Remove the OVMS hard memory ceiling

[8Dionysus]

Summary

• replace the OVMS 4 GiB hard ceiling with its existing 1 GiB soft reclaim reservation
• keep CPU/thread tuning and the independent langchain-api cap unchanged
• record the owner-native health, reload, parity, and rollback boundary

Live evidence

• owner-admitted canary raised the running OVMS limit from 4 GiB to 16 GiB without restart
• container ID, PID, start time, and health remained unchanged
• 8/8 embedding probes were accepted and answered with exact vector parity (max_abs_delta=0)
• rejected, failed, cgroup OOM, and host OOM deltas were all zero
• post-canary PSI remained zero; measured usage did not increase from lifting the ceiling

Validation

• python -m unittest tests.test_compose_contracts (6 tests, PASS)
• python scripts/validate_stack.py (PASS)
• python scripts/ci_gate.py --mode source-fast reaches the unrelated existing aoa-kag-mcp/AGENTS.md snippet mismatch; this branch does not touch that file

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3a41d64c73

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 064e11a2f4

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".