fix(sdk/python): add AGENTFIELD_SKIP_IP_DETECTION opt-out for container IP probing

[AmirF194]

Root cause

_build_callback_candidates() in sdk/python/agentfield/agent.py calls _detect_container_ip() unconditionally whenever _is_running_in_container() is true, with no way to opt out. _detect_container_ip() sends blocking requests to AWS/Azure/GCP cloud-metadata endpoints (169.254.169.254, metadata.google.internal) and falls back to the third-party api.ipify.org, regardless of whether an explicit AGENT_CALLBACK_URL (or constructor callback_url) is already set. In a Kubernetes deployment with a restrictive NetworkPolicy, this fires on every agent start, flooding deny logs, and it also leaks pod egress to a third party for a best-effort candidate that most deployments never use.

Fix

Add AGENTFIELD_SKIP_IP_DETECTION (default false, same boolean-env-var convention already used by AGENTFIELD_LOG_PAYLOADS/AGENTFIELD_LOG_TRACKING/AGENTFIELD_LOG_FIRE in logger.py) and gate the _detect_container_ip() call behind it. Default behavior is unchanged; the Railway-specific hint and the local-network/hostname/localhost fallbacks are untouched. Documented in docs/ENVIRONMENT_VARIABLES.md.

On the prior closed attempts

Four earlier PRs on this issue (#698, #740, #770, #834) were closed as placeholder tests dropped into a root tests/ directory that isn't part of this repo's layout, without an actual fix. This PR is different: it changes the real behavior in sdk/python/agentfield/agent.py and adds tests in sdk/python/tests/test_agent_networking.py, the file that already covers this exact function, following its existing monkeypatch style.

Verification

• New tests (test_build_callback_candidates_skips_ip_detection_when_disabled, test_build_callback_candidates_runs_ip_detection_by_default, test_ip_detection_disabled_is_case_insensitive, test_ip_detection_disabled_defaults_to_false) fail on unmodified main (AttributeError / probe still called) and pass on this branch, in a clean Docker container.
• Full sdk/python suite (./scripts/run_pytest.sh) green on Python 3.10 and 3.12 (the matrix's low and high ends): 1968 passed, 4 pre-existing skips, no new failures.
• ruff check . on this branch reports the same 9 pre-existing findings as unmodified main, all inside a build/lib/ pip-build artifact directory unrelated to this diff; nothing in the changed files.
• Not verified: an actual Kubernetes NetworkPolicy blocking egress (no cluster in this sandbox). The fix is a control-flow gate on an existing, already-tested function, not a change to its detection logic.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ santoshkumarradha
❌ AmirF194
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[santoshkumarradha]

Thanks for tightening this up. The opt-out gate is small, the networking tests passed locally, and the docs note is clear. The only remaining blocker I see is the CLA check. Once that’s signed, this should be good to go.

[AmirF194]

Thanks for the review. Working on getting the CLA signed, will follow up here once it's cleared.

[github-actions]

Performance

SDK	Memory	Δ	Latency	Δ	Tests	Status
Python	9.0 KB	-	0.28 µs	-20%	✓	✓

✓ No regressions detected

[github-actions]

📊 Coverage gate

Thresholds from .coverage-gate.toml: per-surface ≥ 84%, aggregate ≥ 85%, max per-surface regression ≤ 1.0 pp, max aggregate regression ≤ 0.50 pp.

Surface	Current	Baseline	Δ
control-plane	87.10%	87.40%	↓ -0.30 pp	🟡
sdk-go	92.80%	92.00%	↑ +0.80 pp	🟢
sdk-python	94.20%	93.73%	↑ +0.47 pp	🟢
sdk-typescript	91.17%	90.42%	↑ +0.75 pp	🟢
web-ui	84.76%	84.79%	↓ -0.03 pp	🟡
aggregate	85.62%	85.75%	↓ -0.13 pp	🟡

✅ Gate passed

No surface regressed past the allowed threshold and the aggregate stayed above the floor.

[github-actions]

📐 Patch coverage gate

Threshold: 80% on lines this PR touches vs origin/main (from .coverage-gate.toml:thresholds.min_patch).

Surface	Touched lines	Patch coverage	Status
control-plane	0	—	➖ no changes
sdk-go	0	—	➖ no changes
sdk-python	0	—	➖ no changes
sdk-typescript	0	—	➖ no changes
web-ui	0	—	➖ no changes

✅ Patch gate passed

Every surface whose lines were touched by this PR has patch coverage at or above the threshold.

[AbirAbbas]

Closing this in favour of #969, which lands the same opt-out from a clean-room implementation (written from #624's description, not from this diff) and also skips the probe automatically when a callback URL is already configured — the case pocesar originally reported. The CLA has been pending for ten days and we'd rather not leave #624 open longer; the diagnosis and the env-var approach were yours and #969 credits you for them. If you sign the CLA later we'd be glad to take future contributions.

[AmirF194]

Thanks for the update, and for crediting the diagnosis. Glad to see #969 merged and covering the callback URL case too. Working on getting the CLA signed, would like to contribute again here.