feat(OMN-13161): drop hardcoded delegate max_tokens constant; resolve from routing contract

[jonahgabriel]

OMN-13161 — drop hardcoded delegate max_tokens constant (CLI side)

Ticket: OMN-13161 — "Delegation max_tokens belongs in the routing contract (per-backend, overlay-overridable) — not a CLI constant".
Companion PR (omnimarket, routing-authority resolution + per-backend bifrost_delegation.yaml max_tokens + node contract): OmniNode-ai/omnimarket#1230

This PR implements ticket item 4 (the cli_delegate.py CLI side) plus the
parallel skill_mapping.yaml delegate entry that carried the same hardcoded
default.

Problem

onex delegate injected a hardcoded DEFAULT_MAX_TOKENS = 2048 into the
dispatch payload, and the declarative onex skill delegate route carried a
hardcoded default: 2048 in skill_mapping.yaml. The token budget is
configuration and belongs on the selected backend in the routing contract
(per-backend, overlay-overridable), resolved by the delegate node after
backend selection — not as a CLI constant. The 2048 cap truncated long outputs
(a LinkedList suite was cut off) far below the local model's 128k window.

Change

• src/omnibase_infra/cli/cli_delegate.py
  • Removed DEFAULT_MAX_TOKENS = 2048 (and its __all__ entry).
  • --max-tokens default → None (no show_default); help text updated to
    "optional explicit override; omit to let the delegate node resolve
    per-backend from its routing contract".
  • _write_payload now omits the max_tokens key entirely when it is
    None, so the node resolves the budget per-backend (relies on the
    omnimarket fix(ci): re-trigger auto-merge after review and checks complete [OMN-7786] #1230 resolution seam). Signatures widened to int | None.
  • Docstring/comments updated to drop the "historical delegate-shim default"
    reference; documented the omit-on-None behavior.
• src/omnibase_infra/cli/skill_mapping.yaml
  • Delegate max-tokens arg loses default: 2048. Per ModelSkillArgSpec, a
    None default means the field is omitted from the payload so the node
    supplies its own — same per-backend resolution path.
• Tests updated for the new None-default / omitted-key behavior; the explicit
  --max-tokens override path still asserts the key is present in the payload.

--max-tokens remains an optional explicit override: when passed, the value is
included in the payload exactly as before.

Verification (run in the worktree off origin/dev)

$ uv run ruff format src/ tests/ && uv run ruff check --fix src/ tests/
4112 files left unchanged
All checks passed!

$ uv run mypy src/omnibase_infra/ --strict
Success: no issues found in 2441 source files

$ uv run pytest tests/unit/cli/test_cli_delegate.py tests/unit/cli/test_cli_skill.py \
    tests/unit/validation/test_validator_defaults.py::TestUnionCountRegressionGuard -q
42 passed in 3.67s

$ uv run pytest tests/unit -n auto -q
20489 passed, 15 skipped  (after fix)
# 2 pre-existing/env-only failures NOT introduced by this PR:
#  - test_validation/test_validator_defaults union-count guard: fixed in this PR
#    (typed payload dict as dict[str, object] instead of an inline str | int union).
#  - tests/unit/verification/test_cli.py::TestCLIMain::test_registration_only:
#    fails identically on origin/dev baseline; it asserts exit 1/2 when no runtime
#    DB is configured, but this dev machine reaches the .201 LAN Postgres so the
#    check succeeds (exit 0). Environment-only; passes in CI. Untouched by this PR.

$ pre-commit run --files <changed files>
All hooks Passed (no failures)

The full tests/ run (serial) exceeds the local 10-minute harness window; the
unit suite was run in parallel (-n auto) and is green except for the
documented env-only verification-CLI test. CI is the authority for the full
green gate.

dod_evidence (OMN-13161)

DoD per the ticket: zero hardcoded max_tokens constants in the CLI; --max-tokens
default None (omitted = routing authority decides); optional explicit override
only; no env var introduced. This PR satisfies the CLI-side DoD items:

• DEFAULT_MAX_TOKENS = 2048 removed from cli_delegate.py (grep-verified: no
  remaining DEFAULT_MAX_TOKENS reference in cli_delegate.py).
• --max-tokens default is now None; the key is omitted from the dispatch
  payload when no override is supplied (unit-test asserted, both routes).
• skill_mapping.yaml delegate max-tokens default: 2048 removed.
• No env var introduced for the limit.

The per-backend resolution seam (bifrost_delegation.yaml max_tokens +
node_delegate_skill_orchestrator dispatch resolution) is the companion omnimarket
PR #1230 (ticket items 1-3, 5). This CLI PR relies on that seam: when the CLI
omits max_tokens, the node resolves it per-backend.

Receipt Gate / Evidence-Source

This PR pairs with the OCC evidence record for OMN-13161, authored once the
companion omnimarket PR #1230 is green (the standard cross-repo OCC pairing for
this ticket family — cf. OMN-13159 / OCC#2647 which bound omnimarket #1228 +
omnibase_infra #1989). The Evidence-Source: OCC#<num> line will be added when
the OCC evidence PR is minted. Not faking an OCC number here.

---

Evidence-Ticket: OMN-13161
Evidence-Source: OCC#2651

[coderabbitai]

Warning

Review limit reached

@jonahgabriel, we couldn't start this review because you've reached your PR review rate limit.

More reviews will be available in 2 hours, 5 minutes, and 38 seconds. Learn how PR review limits work.

Your organization has used up its prepaid credits, and credit purchases are no longer available. Enable the review add-on in the billing tab to keep reviews running — you're only billed for reviews past your plan's rate limits ($0.25/file).

⌛ How to resolve this issue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

We recommend that you space out your commits to avoid hitting the rate limit.

🚦 How do rate limits work?

CodeRabbit enforces hourly rate limits for each developer per organization.

Our paid plans include higher PR review limits than trial, open-source, and free plans. In all cases, reviews become available again over time. During sustained high-volume PR review activity, CodeRabbit may temporarily slow when the next review becomes available.

Please see our Fair Usage Limits Policy for further information.

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 516027e0-f1aa-4749-ba1e-4114dc011170

📥 Commits

Reviewing files that changed from the base of the PR and between adaf9ce and 9169012.

📒 Files selected for processing (4)

• src/omnibase_infra/cli/cli_delegate.py
• src/omnibase_infra/cli/skill_mapping.yaml
• tests/unit/cli/test_cli_delegate.py
• tests/unit/cli/test_cli_skill.py

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch jonah/omn-13161-max-tokens-cli

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}