feat(hermes): capture session transcripts into Mem threads#194
Merged
wey-gu merged 4 commits intonowledge-co:mainfrom Apr 15, 2026
Merged
Conversation
📝 WalkthroughWalkthroughAdds automatic Hermes session transcript capture into Nowledge Mem via a new Changes
Sequence DiagramsequenceDiagram
participant Session as Hermes Session
participant Provider as NowledgeMemProvider
participant Client as NowledgeMemClient
participant CLI as nmem CLI
participant Mem as Mem Thread Store
Session->>Provider: on_session_end(messages)
Provider->>Provider: Clean & normalize messages (filter roles, extract text)
Provider->>Provider: Derive thread title from first user message
alt First flush (saved_count ≤ 0)
Provider->>Client: import_thread(thread_id, cleaned_messages, title, source="hermes")
Client->>CLI: nmem --json t import --id <id> -s hermes (JSON messages)
CLI->>Mem: create thread with messages
Mem-->>CLI: success + metadata
CLI-->>Client: result
Client-->>Provider: update _saved_message_count
else Subsequent flush (saved_count > 0)
Provider->>Provider: compute delta (new messages only)
Provider->>Client: append_thread(thread_id, delta_messages)
Client->>CLI: nmem --json t append --id <id> (JSON messages)
CLI->>Mem: append messages
Mem-->>CLI: success + metadata
CLI-->>Client: result
Client-->>Provider: update _saved_message_count
end
Provider-->>Session: complete
Estimated code review effort🎯 4 (Complex) | ⏱️ ~40 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: e24f15e6a7
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 4
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@integrations.json`:
- Line 450: The "threads" value currently set to "automatic" violates the
allowed autonomy enum; update the autonomy.threads entry (the JSON key
"threads": "automatic") to one of the valid values listed in the schema (for
example "automatic-capture", "explicit-save", "handoff-only", "import-only", or
"none") so consumers and schema validators accept it; pick the semantics you
need (e.g., use "automatic-capture" if threads should be auto-created and saved)
and replace "automatic" accordingly.
- Around line 443-455: autonomy.threads contains an invalid value; update the
autonomy.threads field from "automatic" to one of the schema's allowed enums
(most likely "automatic-capture") to satisfy the schema, and ensure the change
is applied where autonomy is defined (look for the autonomy object and its
threads key). Also verify that the new threadSave.method value
"cli-import-append" is intentional and documented so it aligns with supported
provider behavior.
In `@nowledge-mem-hermes/provider.py`:
- Around line 608-610: The loop over messages assumes each item is a dict and
uses message.get(...) which will raise for non-dict entries; update the cleaning
loop that iterates over messages (the for message in messages or [] block) to
first guard each item (e.g., isinstance(message, dict)) and skip any non-dict or
malformed entries before accessing message.get("role"), so the sync pass won't
fail; ensure you still strip/validate role and only process items with role in
{"user","assistant"}.
- Around line 358-373: In on_session_end, wrap calls to
self._client.import_thread(...) and self._client.append_thread(...) in
try/except so failures are non-fatal: catch Exception around the import_thread
and append_thread invocations (both places where import_thread and append_thread
are called), log the error (including exception details) and continue execution,
and still update self._saved_message_count appropriately only on successful
persistence; use existing symbols cleaned_messages, session_id, title (from
_build_thread_title) and self._saved_message_count to locate the spots to
change.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 6b58ca79-6335-4221-a99b-470a1922222f
📒 Files selected for processing (7)
integrations.jsonnowledge-mem-hermes/CHANGELOG.mdnowledge-mem-hermes/README.mdnowledge-mem-hermes/client.pynowledge-mem-hermes/plugin.yamlnowledge-mem-hermes/provider.pynowledge-mem-hermes/tests/test_session_sync.py
Member
|
wow @HamsteRider-m 牛逼，i didnt realize they provide the abstractions of hermes hook on thread sync!!! |
There was a problem hiding this comment.
🧹 Nitpick comments (1)
nowledge-mem-hermes/tests/test_session_sync.py (1)
101-101: Consider iterable unpacking for list extension.Static analysis suggests using unpacking instead of concatenation.
Suggested change
- next_messages = first_messages + [{"role": "user", "content": "next step"}] + next_messages = [*first_messages, {"role": "user", "content": "next step"}]🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@nowledge-mem-hermes/tests/test_session_sync.py` at line 101, Replace the list concatenation that builds next_messages from first_messages and the new dict with iterable unpacking: locate the expression assigning next_messages (variable name next_messages and source list first_messages) and construct next_messages using list unpacking of first_messages plus the single new message element to produce the combined list.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Nitpick comments:
In `@nowledge-mem-hermes/tests/test_session_sync.py`:
- Line 101: Replace the list concatenation that builds next_messages from
first_messages and the new dict with iterable unpacking: locate the expression
assigning next_messages (variable name next_messages and source list
first_messages) and construct next_messages using list unpacking of
first_messages plus the single new message element to produce the combined list.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: a36a3499-e550-43b2-87db-3e5f1960d9f0
📒 Files selected for processing (5)
integrations.jsonnowledge-mem-hermes/CHANGELOG.mdnowledge-mem-hermes/README.mdnowledge-mem-hermes/provider.pynowledge-mem-hermes/tests/test_session_sync.py
✅ Files skipped from review due to trivial changes (1)
- nowledge-mem-hermes/CHANGELOG.md
🚧 Files skipped from review as they are similar to previous changes (2)
- nowledge-mem-hermes/README.md
- integrations.json
wey-gu
changed the title
3 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds native Hermes session transcript capture to the Nowledge Mem provider.
Hermes now exposes a provider lifecycle hook for session end. The Nowledge Mem provider uses that hook to capture cleaned
user/assistanttranscript turns into Mem threads at real session boundaries, alongside the existing Working Memory bootstrap, pre-turn recall, and nativenmem_*tools.Behavior
nmem_*toolsContract
This is real transcript capture, but it is session-boundary capture. It runs when Hermes ends or resets a session cleanly, such as normal exit,
/new,/reset, or normal gateway expiry. It is not per-turn persistence and it cannot recover a transcript if the host process is killed before Hermes fireson_session_end.Validation
python3 -m py_compile nowledge-mem-hermes/client.py nowledge-mem-hermes/provider.pynowledge-mem-hermes/tests/test_session_sync.pypython3 -m unittest tests.test_space_resolution -vinsidenowledge-mem-hermesgit diff --check