feat(tasks): route selected sandbox products to the ai-gateway

[brandonleung]

Problem

Signals' sandbox stages (scout, research, implementation, repo_selection) make model calls through the legacy Python llm-gateway. To move them to the new Go ai-gateway, the sandbox needs the gateway URL and the list of products allowed to use it. Neither is plumbed in today.

Changes

Two new settings: SANDBOX_AI_GATEWAY_URL and SANDBOX_AI_GATEWAY_PRODUCTS. When both are set, they're injected into the sandbox as AI_GATEWAY_URL / AI_GATEWAY_PRODUCTS, and the agent-server (PostHog/code#3659) routes the listed products to the new gateway. Everything else stays on the Python gateway.

The sandbox blocks unknown env vars and network egress by default, so most of the diff is allowlisting:

• The two keys are reserved (user-supplied env can't override them) and added to the agentsh env allowlist. Without the allowlist entry, agentsh strips them and the product silently stays on the old gateway.
• The gateway hostname is added to the network allowlist in both layers (agentsh and Modal): static entries for ai-gateway.us/eu.posthog.com, plus the hostname parsed from the URL setting. The parsed entry covers dev's ai-gateway.dev.posthog.dev.
• Setting values are validated as DNS names via Django's DomainNameValidator, so wildcards, IPs, and malformed labels narrow the policy rather than widening it. Misconfigurations log a warning.
• The session-init egress probe now checks the gateway hosts, so a blocked gateway shows up by name in startup diagnostics.

How did you test this code?

Unit tests pin the env and network allowlists per setting, the hostname validation, the reserved-key guard, and that injection only happens when both settings are set. ruff and repo-wide mypy clean.

Inert until the sandbox chart sets both values; PostHog/code#3659 must deploy first. Rollback: clear either setting.

Refs PostHog/ai-gateway#357.

[greptile-apps]

Prompt To Fix All With AI

Fix the following 1 code review issue. Work through them one at a time, proposing concise fixes.

---

### Issue 1 of 1
products/tasks/backend/logic/services/agentsh.py:24-26
**Configured Gateway Host Remains Blocked**

When `SANDBOX_AI_GATEWAY_URL` uses a valid host other than these two PostHog domains, restricted sandboxes receive the URL but the network policy still denies the connection. This makes the new setting fail for self-hosted or environment-specific gateway deployments unless every caller separately adds the host to `allowed_domains`.

_{Reviews (1): Last reviewed commit: "feat(tasks): route selected sandbox prod..." | Re-trigger Greptile}

[greptile-apps]

Configured Gateway Host Remains Blocked

When SANDBOX_AI_GATEWAY_URL uses a valid host other than these two PostHog domains, restricted sandboxes receive the URL but the network policy still denies the connection. This makes the new setting fail for self-hosted or environment-specific gateway deployments unless every caller separately adds the host to allowed_domains.

Prompt To Fix With AI

This is a comment left during a code review.
Path: products/tasks/backend/logic/services/agentsh.py
Line: 24-26

Comment:
**Configured Gateway Host Remains Blocked**

When `SANDBOX_AI_GATEWAY_URL` uses a valid host other than these two PostHog domains, restricted sandboxes receive the URL but the network policy still denies the connection. This makes the new setting fail for self-hosted or environment-specific gateway deployments unless every caller separately adds the host to `allowed_domains`.

How can I resolve this? If you propose a fix, please make it concise.

[Twixes]

I think that's expected

[github-actions]

🤖 CI report

✅ Playwright — all passed

All tests passed.

View test results →

⚠️ Backend snapshots — 7 updated (7 modified, 0 added, 0 deleted)

Query snapshots: Backend query snapshots updated

Changes: 7 snapshots (7 modified, 0 added, 0 deleted)

What this means:

• Query snapshots have been automatically updated to match current output
• These changes reflect modifications to database queries or schema

Next steps:

• Review the query changes to ensure they're intentional
• If unexpected, investigate what caused the query to change

Review snapshot changes →

⚠️ Backend coverage — 74.0% of changed backend lines covered — 50 uncovered

🧪 Backend test coverage

Patch coverage — changed backend lines (products + core): ███████████████░░░░░ 74.0% (143 / 193)

File	Patch	Uncovered changed lines
products/tasks/backend/temporal/process_task/tests/test_provision_sandbox.py	0.0%	110, 146–147, 151, 154–155, 160, 166, 179–181, 184–185, 187, 190–193, 197, 202, 204–205
products/tasks/backend/temporal/process_task/activities/get_sandbox_for_repository.py	0.0%	244
products/tasks/backend/temporal/process_task/tests/test_utils.py	0.0%	1072, 1076, 1080–1082, 1084, 1088–1091, 1093–1094, 1097–1099, 1101–1105
products/tasks/backend/temporal/process_task/utils.py	20.0%	1197, 1229–1230, 1234
products/tasks/backend/temporal/process_task/activities/provision_sandbox.py	33.3%	168, 344
products/tasks/backend/logic/services/agentsh.py	98.3%	112

🤖 Agents: add a test covering the lines above, or note why under "How did you test this code?". Machine-readable gap list: the patch-coverage artifact on this run (gh run download 30415387339 -n patch-coverage), or the coverage-data block at the end of this comment.

Per-product line coverage (touched products)

Product	Coverage	Lines
platform_features	██░░░░░░░░░░░░░░░░░░ 12.1%	7 / 58
batch_exports	████████░░░░░░░░░░░░ 39.5%	8,593 / 21,741
demo	███████████░░░░░░░░░ 56.2%	1,497 / 2,663
warehouse_sources_queue	████████████░░░░░░░░ 59.2%	148 / 250
tasks	██████████████░░░░░░ 69.7%	31,432 / 45,123
data_tools	██████████████░░░░░░ 70.0%	63 / 90
ai_gateway	███████████████░░░░░ 75.0%	9 / 12
cdp	████████████████░░░░ 81.0%	3,146 / 3,883
signals	████████████████░░░░ 81.1%	23,836 / 29,387
wizard	█████████████████░░░ 84.8%	935 / 1,102
data_modeling	█████████████████░░░ 85.3%	7,343 / 8,611
notebooks	█████████████████░░░ 85.7%	7,525 / 8,785
actions	█████████████████░░░ 86.6%	717 / 828
cohorts	█████████████████░░░ 86.9%	5,648 / 6,496
product_tours	██████████████████░░ 87.9%	1,303 / 1,482
data_warehouse	██████████████████░░ 88.2%	12,137 / 13,757
exports	██████████████████░░ 88.4%	6,950 / 7,862
business_knowledge	██████████████████░░ 89.0%	4,391 / 4,936
engineering_analytics	██████████████████░░ 89.4%	6,441 / 7,202
dashboards	██████████████████░░ 89.5%	5,989 / 6,693
visual_review	██████████████████░░ 89.5%	5,837 / 6,522
conversations	██████████████████░░ 89.5%	17,117 / 19,124
alerts	██████████████████░░ 90.0%	4,342 / 4,827
mcp_analytics	██████████████████░░ 90.2%	2,883 / 3,198
links	██████████████████░░ 90.6%	183 / 202
streamlit_apps	██████████████████░░ 90.7%	2,630 / 2,901
error_tracking	██████████████████░░ 91.0%	10,925 / 12,004
slack_app	██████████████████░░ 91.0%	9,533 / 10,472
marketing_analytics	██████████████████░░ 91.1%	11,905 / 13,074
stamphog	██████████████████░░ 91.1%	4,056 / 4,450
mcp_store	██████████████████░░ 91.9%	4,257 / 4,634
product_analytics	███████████████████░ 92.5%	5,849 / 6,321
managed_migrations	███████████████████░ 92.6%	1,556 / 1,681
early_access_features	███████████████████░ 92.6%	1,287 / 1,390
notifications	███████████████████░ 92.6%	1,017 / 1,098
ai_observability	███████████████████░ 92.8%	15,328 / 16,517
surveys	███████████████████░ 93.1%	5,771 / 6,197
web_analytics	███████████████████░ 93.2%	14,826 / 15,906
posthog_ai	███████████████████░ 93.2%	1,326 / 1,422
approvals	███████████████████░ 93.3%	3,437 / 3,682
reminders	███████████████████░ 93.4%	468 / 501
legal_documents	███████████████████░ 93.8%	1,628 / 1,736
workflows	███████████████████░ 93.9%	6,919 / 7,372
endpoints	███████████████████░ 94.2%	8,655 / 9,192
tracing	███████████████████░ 94.5%	2,670 / 2,826
skills	███████████████████░ 94.6%	3,158 / 3,337
review_hog	███████████████████░ 94.6%	6,912 / 7,303
messaging	███████████████████░ 94.7%	2,885 / 3,048
experiments	███████████████████░ 95.4%	25,415 / 26,637
logs	███████████████████░ 95.5%	10,435 / 10,928
growth	███████████████████░ 96.1%	3,245 / 3,376
annotations	███████████████████░ 96.2%	732 / 761
revenue_analytics	███████████████████░ 96.3%	1,887 / 1,960
replay_vision	███████████████████░ 96.4%	15,773 / 16,367
feature_flags	███████████████████░ 96.4%	17,371 / 18,023
user_interviews	███████████████████░ 96.5%	2,638 / 2,734
access_control	███████████████████░ 96.9%	870 / 898
customer_analytics	███████████████████░ 97.1%	9,798 / 10,086
warehouse_sources	███████████████████░ 97.2%	346,658 / 356,568
data_catalog	████████████████████ 97.7%	2,555 / 2,615
analytics_platform	████████████████████ 98.0%	2,153 / 2,197
metrics	████████████████████ 98.2%	2,491 / 2,536
pulse	████████████████████ 98.4%	2,017 / 2,049
live_debugger	████████████████████ 99.2%	613 / 618
field_notes	████████████████████ 99.4%	158 / 159

Report-only. Patch coverage = changed backend lines covered vs origin/master. Sorted lowest first.
Known gaps: lines covered only by Temporal tests show as uncovered; core line numbers may drift if master changed the same file.

[posthog-bot-comment-resolver]

🔀 Tried to auto-resolve conflicts with master but this one needs a human.

I won't retry until the branch or master moves.

[richardsolomou]

could you also add one activity-level case for get_sandbox_for_repository as well? deleting the environment_variables.update(ai_gateway_env_vars()) wiring there still leaves the helper and Modal provisioning tests green.

[trunk-io]

😎 Merged successfully - details.

[brandonleung]

/trunk merge

[brandonleung]

/trunk merge

[deployment-status-posthog]

Deploy status

Environment	Status	Deployed At	Workflow
dev	✅ Deployed	2026-07-29 03:35 UTC	Run
prod-us	✅ Deployed	2026-07-29 03:49 UTC	Run
prod-eu	✅ Deployed	2026-07-29 03:49 UTC	Run