fix(state): resume compression continuation tip by stephenschoettler · Pull Request #15742 · NousResearch/hermes-agent

stephenschoettler · 2026-04-25T17:49:28Z

What does this PR do?

Fixes resume behavior for compressed sessions by redirecting resume targets to the live compression continuation tip.

After context compression, post-compression turns live in a child continuation session. Users may still resume the original root session id they saw before compression. SessionDB.resolve_resume_session_id() previously returned that root unchanged whenever the root already had pre-compression message rows, which can omit post-compaction turns from resumed context.

This PR makes the helper follow the compression chain tip first, while preserving the existing fallback for empty placeholder sessions that have child messages.

Related Issue

Related: #10373

Related PRs:

fix(tui): follow compression tips when resuming sessions #13374 covers complementary TUI resume behavior in tui_gateway/server.py.
fix(gateway): resume should follow compression continuations #15418 merged the gateway /resume path through SessionDB.resolve_resume_session_id().
This PR keeps the fix at the shared helper layer in hermes_state.py.

Type of Change

🐛 Bug fix (non-breaking change that fixes an issue)
✨ New feature (non-breaking change that adds functionality)
🔒 Security fix
📝 Documentation update
✅ Tests (adding or improving test coverage)
♻️ Refactor (no behavior change)
🎯 New skill (bundled or hub)

Changes Made

hermes_state.py
- Update SessionDB.resolve_resume_session_id() to call get_compression_tip() before returning a session with existing message rows.
- Preserve the legacy fallback for empty placeholder sessions by walking descendants until the first child with messages is found.
- Keep the malformed-chain depth guard behavior.
tests/test_hermes_state.py
- Add regression coverage for compressed roots that already have pre-compression messages.
- Add coverage that uncompressed sessions with messages still resolve to themselves.

How to Test

Run the focused regression suite:

python -m pytest \
  tests/test_hermes_state.py::TestCompressionChainProjection::test_resolve_resume_session_id_returns_compression_tip_even_when_parent_has_messages \
  tests/test_hermes_state.py::TestCompressionChainProjection::test_resolve_resume_session_id_returns_self_for_uncompressed_session_with_messages \
  tests/test_hermes_state.py::TestCompressionChainProjection::test_get_compression_tip_walks_full_chain \
  tests/test_hermes_state.py::TestCompressionChainProjection::test_list_surfaces_tip_for_compressed_root \
  -q -o addopts=''

Run the full hermes_state test module:

python -m pytest tests/test_hermes_state.py -q -o addopts=''

Run syntax and whitespace checks:

python -m compileall -q hermes_state.py tests/test_hermes_state.py
git diff --check

Validation Status

Focused regression suite from fresh recheck: 4 passed.
python -m pytest tests/test_hermes_state.py -q -o addopts='': passed locally in the original PR validation.
python -m compileall -q hermes_state.py tests/test_hermes_state.py: passed locally in the original PR validation.
git diff --check: passed locally in the original PR validation.
GitHub checks are not currently all green.
Failed checks: test, e2e.
Current merge state: mergeable but unstable.
Full-suite checklist is not marked complete because this branch needs rebasing/refreshing after the base-CI stabilizer lands.

Checklist

Code

I've read the Contributing Guide
My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
I searched for existing PRs and noted related fix(tui): follow compression tips when resuming sessions #13374 above
My PR contains only changes related to this fix/feature (no unrelated commits)
I've run pytest tests/ -q and all tests pass
I've added tests for my changes (required for bug fixes, strongly encouraged for features)
I've tested on my platform: Linux local checkout

Documentation and Housekeeping

I've updated relevant documentation (README, docs/, docstrings) or N/A
I've updated cli-config.yaml.example if I added/changed config keys or N/A
I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows or N/A
I've considered cross-platform impact (Windows, macOS) per the compatibility guide or N/A
I've updated tool descriptions/schemas if I changed tool behavior or N/A

For New Skills

N/A. This PR does not add a skill.

Screenshots / Logs

Fresh helper-level repro from a previous PR comment against origin/main:

get_compression_tip(root1)= tip1
resolve_resume_session_id(root1)= root1

Expected behavior after this PR:

resolve_resume_session_id(root1)= tip1

Fresh focused validation from the previous PR comment:

4 passed, 4 warnings in 0.23s

stephenschoettler · 2026-04-25T17:56:01Z

CI note: the red Tests job appears to be pre-existing main instability, not from this two-file session-resume change.

Local validation passed:

python -m pytest tests/test_hermes_state.py -q -o addopts=''
python -m compileall -q hermes_state.py tests/test_hermes_state.py
git diff --check

Recent main Tests runs are also failing before this PR, including main run 24936215121 for 648b899. The PR's Nix, lockfile, attribution, and supply-chain checks passed.

alt-glitch · 2026-04-25T18:04:55Z

Competes with open PR #13374 — both fix resolve_resume_session_id() to follow compression chain tips. Fixes #10373.

stephenschoettler · 2026-04-26T00:43:28Z

Thanks, good catch. I should have checked open overlapping PRs before opening this. That was an agent workflow miss, not intentional duplication.

I looked at #13374 now. My read is that #13374 fixes the TUI/gateway session.resume path by projecting the selected target through get_compression_tip(). This PR is narrower at the state helper layer: resolve_resume_session_id() currently returns the root unchanged when the root already has pre-compression message rows, which can still miss post-compression continuation turns for callers that rely on that helper.

If maintainers prefer consolidating this into #13374, I am happy to close/rebase. Otherwise I think the regression in this PR is still relevant because it covers the "compressed root already has messages" edge case in SessionDB.resolve_resume_session_id() directly.

stephenschoettler · 2026-04-28T00:02:07Z

Fresh recheck after #10373 was closed and #16306 merged: I do not think this PR should be closed as a duplicate yet.

Current main still has the helper-level edge case this PR targets. Repro against origin/main:

get_compression_tip(root1)= tip1
resolve_resume_session_id(root1)= root1

That means SessionDB.get_compression_tip() can see the live continuation tip, but SessionDB.resolve_resume_session_id() still returns the compressed root unchanged when the root already has pre-compression message rows.

Why I think this PR is still relevant:

fix(gateway): resume should follow compression continuations #15418 fixed the gateway /resume caller path by routing through SessionDB.resolve_resume_session_id().
This PR fixes that shared helper for compressed roots that already have message rows.
fix(tui): follow compression tips when resuming sessions #13374 is still complementary TUI coverage in tui_gateway/server.py.
fix: signal compression boundary to context engine #16306 is the context-engine compression-boundary signal and does not change this hermes_state.py helper behavior.

Fresh validation on this PR branch:

python -m pytest \
  tests/test_hermes_state.py::TestCompressionChainProjection::test_resolve_resume_session_id_returns_compression_tip_even_when_parent_has_messages \
  tests/test_hermes_state.py::TestCompressionChainProjection::test_resolve_resume_session_id_returns_self_for_uncompressed_session_with_messages \
  tests/test_hermes_state.py::TestCompressionChainProjection::test_get_compression_tip_walks_full_chain \
  tests/test_hermes_state.py::TestCompressionChainProjection::test_list_surfaces_tip_for_compressed_root \
  -q -o addopts=''

Result:

4 passed, 4 warnings in 0.23s

If maintainers prefer folding this helper fix into #13374, that is fine. Otherwise I think this PR should stay open because it covers the helper-level regression directly.

…-main-t39663ba2

stephenschoettler · 2026-05-27T03:16:00Z

Rechecked this against current origin/main: the helper-level resume edge still reproduces.

Local repro on current main:

get_compression_tip(root1)= tip1
resolve_resume_session_id(root1)= root1

I refreshed this branch by merging current origin/main. Targeted validation passed:

scripts/run_tests.sh tests/test_hermes_state.py -- -q --tb=short (218 passed)
python -m compileall -q hermes_state.py tests/test_hermes_state.py

This still looks distinct from the TUI/gateway tip PRs (#13374, #26631): those update gateway/TUI resume projection, while this fixes SessionDB.resolve_resume_session_id() itself. Leaving it open, but now current enough to evaluate on its own merits.

fix: resume compression continuation tip

c6ba71e

alt-glitch added type/bug Something isn't working P2 Medium — degraded but workaround exists comp/agent Core agent loop, run_agent.py, prompt builder labels Apr 25, 2026

stephenschoettler changed the title ~~fix: resume compression continuation tip~~ fix(state): resume compression continuation tip Apr 30, 2026

hejun041 mentioned this pull request May 7, 2026

fix: support proxied dashboard chat sessions #21104

Open

alt-glitch mentioned this pull request May 15, 2026

fix: resolve TUI resume to session tips #26631

Open

Merge remote-tracking branch 'origin/main' into merge/pr15742-current…

551f24f

…-main-t39663ba2

katanumahotori mentioned this pull request Jun 5, 2026

fix(tui): hydrate resumed compression history #39804

Open

pinguarmy mentioned this pull request Jun 5, 2026

fix(state,tui,desktop,api): include compression lineage in continuity hydration #40180

Open

kagura-agent mentioned this pull request Jun 12, 2026

fix(gateway): resolve compression chain in session.resume #44782

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(state): resume compression continuation tip#15742

fix(state): resume compression continuation tip#15742
stephenschoettler wants to merge 2 commits into
NousResearch:mainfrom
stephenschoettler:fix/resume-compression-tip

stephenschoettler commented Apr 25, 2026 •

edited

Loading

Uh oh!

stephenschoettler commented Apr 25, 2026

Uh oh!

alt-glitch commented Apr 25, 2026

Uh oh!

stephenschoettler commented Apr 26, 2026

Uh oh!

stephenschoettler commented Apr 28, 2026

Uh oh!

stephenschoettler commented May 27, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

stephenschoettler commented Apr 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What does this PR do?

Related Issue

Type of Change

Changes Made

How to Test

Validation Status

Checklist

Code

Documentation and Housekeeping

For New Skills

Screenshots / Logs

Uh oh!

stephenschoettler commented Apr 25, 2026

Uh oh!

alt-glitch commented Apr 25, 2026

Uh oh!

stephenschoettler commented Apr 26, 2026

Uh oh!

stephenschoettler commented Apr 28, 2026

Uh oh!

stephenschoettler commented May 27, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

stephenschoettler commented Apr 25, 2026 •

edited

Loading