Fix user text message deduplication to keep best version #65
Conversation
Caches from v0.8.0 and earlier need regeneration because v0.9.0 introduced _compact_ide_tags_for_preview() which transforms first_user_message to use emoji indicators instead of raw IDE tags. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <[email protected]>
Browser tests were failing inconsistently due to vis-timeline CDN load times varying from 2s to 30s+. The 10s timeout caused random failures, especially with parallel test execution where each xdist worker started with a cold cache. Changes: - Use persistent browser context with fixed cache directory instead of per-run temp directories, allowing CDN resources to stay cached - Increase vis-timeline load timeout from 10s to 30s as fallback for cold loads - Add .playwright_cache/ to .gitignore With these changes, browser tests pass consistently (28 passed in ~13s vs previously 2-6 random failures per run). 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <[email protected]>
User text messages with the same timestamp but different UUIDs (can occur during git branch switches) were being deduplicated but always kept the first occurrence, regardless of content quality. Now the deduplication keeps the version with the most content items, ensuring IDE notifications and actual user messages are both preserved when split across multiple entries. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <[email protected]>
WalkthroughAdds Playwright cache ignore, enforces a cache-compatibility cutoff at 0.9.0, swaps TypeGuard helpers for optional-casting helpers and updates renderer usage, implements in-place user-text deduplication, adds persistent per-worker Playwright fixtures and longer timeline timeout, removes ty.toml, and updates tests and HTML id/dedup linking. Changes
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes
Possibly related PRs
Poem
Pre-merge checks and finishing touches✅ Passed checks (3 passed)
✨ Finishing touches
🧪 Generate unit tests (beta)
📜 Recent review detailsConfiguration used: defaults Review profile: CHILL Plan: Pro 📒 Files selected for processing (6)
✅ Files skipped from review due to trivial changes (1)
🚧 Files skipped from review as they are similar to previous changes (1)
🧰 Additional context used📓 Path-based instructions (4)test/**/*.py📄 CodeRabbit inference engine (CLAUDE.md)
Files:
**/*.py📄 CodeRabbit inference engine (CLAUDE.md)
Files:
claude_code_log/*.py📄 CodeRabbit inference engine (CLAUDE.md)
Files:
claude_code_log/renderer.py📄 CodeRabbit inference engine (CLAUDE.md)
Files:
🧠 Learnings (5)📓 Common learnings📚 Learning: 2025-11-30T23:24:07.840ZApplied to files:
📚 Learning: 2025-12-09T23:52:47.578ZApplied to files:
📚 Learning: 2025-11-30T17:16:32.495ZApplied to files:
📚 Learning: 2025-11-30T17:16:32.495ZApplied to files:
🧬 Code graph analysis (1)claude_code_log/renderer.py (3)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (10)
🔇 Additional comments (11)
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 |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (1)
pyproject.toml (1)
82-86: Configuration consolidation looks good; use per-file overrides to scope suppression.Moving the ty configuration into
pyproject.tomlis a good practice for consolidating project settings. The comment clearly explains the TypeGuard limitation.However, this global suppression applies to all files and could mask legitimate "possibly-missing-attribute" issues elsewhere in the codebase. ty supports per-file overrides, which would allow more precise scoping. Instead of a global rule, use a
[[tool.ty.overrides]]section with glob patterns to apply different rule configurations to specific files or directories containingis_user_entry()andis_assistant_entry(). Alternatively, use inline suppression comments like# ty: ignore[possibly-missing-attribute]at the specific locations where the false positive occurs.
📜 Review details
Configuration used: defaults
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (9)
.gitignore(1 hunks)claude_code_log/cache.py(1 hunks)claude_code_log/converter.py(3 hunks)pyproject.toml(1 hunks)test/conftest.py(1 hunks)test/test_cache.py(1 hunks)test/test_timeline_browser.py(1 hunks)test/test_version_deduplication.py(1 hunks)ty.toml(0 hunks)
💤 Files with no reviewable changes (1)
- ty.toml
🧰 Additional context used
📓 Path-based instructions (5)
claude_code_log/*.py
📄 CodeRabbit inference engine (CLAUDE.md)
Use dateparser for natural language date parsing to support date range filtering with expressions like 'today', 'yesterday', 'last week', and relative dates
Files:
claude_code_log/cache.pyclaude_code_log/converter.py
**/*.py
📄 CodeRabbit inference engine (CLAUDE.md)
**/*.py: Use ruff for code formatting and linting, with ruff check --fix for automatic fixes
Use pyright and mypy for type checking in Python code
Target Python 3.10+ with support for modern Python features and type hints
Files:
claude_code_log/cache.pytest/test_timeline_browser.pytest/test_cache.pytest/conftest.pytest/test_version_deduplication.pyclaude_code_log/converter.py
test/**/*.py
📄 CodeRabbit inference engine (CLAUDE.md)
Organize tests into categories with pytest markers to avoid async event loop conflicts: unit tests (no mark), TUI tests (@pytest.mark.tui), browser tests (@pytest.mark.browser), and snapshot tests
Files:
test/test_timeline_browser.pytest/test_cache.pytest/conftest.pytest/test_version_deduplication.py
test/**/*browser*.py
📄 CodeRabbit inference engine (CLAUDE.md)
Use Playwright for browser-based testing of interactive HTML features, particularly for testing the timeline visualization and filter controls
Files:
test/test_timeline_browser.py
pyproject.toml
📄 CodeRabbit inference engine (CLAUDE.md)
Use uv for dependency management with commands like 'uv add' to add dependencies and 'uv remove' to remove dependencies
Files:
pyproject.toml
🧠 Learnings (6)
📚 Learning: 2025-11-30T17:16:32.495Z
Learnt from: CR
Repo: daaain/claude-code-log PR: 0
File: CLAUDE.md:0-0
Timestamp: 2025-11-30T17:16:32.495Z
Learning: Applies to test/**/*browser*.py : Use Playwright for browser-based testing of interactive HTML features, particularly for testing the timeline visualization and filter controls
Applied to files:
.gitignoretest/test_timeline_browser.pytest/conftest.py
📚 Learning: 2025-12-09T23:52:47.578Z
Learnt from: daaain
Repo: daaain/claude-code-log PR: 59
File: test/test_cache.py:135-165
Timestamp: 2025-12-09T23:52:47.578Z
Learning: SQLite supports NULLS FIRST and NULLS LAST in ORDER BY since v3.30.0 (Oct 2019). Do not flag SQL that uses these clauses as an error when reviewing Python tests or code that interacts with SQLite. If reviewing SQL strings, verify the target SQLite version supports NULLS FIRST/LAST and ensure the syntax is used correctly for the database in use.
Applied to files:
test/test_timeline_browser.pytest/test_cache.pytest/conftest.pytest/test_version_deduplication.py
📚 Learning: 2025-11-30T17:16:32.495Z
Learnt from: CR
Repo: daaain/claude-code-log PR: 0
File: CLAUDE.md:0-0
Timestamp: 2025-11-30T17:16:32.495Z
Learning: Applies to **/*.py : Use ruff for code formatting and linting, with ruff check --fix for automatic fixes
Applied to files:
pyproject.toml
📚 Learning: 2025-11-30T17:16:32.495Z
Learnt from: CR
Repo: daaain/claude-code-log PR: 0
File: CLAUDE.md:0-0
Timestamp: 2025-11-30T17:16:32.495Z
Learning: Applies to **/*.py : Use pyright and mypy for type checking in Python code
Applied to files:
pyproject.toml
📚 Learning: 2025-11-30T17:16:32.495Z
Learnt from: CR
Repo: daaain/claude-code-log PR: 0
File: CLAUDE.md:0-0
Timestamp: 2025-11-30T17:16:32.495Z
Learning: Applies to **/*.py : Target Python 3.10+ with support for modern Python features and type hints
Applied to files:
pyproject.toml
📚 Learning: 2025-11-30T17:16:32.494Z
Learnt from: CR
Repo: daaain/claude-code-log PR: 0
File: CLAUDE.md:0-0
Timestamp: 2025-11-30T17:16:32.494Z
Learning: Separate TUI tests using run_test() from browser tests using Playwright and unit tests to avoid 'RuntimeError: This event loop is already running' conflicts due to incompatible async frameworks
Applied to files:
test/conftest.py
🧬 Code graph analysis (4)
test/test_timeline_browser.py (1)
test/conftest.py (1)
page(89-93)
test/test_cache.py (1)
claude_code_log/cache.py (1)
_is_cache_version_compatible(411-451)
test/test_version_deduplication.py (3)
claude_code_log/models.py (3)
TextContent(682-684)UserTranscriptEntry(812-816)UserMessage(749-752)claude_code_log/converter.py (1)
deduplicate_messages(310-394)claude_code_log/html/renderer.py (1)
generate_html(305-314)
claude_code_log/converter.py (1)
claude_code_log/models.py (2)
SummaryTranscriptEntry(825-830)UserTranscriptEntry(812-816)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (10)
- GitHub Check: test (windows-latest, 3.11)
- GitHub Check: test (windows-latest, 3.12)
- GitHub Check: test (windows-latest, 3.13)
- GitHub Check: test (ubuntu-latest, 3.13)
- GitHub Check: test (ubuntu-latest, 3.10)
- GitHub Check: test (ubuntu-latest, 3.14)
- GitHub Check: test (windows-latest, 3.14)
- GitHub Check: test (ubuntu-latest, 3.12)
- GitHub Check: test (ubuntu-latest, 3.11)
- GitHub Check: test (windows-latest, 3.10)
🔇 Additional comments (15)
.gitignore (1)
53-53: LGTM!The addition of
.playwright_cache/is appropriate for ignoring the persistent Playwright browser cache directory introduced in test/conftest.py. The placement in the test artifacts section is logical.claude_code_log/converter.py (4)
321-322: LGTM!The docstring addition clearly explains the new deduplication behavior for user text messages with identical timestamps but different UUIDs, addressing the branch switch artifact scenario mentioned in the PR objectives.
332-332: LGTM!The change from a set to a dict mapping dedup_key to list index is necessary to enable in-place replacement of duplicate messages. This allows the deduplicator to replace earlier entries with later, more complete versions.
359-372: LGTM!The logic correctly identifies user text messages by checking for the absence of ToolResultContent. The use of for-else is appropriate here: the else block executes only when no ToolResultContent is found (loop completes without break), correctly setting
is_user_text = Trueand leavingcontent_keyempty for deduplication by timestamp alone.
380-392: LGTM!The in-place replacement logic correctly implements the "keep the version with most content" strategy. The comparison of content item counts ensures that the most complete message is retained regardless of input order. The defensive type check is appropriate, and the logic handles all ordering scenarios correctly.
test/test_timeline_browser.py (1)
54-55: LGTM!The timeout increase from 10s to 30s is appropriate for handling CDN cold loads on first access per xdist worker. The comment clearly explains the rationale, and this aligns with the persistent browser cache implementation that reduces subsequent load times.
test/test_cache.py (1)
572-583: LGTM!The test correctly validates the 0.8.0 → 0.9.0 breaking change rule. The assertions properly verify that:
- Caches from 0.9.0+ are compatible (same or newer than breaking change threshold)
- Caches from 0.8.0 and earlier are invalidated (at or below breaking version when current >= 0.9.0)
This aligns with the breaking change rule added in claude_code_log/cache.py.
claude_code_log/cache.py (1)
424-426: LGTM!The breaking change entry is well-documented with a clear explanation of why cache invalidation is needed. The mapping correctly invalidates caches from 0.8.0 and earlier when upgrading to 0.9.0+, which aligns with the PR's objective of adding a cache-breaking change marker for the IDE tag transformation.
test/test_version_deduplication.py (3)
274-283: LGTM!The test is well-documented with a clear explanation of the git branch switch scenario. The import of TextContent is necessary for constructing the test messages, and using a fixed timestamp is appropriate for testing deduplication behavior.
285-352: LGTM!The three test messages correctly represent the branch switch artifact scenario:
- msg1 with 2 content items (IDE notification + actual text) represents the complete message
- msg2 and msg3 with 1 content item each represent partial versions
- All share the same timestamp but have different UUIDs, matching the real-world scenario
354-370: LGTM!The test logic comprehensively validates the deduplication behavior by testing three different orderings, ensuring the in-place replacement logic works regardless of message order. The assertions correctly verify:
- Deduplication occurred (content appears exactly once)
- The most complete message (msg1) was selected (IDE notification present)
test/conftest.py (4)
42-56: LGTM!The fixture correctly implements per-worker persistent cache directories:
- Session scope is appropriate for sharing across tests
- worker_id parameter enables pytest-xdist parallel execution without lock conflicts
- Using a fixed directory (not temporary) enables caching across test runs
- The path matches the .gitignore entry added in this PR
59-76: LGTM!The persistent context fixture correctly implements HTTP caching for CDN resources:
- Uses
launch_persistent_contextwhich is the appropriate Playwright API- Session scope shares the context across all tests in a worker
- Configuration properly merges launch args with viewport and HTTPS settings
- Context is properly cleaned up using the yield pattern
79-85: LGTM!The context fixture override correctly redirects pytest-playwright's default context to use the persistent context. This ensures all browser tests benefit from the shared HTTP cache while maintaining the standard pytest-playwright API.
88-93: LGTM!The page fixture correctly provides per-test page isolation while using the shared persistent context. Each test gets a fresh page (for proper isolation) but benefits from the shared HTTP cache. The cleanup pattern is correct.
Add explicit isinstance check so pyright can track that message is a UserTranscriptEntry when accessing message.message.content. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <[email protected]>
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (1)
claude_code_log/converter.py (1)
380-392: Implementation looks correct.The replacement logic properly compares content item counts and performs in-place replacement only when beneficial. The
isinstance(existing, UserTranscriptEntry)check at line 385 is a good defensive type guard.One subtle invariant to note: when
is_user_textis True for the current message, the existing entry at the same dedup_key must also be aUserTranscriptEntry(since dedup_key includesmessage_type). The current code handles this correctly, but if this invariant ever breaks, the replacement would silently fail. Consider adding a brief comment or assertion if you want to catch unexpected invariant violations in the future.
📜 Review details
Configuration used: defaults
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
claude_code_log/converter.py(3 hunks)
🧰 Additional context used
📓 Path-based instructions (2)
claude_code_log/*.py
📄 CodeRabbit inference engine (CLAUDE.md)
Use dateparser for natural language date parsing to support date range filtering with expressions like 'today', 'yesterday', 'last week', and relative dates
Files:
claude_code_log/converter.py
**/*.py
📄 CodeRabbit inference engine (CLAUDE.md)
**/*.py: Use ruff for code formatting and linting, with ruff check --fix for automatic fixes
Use pyright and mypy for type checking in Python code
Target Python 3.10+ with support for modern Python features and type hints
Files:
claude_code_log/converter.py
🧠 Learnings (1)
📓 Common learnings
Learnt from: CR
Repo: daaain/claude-code-log PR: 0
File: CLAUDE.md:0-0
Timestamp: 2025-11-30T17:16:32.495Z
Learning: Applies to test/**/*browser*.py : Use Playwright for browser-based testing of interactive HTML features, particularly for testing the timeline visualization and filter controls
🧬 Code graph analysis (1)
claude_code_log/converter.py (1)
claude_code_log/models.py (2)
SummaryTranscriptEntry(825-830)UserTranscriptEntry(812-816)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (10)
- GitHub Check: test (windows-latest, 3.14)
- GitHub Check: test (ubuntu-latest, 3.14)
- GitHub Check: test (windows-latest, 3.13)
- GitHub Check: test (windows-latest, 3.12)
- GitHub Check: test (windows-latest, 3.11)
- GitHub Check: test (ubuntu-latest, 3.12)
- GitHub Check: test (windows-latest, 3.10)
- GitHub Check: test (ubuntu-latest, 3.11)
- GitHub Check: test (ubuntu-latest, 3.10)
- GitHub Check: test (ubuntu-latest, 3.13)
🔇 Additional comments (3)
claude_code_log/converter.py (3)
321-329: LGTM!The updated docstrings clearly document the new deduplication behavior: user text messages with matching timestamps are deduplicated while preserving the entry with the most content items. The in-place replacement semantics are properly explained in the return description.
331-332: LGTM!The change from a set to a dict mapping dedup keys to indices is the correct approach for enabling in-place replacement of existing entries.
356-372: LGTM!The
for-elsepattern cleanly identifies user text messages (those withoutToolResultContent). Using an emptycontent_keyfor these messages ensures they deduplicate by timestamp alone, while theis_user_textflag gates the content-count comparison logic.
|
Not sure why we had one failure: Only for Python 3.10 on Windows, first attempt. Second worked. I can't reproduce it locally. |
Replace TypeGuard[T] with T | None pattern in is_user_entry() and is_assistant_entry(). This enables walrus operator usage and works with both pyright and ty type checkers, allowing removal of the ty rules workaround. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <[email protected]>
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (1)
claude_code_log/parser.py (1)
480-491: Consider renaming functions to reflect non-boolean return.The refactoring from
TypeGuardtoOptional[T]is correct and improves call-site efficiency (as seen inrenderer.py). However, theis_prefix conventionally suggests boolean returns in Python. Names likeget_user_entryoras_user_entrywould better signal that these return typed entries orNone.The implementation itself is sound—the type checks are safe given Pydantic's
Literaltypes, and the walrus-operator usage inrenderer.pyreads naturally.
📜 Review details
Configuration used: defaults
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
claude_code_log/parser.py(2 hunks)claude_code_log/renderer.py(4 hunks)
🧰 Additional context used
📓 Path-based instructions (3)
claude_code_log/*.py
📄 CodeRabbit inference engine (CLAUDE.md)
Use dateparser for natural language date parsing to support date range filtering with expressions like 'today', 'yesterday', 'last week', and relative dates
Files:
claude_code_log/renderer.pyclaude_code_log/parser.py
claude_code_log/renderer.py
📄 CodeRabbit inference engine (CLAUDE.md)
Use mistune for quick Markdown rendering with syntax highlighting support in server-side template rendering
Files:
claude_code_log/renderer.py
**/*.py
📄 CodeRabbit inference engine (CLAUDE.md)
**/*.py: Use ruff for code formatting and linting, with ruff check --fix for automatic fixes
Use pyright and mypy for type checking in Python code
Target Python 3.10+ with support for modern Python features and type hints
Files:
claude_code_log/renderer.pyclaude_code_log/parser.py
🧠 Learnings (1)
📚 Learning: 2025-11-30T17:16:32.495Z
Learnt from: CR
Repo: daaain/claude-code-log PR: 0
File: CLAUDE.md:0-0
Timestamp: 2025-11-30T17:16:32.495Z
Learning: Applies to claude_code_log/models.py : Use Pydantic models for parsing and validating transcript JSON data, including TranscriptEntry (union of UserTranscriptEntry, AssistantTranscriptEntry, SummaryTranscriptEntry), UsageInfo, and ContentItem
Applied to files:
claude_code_log/parser.py
🧬 Code graph analysis (2)
claude_code_log/renderer.py (2)
claude_code_log/parser.py (3)
is_user_entry(480-484)extract_text_content(64-80)is_assistant_entry(487-491)claude_code_log/utils.py (1)
should_use_as_session_starter(110-129)
claude_code_log/parser.py (1)
claude_code_log/models.py (3)
UserTranscriptEntry(812-816)MessageType(17-52)AssistantTranscriptEntry(819-822)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (10)
- GitHub Check: test (ubuntu-latest, 3.12)
- GitHub Check: test (windows-latest, 3.10)
- GitHub Check: test (ubuntu-latest, 3.10)
- GitHub Check: test (ubuntu-latest, 3.13)
- GitHub Check: test (windows-latest, 3.13)
- GitHub Check: test (ubuntu-latest, 3.11)
- GitHub Check: test (windows-latest, 3.11)
- GitHub Check: test (windows-latest, 3.14)
- GitHub Check: test (windows-latest, 3.12)
- GitHub Check: test (ubuntu-latest, 3.14)
🔇 Additional comments (4)
claude_code_log/renderer.py (4)
1800-1804: LGTM!The walrus-operator pattern efficiently combines the entry retrieval and type check, avoiding redundant function calls. Field access on the returned entry is safe and correct.
1821-1828: LGTM!Consistent usage pattern with proper conditional checks. The
elifstructure correctly handles the case where a session already exists but needs its first user message populated.
1839-1842: LGTM!Clean extraction of multiple fields from the returned entry. This pattern is particularly beneficial here where three attributes need to be accessed from the same entry.
2020-2022: LGTM!Consistent application of the walrus-operator pattern for accessing assistant entry fields. The refactoring maintains correctness while reducing redundant function calls throughout the file.
The is_ prefix conventionally implies boolean returns in Python. Rename to as_user_entry() and as_assistant_entry() to better signal that these return typed entries or None. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <[email protected]>
UserTextContent is now instantiated in parser.py and renderer.py, so the TODO claiming it wasn't used is stale. Also use isinstance() for type narrowing in test_ide_tags.py to satisfy ty. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <[email protected]>
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (2)
test/test_ide_tags.py (1)
314-314: Consider adding test coverage for plain text without IDE tags.The assertion correctly validates that
parse_user_message_contentreturnsUserTextContentwhen IDE notifications are present. However, there's no test verifying the behavior when parsing plain user text without any IDE tags.Consider adding a test case like:
def test_parse_user_message_without_ide_tags(self): """Test parsing plain user message without IDE tags.""" plain_text = "This is a simple user message." content_list = [TextContent(type="text", text=plain_text)] content_model = parse_user_message_content(content_list) # Verify the type and that ide_notifications is None assert isinstance(content_model, UserTextContent) assert content_model.ide_notifications is None assert content_model.text == plain_textThis would ensure both code paths are tested and document the expected behavior for plain text inputs.
claude_code_log/renderer.py (1)
1800-1804: Minor redundancy:text_contentis already extracted from the same source.For user entries,
message_content(line 1787) equalsuser_entry.message.content, so re-extracting at line 1803 yields the same value astext_content(line 1788). Consider reusingtext_contentdirectly:if (user_entry := as_user_entry(message)) and should_use_as_session_starter( text_content ): - content = extract_text_content(user_entry.message.content) - first_user_message = create_session_preview(content) + first_user_message = create_session_preview(text_content)
📜 Review details
Configuration used: defaults
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (4)
claude_code_log/models.py(0 hunks)claude_code_log/parser.py(2 hunks)claude_code_log/renderer.py(5 hunks)test/test_ide_tags.py(2 hunks)
💤 Files with no reviewable changes (1)
- claude_code_log/models.py
🚧 Files skipped from review as they are similar to previous changes (1)
- claude_code_log/parser.py
🧰 Additional context used
📓 Path-based instructions (4)
test/**/*.py
📄 CodeRabbit inference engine (CLAUDE.md)
Organize tests into categories with pytest markers to avoid async event loop conflicts: unit tests (no mark), TUI tests (@pytest.mark.tui), browser tests (@pytest.mark.browser), and snapshot tests
Files:
test/test_ide_tags.py
**/*.py
📄 CodeRabbit inference engine (CLAUDE.md)
**/*.py: Use ruff for code formatting and linting, with ruff check --fix for automatic fixes
Use pyright and mypy for type checking in Python code
Target Python 3.10+ with support for modern Python features and type hints
Files:
test/test_ide_tags.pyclaude_code_log/renderer.py
claude_code_log/*.py
📄 CodeRabbit inference engine (CLAUDE.md)
Use dateparser for natural language date parsing to support date range filtering with expressions like 'today', 'yesterday', 'last week', and relative dates
Files:
claude_code_log/renderer.py
claude_code_log/renderer.py
📄 CodeRabbit inference engine (CLAUDE.md)
Use mistune for quick Markdown rendering with syntax highlighting support in server-side template rendering
Files:
claude_code_log/renderer.py
🧠 Learnings (3)
📓 Common learnings
Learnt from: CR
Repo: daaain/claude-code-log PR: 0
File: CLAUDE.md:0-0
Timestamp: 2025-11-30T17:16:32.495Z
Learning: Applies to test/**/*browser*.py : Use Playwright for browser-based testing of interactive HTML features, particularly for testing the timeline visualization and filter controls
📚 Learning: 2025-11-30T17:16:32.495Z
Learnt from: CR
Repo: daaain/claude-code-log PR: 0
File: CLAUDE.md:0-0
Timestamp: 2025-11-30T17:16:32.495Z
Learning: Applies to claude_code_log/models.py : Use Pydantic models for parsing and validating transcript JSON data, including TranscriptEntry (union of UserTranscriptEntry, AssistantTranscriptEntry, SummaryTranscriptEntry), UsageInfo, and ContentItem
Applied to files:
test/test_ide_tags.py
📚 Learning: 2025-12-09T23:52:47.578Z
Learnt from: daaain
Repo: daaain/claude-code-log PR: 59
File: test/test_cache.py:135-165
Timestamp: 2025-12-09T23:52:47.578Z
Learning: SQLite supports NULLS FIRST and NULLS LAST in ORDER BY since v3.30.0 (Oct 2019). Do not flag SQL that uses these clauses as an error when reviewing Python tests or code that interacts with SQLite. If reviewing SQL strings, verify the target SQLite version supports NULLS FIRST/LAST and ensure the syntax is used correctly for the database in use.
Applied to files:
test/test_ide_tags.py
🧬 Code graph analysis (2)
test/test_ide_tags.py (1)
claude_code_log/models.py (2)
TextContent(679-681)UserTextContent(265-275)
claude_code_log/renderer.py (2)
claude_code_log/parser.py (3)
as_assistant_entry(487-491)as_user_entry(480-484)extract_text_content(64-80)claude_code_log/utils.py (1)
should_use_as_session_starter(110-129)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (10)
- GitHub Check: test (windows-latest, 3.12)
- GitHub Check: test (windows-latest, 3.14)
- GitHub Check: test (ubuntu-latest, 3.11)
- GitHub Check: test (windows-latest, 3.10)
- GitHub Check: test (windows-latest, 3.11)
- GitHub Check: test (windows-latest, 3.13)
- GitHub Check: test (ubuntu-latest, 3.13)
- GitHub Check: test (ubuntu-latest, 3.14)
- GitHub Check: test (ubuntu-latest, 3.10)
- GitHub Check: test (ubuntu-latest, 3.12)
🔇 Additional comments (5)
test/test_ide_tags.py (1)
20-21: LGTM!The addition of
UserTextContentto the imports is necessary to support the updated type assertion in the test. Both imports serve distinct purposes:TextContentfor constructing test input andUserTextContentfor validating the parser output type.claude_code_log/renderer.py (4)
43-44: LGTM!The new imports align with the updated entry-based API from the parser module, enabling type-safe access to entry properties.
1821-1828: LGTM!The walrus operator pattern correctly captures the typed entry, and the extracted content is consistently used for both the starter check and preview creation.
1839-1842: LGTM!Clean extraction of assistant entry properties. The subsequent usage check (line 1845) properly handles the case where
usagemay be None.
2020-2022: LGTM!Consistent use of the
as_assistant_entrypattern, matching the approach in_collect_session_info.
The text_content variable is already extracted from message_content, which equals user_entry.message.content for user entries. No need to extract it again. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <[email protected]>
DedupNoticeContent now stores target_uuid and resolves it to target_message_id after hierarchy building. The formatter creates a clickable anchor link to the Task result message. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <[email protected]>
Summary
Background
User text messages with the same timestamp but different UUIDs can occur when resuming a Claude Code session after a git branch switch. Claude Code appears to re-log the user input multiple times, with each entry getting a different UUID but the same timestamp. The content gets split inconsistently across these entries - one might have just the IDE notification (
<ide_opened_file>), another might have just the user's actual text, and one might have both combined.Previously, deduplication would not happen in this case. Now it keeps the version with the most content items, ensuring both IDE notifications and actual user messages are preserved.
Test plan
uv run pytest test/test_version_deduplication.py -vfor new deduplication testsjust test-browserto verify browser test stabilityNotes
The deduplication issue was discovered in session
741790a4-4fe2-4644-9a51-fb4482074060.jsonlfrom the coderabbit-review-helper project (cc @daaain).🤖 Generated with Claude Code
Summary by CodeRabbit
Bug Fixes
Rendering
Tests
Chores
Public API
✏️ Tip: You can customize this high-level summary in your review settings.