PR #17141: fix(cli): prevent .env sanitizer from splitting GLM_API_KEY by LM_API_KEY suffix

• Repository: NousResearch/hermes-agent
• Author: jackjin1997
• State: closed | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/17141

Description (problem / solution / changelog)

What does this PR do?

Fix a corruption bug in _sanitize_env_lines where any registered env var name that is a suffix of another (e.g. LM_API_KEY is a suffix of GLM_API_KEY) would cause the longer key's line to be silently split into corrupted halves like G\nLM_API_KEY=.... This rewrote the user's .env file on disk and disabled provider auth for the affected key (Z.AI/GLM in the report; same shape applies to GLM_BASE_URL vs LM_BASE_URL).

The reporter pinpointed the exact mechanism: the splitter walks each .env line scanning for any known-key name as a substring, with no word-boundary check. stripped.find(\"LM_API_KEY=\") returns 1 inside \"GLM_API_KEY=...\" because LM_API_KEY= is a literal suffix of GLM_API_KEY=.

Related Issue

Fixes #17138

Type of Change

• Bug fix (non-breaking change which fixes an issue)

Changes Made

• hermes_cli/config.py — _sanitize_env_lines: collect full needle ranges (start, end) for every known-key match, then drop matches whose range is fully contained within a longer overlapping match. This kills suffix-collision splits while keeping every legitimate concatenation case (where matches do not nest).
• tests/hermes_cli/test_config.py — two new tests under TestSanitizeEnvLines:
  • test_glm_suffix_collision_not_split — direct regression for #17138.
  • test_suffix_collision_does_not_break_real_concatenation — confirms a genuine concat GLM_API_KEY=glmLM_API_KEY=lm-key still splits at the second key, since that match is not nested inside the first.

How to Test

pytest tests/hermes_cli/test_config.py::TestSanitizeEnvLines -v

All 13 tests (11 existing + 2 new) pass. Full `tests/hermes_cli/test_config.py` suite (52 tests) also passes.

Manual repro:

1. Put `GLM_API_KEY=glm-secret` in `~/.hermes/.env`.
2. Run `hermes` (or any code path that triggers `sanitize_env_file`).
3. Before the fix: the file is rewritten as `G\nLM_API_KEY=glm-secret` and `os.getenv("GLM_API_KEY")` returns `None`. After the fix: the file is unchanged.

Checklist

• Tests added and pass locally
• Conventional Commit format
• Single logical change, scope limited to the sanitizer
• No hardcoded `~/.hermes` paths (uses existing `HERMES_HOME` flow)
• No `simple_term_menu` / ANSI erase / cross-toolset schema references
• No `prompt_toolkit` cache-breaking changes

AI Disclosure

This bug was identified and fixed with AI assistance.

Changed files

• hermes_cli/config.py (modified, +14/-5)
• tests/hermes_cli/test_config.py (modified, +17/-0)

PR #17241: fix(config): sanitize_env substring false positive breaks valid API keys

• Repository: NousResearch/hermes-agent
• Author: Feng-H
• State: open | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/17241

Description (problem / solution / changelog)

Problem

_sanitize_env_lines() in hermes_cli/config.py uses str.find() to detect concatenated KEY=VALUE pairs on a single line. This substring search causes false positives when one known env var name is a substring of another:

• LM_API_KEY= matches at position 1 inside GLM_API_KEY=<value>, because find() does not require a word boundary.
• The sanitizer then splits the valid entry into a bare G line and an LM_API_KEY=... line, destroying the key.
• On the next gateway startup, sanitize_env_file() runs again and sees the two fragments as-is (neither is a valid known key), preserving the corruption permanently.

This is a general bug — any pair of known keys where one name contains another as a substring is affected (e.g., API_KEY= inside OPENAI_API_KEY=, ANTHROPIC_TOKEN= inside ANTHROPIC_API_KEY=, etc.). The practical impact is limited because most users do not encounter the initial corruption that triggers the cycle.

Trigger scenario (CJK input methods)

The bug becomes a persistent data-loss issue when combined with CJK input methods (fcitx5, ibus). These can occasionally insert a stray newline in the middle of a pasted env var name, producing:

G
LM_API_KEY=<actual_value>

Once this corruption exists:

1. The old sanitizer has no merge logic — it preserves both fragments as-is.
2. If the user manually fixes the split (e.g., via sed), the sanitizer re-splits it on the next startup due to the substring match described above.
3. This creates a repair loop that makes the .env file unfixable without patching the code.

Fix

Pass 0 (new): Detect and merge split key names. If a line contains only uppercase letters/underscores (no =) and the next non-empty line, when concatenated, forms a known env var name with =, merge them.

Pass 1 (improved): Replace str.find() (arbitrary substring search) with str.startswith() at position 0 + forward offset scanning. This ensures concatenated key detection only matches actual key names at line/segment boundaries, never as substrings inside other key names.

Testing

Verified with these scenarios:

• Split key merge: G + newline + LM_API_KEY=<value> → GLM_API_KEY=<value>
• Normal key preserved: GLM_API_KEY=abc → GLM_API_KEY=abc
• Concatenated keys still split: KEY1=val1KEY2=val2 → two lines
• Comments/blanks preserved
• No false merge on unrelated uppercase text

Changed files

• hermes_cli/config.py (modified, +82/-19)

PR #17273: fix(config): prevent .env sanitizer from corrupting GLM_API_KEY and other suffix-trap keys

• Repository: NousResearch/hermes-agent
• Author: searchonedev
• State: open | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/17273

Description (problem / solution / changelog)

Summary

The _sanitize_env_lines() function in hermes_cli/config.py silently corrupts API keys on every startup. It uses str.find() to detect accidentally-concatenated KEY=VALUE pairs, but since LM_API_KEY is a substring of GLM_API_KEY, scanning GLM_API_KEY=abc123 finds LM_API_KEY= at position 1, triggering a false split into G and LM_API_KEY=abc123.

Because sanitize_env_file() is called unconditionally from migrate_config() on every startup, the .env file gets rewritten with the corruption each time.

Affected key pairs (5 total)

Fix

Pre-compute a "suffix traps" set of known keys that are suffixes of other known keys, and skip them during the scan. This eliminates substring false-positives while still correctly splitting truly concatenated lines.

Testing

• GLM_API_KEY=value → unchanged (was split into G + LM_API_KEY=value)
• LM_API_KEY=value → unchanged (still works standalone)
• HERMES_LANGFUSE_PUBLIC_KEY=value → unchanged
• OPENROUTER_API_KEY=abcGOOGLE_API_KEY=def → correctly split into 2 lines (real concatenation still handled)

Changed files

• hermes_cli/config.py (modified, +9/-1)

PR #17288: fix(cli): prevent .env sanitizer from splitting GLM_API_KEY by LM_API_KEY suffix

• Repository: NousResearch/hermes-agent
• Author: teknium1
• State: closed | merged: True
• Link: https://github.com/NousResearch/hermes-agent/pull/17288

Description (problem / solution / changelog)

Salvage of #17141 by @jackjin1997 onto current main.

Fixes #17138.

_sanitize_env_lines searched every registered env key name as a substring with no word-boundary check. LM_API_KEY= (LM Studio) is a literal suffix of GLM_API_KEY= (Z.AI/GLM), so doctor — or any code path that sanitizes .env — rewrote user files from GLM_API_KEY=... into G\nLM_API_KEY=..., silently killing Z.AI auth. Same shape for GLM_BASE_URL vs LM_BASE_URL.

Fix collects full (start, end) ranges per match and drops ranges fully contained inside a longer overlapping match. Real concatenations (no nesting) still split.

Validation

Contributor authorship preserved via rebase-merge.

Changed files

• hermes_cli/config.py (modified, +14/-5)
• tests/hermes_cli/test_config.py (modified, +17/-0)

PR #17291: fix(config): prevent _sanitize_env_lines from splitting valid keys via substring match

• Repository: NousResearch/hermes-agent
• Author: Luojz
• State: open | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/17291

Description (problem / solution / changelog)

Fixes #17138

Problem

_sanitize_env_lines() uses str.find() to detect concatenated KEY=VALUE pairs. This causes false positives when one known key name is a substring of another:

• LM_API_KEY= matches at position 1 inside GLM_API_KEY=xxx
• The sanitizer splits valid entries into garbage: G + LM_API_KEY=xxx
• Every save_env_value() call triggers this, permanently corrupting .env

Same issue affects GLM_BASE_URL / LM_BASE_URL and potentially any key pair where one name contains another as a substring.

Root Cause Timeline

• 2026-03-06: GLM support added (GLM_API_KEY, GLM_BASE_URL)
• 2026-04-13: _sanitize_env_lines added (substring detection via str.find)
• 2026-04-25: LM Studio integration added (LM_API_KEY, LM_BASE_URL) — substring collision begins

Fix

Two-pass approach with zero new lists to maintain:

Pass 0 (new): Merge split key names. Detect bare uppercase fragments (e.g. G) and merge with the next line when the result is a known key (e.g. GLM_API_KEY=). Uses the existing known_keys set — no additional key registry needed.

Pass 1 (improved): Collect all KEY= matches with (position, key_name_length), then filter out matches embedded inside another key's name via position comparison: other_start < this_start < other_start + other_key_length → skip (substring false positive)

Deterministic (sorted before filtering) and O(matches²) where matches per line is typically 1-3.

Design Advantages

1. Zero maintenance — uses existing OPTIONAL_ENV_VARS | _EXTRA_ENV_KEYS. No new key lists. Future keys auto-covered.
2. Deterministic — sorted before filtering. Same input, same output, regardless of set iteration order.
3. Position-based logic — one comparison per match pair. No character-type heuristics, no startswith scanning, no longest-match ambiguity.
4. Minimal diff — ~45 lines within a single function. Easy to review and revert.

Test Plan

18 tests in TestSanitizeEnvLines, all passing:

Changed files

• hermes_cli/config.py (modified, +89/-15)
• tests/hermes_cli/test_config.py (modified, +51/-0)