PR #15293: chore(nix): probe hermes-web npm hash

• Repository: NousResearch/hermes-agent
• Author: LeonSGP43
• State: closed | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/15293

Description (problem / solution / changelog)

Temporary draft PR to trigger CI and capture the updated hermes-web npm deps hash for issue #15272. This will be superseded by a clean final PR with the real hash.

Changed files

• nix/web.nix (modified, +1/-1)

---

PR #15295: fix(nix): refresh hermes-web npm lockfile hash

• Repository: NousResearch/hermes-agent
• Author: LeonSGP43
• State: open | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/15295

Description (problem / solution / changelog)

Summary\n- refresh the stale npm deps hash for \n- unblock the shared Nix failure tracked in #15272\n\n## Verification\n- extracted the updated hash from the repository's workflow report for probe PR #15293\n- final verification is pending on this PR's CI

Changed files

• nix/web.nix (modified, +1/-1)

---

PR #15260: fix(docs): add ┴ junctions to two outer-box bottom borders to satisfy ascii-guard

• Repository: NousResearch/hermes-agent
• Author: perlowja
• State: open | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/15260

Description (problem / solution / changelog)

What does this PR do?

Fixes 6 pre-existing ascii-guard 2.3.0 warnings that currently block the `docs-site-checks` CI on every PR touching `website/**`.

• `website/docs/developer-guide/architecture.md` — 3 warnings on the bottom border of the outer AIAgent box (line 38)
• `website/docs/developer-guide/gateway-internals.md` — 3 warnings on the bottom border of the outer GatewayRunner box (line 49)

Both diagrams have a two-row nested-box pattern where the first inner-box row's bottom uses `┬` to connect downward to a second inner-box row. ascii-guard's `validate_box()` routine looks for `┬`/`┼` characters anywhere in the outer box's content lines (not just in its top border) via `get_column_positions()` and concludes those are table-column boundaries belonging to the outer box. It then expects a matching `┴` in the outer box's bottom border — and warns when it finds `─` instead. That interpretation is wrong for this geometry (the `┬` is a purely internal inner-box-to-inner-box connection), but the linter rule is what it is, and the fix is trivial: add `┴` at the three flagged columns of each outer bottom border.

After the fix, the `┴` reads visually as "the column space continues past this box," which matches the arrows already sitting below each outer box. No readability regression.

Related Issue

No existing issue — noticed while filing #15214 (docs cookbook). Verified the warnings are pre-existing on `main` (`architecture.md` last touched in #11373, `gateway-internals.md` much older).

Fixes #

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue) — unblocks `docs-site-checks` CI
• ✨ New feature
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests
• ♻️ Refactor
• 🎯 New skill

Changes Made

• `website/docs/developer-guide/architecture.md` line 38: positions 11, 29, 47 (1-indexed) on the outer bottom border changed from `─` to `┴`. Those positions line up with the `┬` junctions on line 30 that connect the "Prompt Builder / Provider Resolution / Tool Dispatch" row to the "Compression / 3 API Modes / Tool Registry" row below.
• `website/docs/developer-guide/gateway-internals.md` line 49: positions 9, 23, 37 (1-indexed) on the outer bottom border changed from `─` to `┴`. Those positions line up with the `┬` junctions on line 35 that connect the "Telegram / Discord / Slack Adapter" row to the `_handle_message()` merge below.

Diff: 2 files changed, 2 insertions(+), 2 deletions(-) — every changed character is a single Unicode swap.

How to Test

Validated locally against ascii-guard 2.3.0 (same version the CI workflow installs via `python -m pip install ascii-guard==2.3.0`).

Before (on clean `upstream/main`): ``` $ ascii-guard lint website/docs/developer-guide/architecture.md Checking .../architecture.md... Found 10 ASCII box(es) ⚠ Line 39, Col 12: Bottom border missing junction point at column 11 (expected ┴, got '─') ⚠ Line 39, Col 30: Bottom border missing junction point at column 29 (expected ┴, got '─') ⚠ Line 39, Col 48: Bottom border missing junction point at column 47 (expected ┴, got '─')

$ ascii-guard lint website/docs/developer-guide/gateway-internals.md Checking .../gateway-internals.md... Found 4 ASCII box(es) ⚠ Line 50, Col 10: Bottom border missing junction point at column 9 (expected ┴, got '─') ⚠ Line 50, Col 24: Bottom border missing junction point at column 23 (expected ┴, got '─') ⚠ Line 50, Col 38: Bottom border missing junction point at column 37 (expected ┴, got '─') ```

After (on this branch): ``` $ ascii-guard lint website/docs/developer-guide/architecture.md Checking .../architecture.md... Found 10 ASCII box(es) ✓ No issues found ✓ Errors: 0

$ ascii-guard lint website/docs/developer-guide/gateway-internals.md Checking .../gateway-internals.md... Found 4 ASCII box(es) ✓ No issues found ✓ Errors: 0 ```

No other files changed. Other pre-existing ascii-guard findings on committed-but-regenerated-in-CI skill-docs (`skills/bundled/research/research-research-paper-writing.md`, `skills/bundled/mlops/mlops-models-segment-anything.md`) are out of scope for this PR — CI runs `generate-skill-docs.py` before the lint step, which overwrites those files with clean output, so they don't actually fail CI.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (`fix(docs): …`)
• I searched for existing PRs — nothing duplicative
• My PR contains only changes related to this fix (2 `─` → `┴` swaps on 2 lines)
• I've run `pytest tests/ -q` — N/A for a docs-only change to 2 markdown files; the Python test suite does not exercise these files
• I've added tests — N/A, existing CI `docs-site-checks` workflow is the regression guard
• I've tested on my platform: macOS 15 / darwin-arm64, ascii-guard 2.3.0

Documentation & Housekeeping

• Documentation (README, `docs/`, docstrings) — this PR IS a docs fix
• `cli-config.yaml.example` — N/A
• `CONTRIBUTING.md` or `AGENTS.md` — N/A
• Cross-platform impact considered — ASCII art renders identically on all platforms
• Tool descriptions/schemas — N/A

For New Skills

N/A.

Screenshots / Logs

The 3 `┴` characters on `architecture.md` line 38 now align with the `┬` connectors on line 30 (same columns: 11, 29, 47). Same pattern for `gateway-internals.md`.

Unblocks #15214 (and every other open PR touching `website/**`) from the same CI failure.

Changed files

• website/docs/developer-guide/architecture.md (modified, +1/-1)
• website/docs/developer-guide/gateway-internals.md (modified, +1/-1)

---

PR #15420: fix(nix): refresh tui + web npmDepsHash to unblock CI on every PR (#15244, #15272, #15314)

• Repository: NousResearch/hermes-agent
• Author: briandevans
• State: closed | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/15420

Description (problem / solution / changelog)

What does this PR do?

The `Nix Lockfile Check` on `main` has been failing across every open PR for several days because two pinned `fetchNpmDeps` hashes drifted out of sync with the actual `package-lock.json` contents in `ui-tui/` and `web/`.

The CI run prints the correct hash on each failure, so this is a one-character-precision drive-by — both new hashes are exactly what every recent `Nix Lockfile Check` run is reporting (e.g. run 24916813199 from a few minutes ago):

The drift was caused by recent dependency-touching commits in:

• `ui-tui/`: PTY WebSocket bridge (`f49afd31`), package.json updates across the chat-unification series (`c61547c0`), `fix(tui): warn on bare null sections in config.yaml` (`bfa60234`)
• `web/`: docs embedding (`0fdbfad2`), sidebar layout (`e5d2815b`), dashboard extension points (`f593c367`)

None of those PRs ran the `nix flake update` follow-up, so the cached hashes never got refreshed.

Related Issues

• Closes #15244 (`nix ubuntu check failing on main again`)
• Closes #15272 (`nix CI broken on main — npmDepsHash out of date`)
• Closes #15314 (`hermes-tui npmDepsHash is out of date`)

Type of Change

• 🐛 Bug fix (CI infrastructure)
• No tests required — pure config update; the existing `Nix Lockfile Check` workflow IS the regression guard for this kind of drift

Test plan

• Both new hashes are taken verbatim from `Nix Lockfile Check` failure output: the CI computes the correct hash from the actual `package-lock.json` and prints "To correct the hash mismatch for npm-deps, use ..."
• No production code touched — only the two pinned hashes
• Once merged: every open PR's `Nix Lockfile Check` lane goes from FAILURE to SUCCESS on its next CI run

Why a separate PR (vs piggybacking on another fix)

This fix unblocks ~30 open PRs at once, including all of mine. Keeping it isolated makes the merge low-risk (1 line per file), explicit in git history, and easy to cherry-pick onto release branches if needed.

Out of scope

• Automating the hash refresh on every `package-lock.json` change (would need a CI job to compute and commit the hash). `#13136` proposed this with `automatic lockfile fixing to keep main building with nix` but was reverted in `688c9f5b`. Restoring that automation is a bigger conversation; this PR is a narrow unblock for today.

Changed files

• nix/tui.nix (modified, +1/-1)
• nix/web.nix (modified, +1/-1)