fix(honcho): restore profile fallback and backfill memory files by stephenschoettler · Pull Request #13375 · NousResearch/hermes-agent

stephenschoettler · 2026-04-21T06:45:11Z

What does this PR do?

Fixes Honcho profile recall and memory-file migration for existing sessions.

honcho_profile could return an empty profile even when Honcho session context already had a populated peer card. Existing non-per-session Honcho sessions also did not import USER.md, MEMORY.md, or SOUL.md if those sessions were created before the migration path ran.

This PR adds fallback lookup through session-context peer-card data and backfills local memory files into existing non-per-session Honcho sessions once per content hash.

Related to #17124. This PR is broader than #17124 because it includes the memory-file backfill path in addition to profile fallback behavior.

Related Issue

No linked GitHub issue.

Related PR: #17124

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

plugins/memory/honcho/__init__.py
- Make honcho_profile fall back to session context before returning an empty profile result.
plugins/memory/honcho/session.py
- Fall back to session-context peer-card data when direct peer-card lookup is empty.
- Backfill USER.md, MEMORY.md, and SOUL.md into existing non-per-session Honcho sessions.
- Persist a local content-hash marker so the same memory content is not imported repeatedly.
tests/honcho_plugin/test_session.py
- Add regression coverage for peer-card fallback and existing-session memory backfill behavior.

How to Test

Run the Honcho session test module:

source venv/bin/activate
python -m pytest tests/honcho_plugin/test_session.py -q -o 'addopts='

Confirm profile fallback behavior:

honcho_profile returns session-context facts when direct peer-card lookup is empty.

Confirm existing-session backfill behavior:

Existing non-per-session Honcho sessions import local memory files once for a given memory-file hash.

Optional live harness:

source venv/bin/activate
python - <<'PY'
from plugins.memory.honcho.client import HonchoClientConfig, get_honcho_client
from plugins.memory.honcho.session import HonchoSessionManager

cfg = HonchoClientConfig.from_global_config()
client = get_honcho_client(cfg)
mgr = HonchoSessionManager(honcho=client, config=cfg)
session = mgr.get_or_create(cfg.resolve_session_name(cwd='/home/w0lf'))
print(mgr.get_peer_card(session.key, peer='user')[:8])
PY

Validation Status

Targeted regression coverage was added for this bug.
Original focused validation command: python -m pytest tests/honcho_plugin/test_session.py -q.
GitHub checks are not currently all green.
Current merge state: conflicting with main.
Failed check: test.
Full-suite checklist is not marked complete because this branch needs rebasing on a green main before merge validation.

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(honcho): fallback to durable peer cards for profiles #17124 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 venv

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

Live verification harness from the original PR description:

source venv/bin/activate
python - <<'PY'
from plugins.memory.honcho.client import HonchoClientConfig, get_honcho_client
from plugins.memory.honcho.session import HonchoSessionManager

cfg = HonchoClientConfig.from_global_config()
client = get_honcho_client(cfg)
mgr = HonchoSessionManager(honcho=client, config=cfg)
s = mgr.get_or_create(cfg.resolve_session_name(cwd='/home/w0lf'))
print(mgr.get_peer_card(s.key, peer='user')[:8])
PY

- fall back to session context when direct peer-card lookup is empty - backfill USER.md, MEMORY.md, and SOUL.md into existing sessions once per content hash - add regression tests for both paths

stephenschoettler · 2026-04-21T06:52:18Z

CI audit: the remaining failure is the shared Tests workflow, not the Honcho patch. This PR changes only plugins/memory/honcho/init.py, plugins/memory/honcho/session.py, and tests/honcho_plugin/test_session.py; targeted Honcho tests pass locally (117 passed). The failing workflow is also red on current main (run 24707349937), with unrelated failures in provider config / skills config / command guard tests. I also re-ran the four failing test nodes from this PR against a clean origin/main worktree and they passed in isolation, which points to broader suite instability rather than this patch.

# Conflicts: # plugins/memory/honcho/__init__.py

stephenschoettler · 2026-05-07T02:14:58Z

Updated against current green main (49c3c2e0).

What changed on the branch:

Merged current main into fix/honcho-profile-fallback-backfill-v2.
Resolved the Honcho profile conflict by preserving the current _empty_profile_hint() behavior and keeping this PR's session-context card fallback.
PR is now mergeable.

Focused local validation:

python -m pytest tests/honcho_plugin/test_session.py -q -o addopts='': 118 passed
python -m compileall plugins/memory/honcho tests/honcho_plugin/test_session.py -q: passed
git diff --check origin/main...HEAD: passed

Current CI notes:

nix, e2e, attribution, and supply-chain checks are green.
ruff + ty diff is red because the advisory workflow tries to post a PR comment from a fork PR and GitHub returns 403 Resource not accessible by integration. The generated lint summary reports 0 ruff issues and no net new ty count.
test is red with broad failures outside this Honcho change set, including Bedrock, gateway restart, update, browser Chromium, delegate, and sandbox tests. Focused Honcho validation is clean.

stephenschoettler · 2026-05-27T02:54:16Z

Closing this lane in favor of the narrower #17124.

Current Honcho is operational on my local setup without this broader backfill patch: memory status and honcho status are OK, honcho_search/context/reasoning return data, and the remaining profile-card gap is the narrower peer-card lookup path covered by #17124. If a distinct memory-file backfill regression still reproduces on current main, I will split/reopen that as a focused PR instead of keeping this broader overlapping branch alive.

honcho_profile(peer="user") returned an empty card even when Honcho held a populated peer card for the user. Two independent bugs combined to produce the symptom: 1. Read path: get_peer_card() called _fetch_peer_card(observer, target=user), which hits GET /peers/{observer}/card?target={user} — the observer's local card of the user. On self-hosted Honcho v3 this slot is empty unless writes also use it. The peer card lives on the user peer itself (GET /peers/{user}/card). Add a fallback: when the observer-target slot is empty and a target exists, retry against the target peer's own card. 2. Write path: set_peer_card() resolved only the target peer and called user_peer.set_card(card). The read path uses the assistant peer as observer, so writes and reads addressed different Honcho card scopes. Align set_peer_card() with _resolve_observer_target() so writes go to assistant_peer.set_card(card, target=user_peer_id), matching the read. Both paths now use the same observer/target resolution, and the read path additionally falls back to the target's own card for compatibility with deployments where cards were written directly to the peer. Closes: related to NousResearch#13375, NousResearch#17124, NousResearch#20729

honcho_profile(peer="user") returned an empty card even when Honcho held a populated peer card for the user. Two independent bugs combined to produce the symptom: 1. Read path: get_peer_card() called _fetch_peer_card(observer, target=user), which hits GET /peers/{observer}/card?target={user} — the observer's local card of the user. On self-hosted Honcho v3 this slot is empty unless writes also use it. The peer card lives on the user peer itself (GET /peers/{user}/card). Add a fallback: when the observer-target slot is empty and a target exists, retry against the target peer's own card. 2. Write path: set_peer_card() resolved only the target peer and called user_peer.set_card(card). The read path uses the assistant peer as observer, so writes and reads addressed different Honcho card scopes. Align set_peer_card() with _resolve_observer_target() so writes go to assistant_peer.set_card(card, target=user_peer_id), matching the read. Both paths now use the same observer/target resolution, and the read path additionally falls back to the target's own card for compatibility with deployments where cards were written directly to the peer. Closes: related to #13375, #17124, #20729

honcho_profile(peer="user") returned an empty card even when Honcho held a populated peer card for the user. Two independent bugs combined to produce the symptom: 1. Read path: get_peer_card() called _fetch_peer_card(observer, target=user), which hits GET /peers/{observer}/card?target={user} — the observer's local card of the user. On self-hosted Honcho v3 this slot is empty unless writes also use it. The peer card lives on the user peer itself (GET /peers/{user}/card). Add a fallback: when the observer-target slot is empty and a target exists, retry against the target peer's own card. 2. Write path: set_peer_card() resolved only the target peer and called user_peer.set_card(card). The read path uses the assistant peer as observer, so writes and reads addressed different Honcho card scopes. Align set_peer_card() with _resolve_observer_target() so writes go to assistant_peer.set_card(card, target=user_peer_id), matching the read. Both paths now use the same observer/target resolution, and the read path additionally falls back to the target's own card for compatibility with deployments where cards were written directly to the peer. Closes: related to NousResearch#13375, NousResearch#17124, NousResearch#20729

honcho_profile(peer="user") returned an empty card even when Honcho held a populated peer card for the user. Two independent bugs combined to produce the symptom: 1. Read path: get_peer_card() called _fetch_peer_card(observer, target=user), which hits GET /peers/{observer}/card?target={user} — the observer's local card of the user. On self-hosted Honcho v3 this slot is empty unless writes also use it. The peer card lives on the user peer itself (GET /peers/{user}/card). Add a fallback: when the observer-target slot is empty and a target exists, retry against the target peer's own card. 2. Write path: set_peer_card() resolved only the target peer and called user_peer.set_card(card). The read path uses the assistant peer as observer, so writes and reads addressed different Honcho card scopes. Align set_peer_card() with _resolve_observer_target() so writes go to assistant_peer.set_card(card, target=user_peer_id), matching the read. Both paths now use the same observer/target resolution, and the read path additionally falls back to the target's own card for compatibility with deployments where cards were written directly to the peer. Closes: related to NousResearch#13375, NousResearch#17124, NousResearch#20729 #AI commit#

honcho_profile(peer="user") returned an empty card even when Honcho held a populated peer card for the user. Two independent bugs combined to produce the symptom: 1. Read path: get_peer_card() called _fetch_peer_card(observer, target=user), which hits GET /peers/{observer}/card?target={user} — the observer's local card of the user. On self-hosted Honcho v3 this slot is empty unless writes also use it. The peer card lives on the user peer itself (GET /peers/{user}/card). Add a fallback: when the observer-target slot is empty and a target exists, retry against the target peer's own card. 2. Write path: set_peer_card() resolved only the target peer and called user_peer.set_card(card). The read path uses the assistant peer as observer, so writes and reads addressed different Honcho card scopes. Align set_peer_card() with _resolve_observer_target() so writes go to assistant_peer.set_card(card, target=user_peer_id), matching the read. Both paths now use the same observer/target resolution, and the read path additionally falls back to the target's own card for compatibility with deployments where cards were written directly to the peer. Closes: related to NousResearch#13375, NousResearch#17124, NousResearch#20729

fix(honcho): restore profile fallback and backfill memory files

8140f1b

- fall back to session context when direct peer-card lookup is empty - backfill USER.md, MEMORY.md, and SOUL.md into existing sessions once per content hash - add regression tests for both paths

This was referenced Apr 21, 2026

fix(honcho): restore profile fallback and backfill memory files #13369

Closed

fix(honcho): restore profile fallback and backfill memory files #13359

Closed

alt-glitch added type/bug Something isn't working P2 Medium — degraded but workaround exists comp/plugins Plugin system and bundled plugins tool/memory Memory tool and memory providers labels Apr 22, 2026

stephenschoettler mentioned this pull request Apr 28, 2026

fix(honcho): fallback to durable peer cards for profiles #17124

Open

19 tasks

stephenschoettler added 2 commits May 6, 2026 19:03

fix(honcho): update profile fallback backfill branch

d7e507d

Merge current main into PR NousResearch#13375

0d2d655

# Conflicts: # plugins/memory/honcho/__init__.py

github-actions Bot mentioned this pull request May 7, 2026

🦞 OpenClaw 生态日报 2026-05-07 ivanweng2077/big_model_radar#6

Open

stephenschoettler mentioned this pull request May 7, 2026

fix(ci): unblock shared PR checks #21012

Merged

25 tasks

cyb0rgk1tty mentioned this pull request May 18, 2026

fix(honcho): align peer-card read and write paths #27979

Closed

9 tasks

stephenschoettler closed this May 27, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(honcho): restore profile fallback and backfill memory files#13375

fix(honcho): restore profile fallback and backfill memory files#13375
stephenschoettler wants to merge 3 commits into
NousResearch:mainfrom
stephenschoettler:fix/honcho-profile-fallback-backfill-v2

stephenschoettler commented Apr 21, 2026 •

edited

Loading

Uh oh!

stephenschoettler commented Apr 21, 2026

Uh oh!

stephenschoettler commented May 7, 2026

Uh oh!

stephenschoettler commented May 27, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

stephenschoettler commented Apr 21, 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 21, 2026

Uh oh!

stephenschoettler commented May 7, 2026

Uh oh!

stephenschoettler commented May 27, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

stephenschoettler commented Apr 21, 2026 •

edited

Loading