSON config layer's typo detector
# (``validation.strategy_validator.known_market_making_keys``) can
# introspect them without circular-import gymnastics. The call sites
# below reference the module-level definitions directly.

Non-blocking.

---

2. Consider an end-to-end typo-warning test (nit)

tests/test_strategy_validator.py:

The new TestKnownMarketMakingKeys covers the symptom (function returns / direct import). One additional integration-style test that drives load_json_configs with a deliberate typo and asserts the warning fires would catch the whole failure mode (loader → validator → bot module) in a single check. Nice-to-have, not required — the current six tests already prevent the specific regression that motivated this PR.

---

Verdict: LGTM

No blockers, no should-fix. Two nits at most. The contract-pinning tests in TestKnownMarketMakingKeys::test_bot_module_exports_param_lists are exactly what was needed — the next time someone is tempted to re-localise these lists, CI will stop them.