Ignore 'last-prompt' message type#112
Merged
cboos merged 3 commits intodaaain:mainfrom Apr 19, 2026
Merged
Conversation
📝 WalkthroughWalkthrough
Changes
Estimated code review effort🎯 2 (Simple) | ⏱️ ~10 minutes 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 |
Collaborator
|
Thanks for your first contributed line ;-) |
Collaborator
|
"This branch has conflicts that must be resolved" - at first, I was puzzled by that, but there's an explanation:
So, basically, #102 is already fixed. I'll add a regression test, instead. |
Claude Code now writes a trailing {type:last-prompt,...} line as the
last entry of session .jsonl files. Add it to the silent-skip list
alongside file-history-snapshot and progress, so it does not print a
'not a recognised message type' warning.
Fixes daaain#102
…elds Before PR daaain#97's DAG refactor, load_transcript explicitly listed internal message types (file-history-snapshot, progress) as silent skips, and warned about anything else. The refactor replaced that with an unknown → PassthroughTranscriptEntry fallback for entries with uuid+sessionId, and an implicit drop for everything else — which also silently drops types like custom-title and agent-name that Claude Code introduced later, so they never show up for maintainers to notice. Restore the old signal by reintroducing an explicit SILENT_SKIP_TYPES frozenset and an else-branch warning: - file-history-snapshot and last-prompt (fixes daaain#102) are listed and dropped without warning. Both lack uuid/sessionId in the wild. - progress is intentionally NOT in the skip list — it has uuid, so the uuid+sessionId branch keeps it as PassthroughTranscriptEntry for DAG continuity (the existing test coverage in test_dag_integration.py relies on that). - Everything else that reaches the else-branch prints a single-line warning quoting the unknown type so future metadata surfaces. Update test_multiple_passthrough_types to use a generic "unknown-future-type" label instead of "file-history-snapshot" for its synthetic passthrough entry — the test was never representative of real file-history-snapshot data (which has no uuid) and conflicted with the silent-skip list.
Covers the four paths through load_transcript's non-conversation type handling: - file-history-snapshot, last-prompt: silent-skip path. No output on stderr/stdout, zero messages. Direct regression for daaain#102. - progress: uuid+sessionId path. Preserved as PassthroughTranscriptEntry, silently (verifies it stays out of SILENT_SKIP_TYPES). - custom-title, agent-name, future-metadata-type: else-branch warning path. Parameterised to make "what's on this path" explicit. Also checks that silent=True suppresses the warning. - Unknown type with uuid: PassthroughTranscriptEntry fallback path, no warning.
cboos
force-pushed
the
fix/ignore-last-prompt-message-type
branch
from
16c3b0d to
a2f9b4c
Compare
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
🧹 Nitpick comments (1)
claude_code_log/converter.py (1)
343-360: Preserve DAG nodes before applying silent skips.
SILENT_SKIP_TYPESis documented as non-DAG, but this branch now runs before theuuid+sessionIdpassthrough path. If Claude adds DAG fields to one of these metadata types later, it would be dropped and could create false roots. Prefer preserving DAG-bearing entries first, then silently skipping non-DAG metadata.Proposed branch-order tweak
- elif entry_type in SILENT_SKIP_TYPES: - # Internal Claude Code entries with no DAG fields. - pass - elif entry_dict.get("uuid") and entry_dict.get("sessionId"): + elif entry_dict.get("uuid") and entry_dict.get("sessionId"): # Unknown type with DAG-relevant fields — create a # PassthroughTranscriptEntry to preserve DAG chain # continuity (e.g. "attachment", "permission-mode"). messages.append( PassthroughTranscriptEntry( uuid=entry_dict["uuid"], parentUuid=entry_dict.get("parentUuid"), sessionId=entry_dict["sessionId"], timestamp=entry_dict.get("timestamp", ""), type=entry_type, isSidechain=entry_dict.get("isSidechain", False), agentId=entry_dict.get("agentId"), ) ) + elif entry_type in SILENT_SKIP_TYPES: + # Internal Claude Code entries with no DAG fields. + pass🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@claude_code_log/converter.py` around lines 343 - 360, The current branch checks SILENT_SKIP_TYPES before the passthrough DAG-preserving branch, which can drop entries that actually carry DAG fields; update the condition order in converter.py so the branch that detects DAG-relevant entries (the check using entry_dict.get("uuid") and entry_dict.get("sessionId") that constructs a PassthroughTranscriptEntry and appends to messages) runs before the SILENT_SKIP_TYPES check (entry_type in SILENT_SKIP_TYPES), or alternatively enhance the SILENT_SKIP_TYPES branch to skip only when uuid/sessionId are absent; ensure you reference entry_type, entry_dict, PassthroughTranscriptEntry and messages when making the change so DAG-bearing entries are preserved.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Nitpick comments:
In `@claude_code_log/converter.py`:
- Around line 343-360: The current branch checks SILENT_SKIP_TYPES before the
passthrough DAG-preserving branch, which can drop entries that actually carry
DAG fields; update the condition order in converter.py so the branch that
detects DAG-relevant entries (the check using entry_dict.get("uuid") and
entry_dict.get("sessionId") that constructs a PassthroughTranscriptEntry and
appends to messages) runs before the SILENT_SKIP_TYPES check (entry_type in
SILENT_SKIP_TYPES), or alternatively enhance the SILENT_SKIP_TYPES branch to
skip only when uuid/sessionId are absent; ensure you reference entry_type,
entry_dict, PassthroughTranscriptEntry and messages when making the change so
DAG-bearing entries are preserved.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: d7905d07-dfac-4e0b-8570-64821e18ac38
📒 Files selected for processing (3)
claude_code_log/converter.pytest/test_dag_integration.pytest/test_silent_skip.py
✅ Files skipped from review due to trivial changes (1)
- test/test_dag_integration.py
Collaborator
|
Sorry for the "force push" ;-) I believe it was either that or closing the PR and recreating a new one.
Net effect: fixes #102, restores the "notice new Claude Code types" signal that went dark after PR #97, and adds the first explicit test coverage for the non-conversation type dispatch. Local just test passes 871/871 (excluding pre-existing pagination env failures). |
cboos
added a commit
to fuleinist/claude-code-log
that referenced
this pull request
Apr 19, 2026
Claude Code writes per-session state snapshots to the transcript with no uuid/parentUuid/timestamp — they are positional markers, not DAG nodes. They arrive frequently (e.g. a `permission-mode` after every mode toggle, an `agent-name`+`custom-title` pair after every /rename) and would otherwise drown the unrecognised-type warning introduced on top of daaain#112. Four types land in the silent-skip list: - permission-mode : {permissionMode} - custom-title : {customTitle} - agent-name : {agentName} - agent-color : {agentColor} All four will become load-bearing once we propagate their state onto conversational messages in a follow-up (see daaain#94). Dropping them silently for now is a strict improvement over the current state: either noisy warnings (reinstated warning branch on top of daaain#112) or silent loss (main). Extend test_silent_skip with a parameterised case covering all four and repoint the unrecognised-type parametrisation onto two hypothetical future types, since the original custom-title / agent-name samples now land on the silent path.
cboos
added a commit
that referenced
this pull request
Apr 19, 2026
…es (#113) * Handle custom-title, agent-name, and agent-color transcript entry types These session metadata entry types are generated by Claude Code but were not recognized by the parser, causing 'Unknown transcript entry type' errors. Changes: - Make uuid and timestamp optional in PassthroughTranscriptEntry to support entry types that lack these fields - Add custom-title, agent-name, and agent-color to ENTRY_CREATORS registry - Update fallback in create_transcript_entry to handle any unknown type with a sessionId gracefully * Extend SILENT_SKIP_TYPES with session-metadata types (#94) Claude Code writes per-session state snapshots to the transcript with no uuid/parentUuid/timestamp — they are positional markers, not DAG nodes. They arrive frequently (e.g. a `permission-mode` after every mode toggle, an `agent-name`+`custom-title` pair after every /rename) and would otherwise drown the unrecognised-type warning introduced on top of #112. Four types land in the silent-skip list: - permission-mode : {permissionMode} - custom-title : {customTitle} - agent-name : {agentName} - agent-color : {agentColor} All four will become load-bearing once we propagate their state onto conversational messages in a follow-up (see #94). Dropping them silently for now is a strict improvement over the current state: either noisy warnings (reinstated warning branch on top of #112) or silent loss (main). Extend test_silent_skip with a parameterised case covering all four and repoint the unrecognised-type parametrisation onto two hypothetical future types, since the original custom-title / agent-name samples now land on the silent path. * Document session-state propagation plan for #94 follow-up Spells out the deferred half of issue #94 — turning the now-silent session-metadata types into a visible "Assistant · CCL (Monk)" decoration on conversational messages, with colour tinting from agent-color. Covers: data shape (no uuid/timestamp, pure positional markers), file-position propagation semantics (single self-contained session file), six concrete change sites (MessageMeta fields, load_transcript state tracker, private-attr channel onto pydantic entries, create_meta forwarding, title_AssistantTextMessage decoration, CSS), cache concerns, open questions (separator, colour palette, permission-mode surfacing, session nav integration), and risks (snapshot churn, private-attr subtlety, markdown heading compactor). This PR does NOT complete #94. Wording here is deliberately free of GitHub close keywords: #94 stays open as the tracker for the state-propagation implementation documented in work/session-state-propagation.md. The PR description should also be edited to use a plain `#94` reference rather than a closing keyword so merging does not auto-close the issue. * Switch warning text to American spelling; tighten silent-skip tests Two small follow-ups on top of the SILENT_SKIP_TYPES work. The "unrecognised message type" warning landed in the prior PR using the British spelling, matching the pre-#97 original phrasing. #94 explicitly called out the preferred `s/recognised/recognized/`, so flip it here. Single user-facing string; the else-branch is only reached for unknown types, so unlikely to bite anyone scraping logs. CodeRabbit flagged the silent-skip test assertions — they used the weak form `"unrecognised" not in captured.out`, which only catches a specific substring and silently tolerates any other output the loader might accidentally start printing. Switch every silent-path test to call `load_transcript(..., silent=True)` and assert `captured.out == ""`; the warn-path test keeps `silent=False` and now matches the new American spelling. --------- Co-authored-by: fuleinist <fuleinist@gmail.com> Co-authored-by: Christian Boos <christian.boos@bct-technology.com>
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
3 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.
Fixes #102. Added
last-promptto the silent-skip list inconverter.pyalongsidefile-history-snapshotandprogress, so Claude Code's new trailing{type:last-prompt,...}line no longer triggers a 'not a recognised message type' warning.Also maintain a small open source tool in the Claude Code space, happy to contribute more if useful.
Summary by CodeRabbit