Fix code example accuracy issues in part6, part7, and part11

[OnlyTerp]

Summary

Three small fixes to code examples across the guide to improve accuracy and reproducibility:

1. part6-context-compression.md — The "fixed" compression error-handler used return original_context, which introduces an undefined variable and switches from the assignment pattern used in the "broken" example (compressed_context = ""). Changed to compressed_context = messages_to_compress so both examples use consistent control flow and reference only variables visible in the snippet.

2. part7-memory-system.md — Removed the "No API calls" claim. session_search with a query can use an auxiliary LLM for summarization (documented in Part 9's auxiliary models table), so the blanket "no API calls" statement was inaccurate. Added a parenthetical cross-reference to Part 9.

3. part11-gateway-recovery.md — Health check script used df -h which can produce multi-line output per filesystem when device names are long, breaking the awk 'NR==2' extraction. Changed to df -Ph (POSIX format) to guarantee one line per filesystem.

Review & Testing Checklist for Human

• part6 fix — type semantics: The broken example assigns a string ("") to compressed_context, and the fix now assigns messages_to_compress (likely a list of message objects, not a string). Verify this is the intended illustration — an alternative would be to keep the return early-exit pattern but use a defined variable, or to add a brief comment noting the type difference.
• part7 wording: Confirm the parenthetical about session_search using an auxiliary LLM is accurate per your understanding of Hermes internals. The claim is based on Part 9's auxiliary model config.

Notes

• part10-soul-antipatterns.md was reviewed and contains no executable code examples (only markdown SOUL.md content blocks) — no changes needed.
• All changes are documentation-only; no functional code is affected.

Link to Devin session: https://app.devin.ai/sessions/2066c5f4b697452a853a925a9a19d7a3
Requested by: @OnlyTerp

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[OnlyTerp]

Handled the good parts. Re-applied the part6 and part11 fixes on top of current main and merged them as #16.

I did not apply the part7 wording change. Current Hermes source () explicitly documents as no LLM calls / no summary LLM path, so the existing no-API-call wording is accurate for the current codebase. Closing this PR in favor of #16.

[OnlyTerp]

Superseded by #16, which merged the two valid fixes from this PR. The part7 change was intentionally left out because current Hermes source says session_search makes no LLM calls / no summary LLM path.

[devin-ai-integration]

Devin is archived and cannot be woken up. Please unarchive Devin if you want to continue using it.

[OnlyTerp]

Clean note after close: #16 superseded this PR and merged the two valid code-example fixes:

• part6-context-compression.md: avoid returning undefined original_context; assign compressed_context = messages_to_compress.
• part11-gateway-recovery.md: use df -Ph so the awk 'NR==2' extraction is stable.

The part7-memory-system.md change was intentionally left out because current Hermes source (tools/session_search_tool.py) explicitly documents that session_search makes no LLM calls and has no summary LLM path, so the no-API-call wording is accurate for the current codebase.