Add a Claude Code session-transcript adapter#8
Open
ejentum wants to merge 2 commits into
Open
Conversation
Adds skillloop/adapters/claude_code.py and wires `ingest claude-code`, so SkillLoop can ingest Claude Code sessions alongside the generic and hermes adapters. Claude Code stores sessions as JSONL under ~/.claude/projects/<slug>/*.jsonl with a nested envelope and block-array content, so the generic adapter cannot read them. The adapter normalizes that into AgentTrace: - two-pass parse matching tool_use blocks to their tool_result across turns, populating ToolCall name/arguments/result/status/started_at/ended_at/duration_ms - flattens text blocks into content; tolerates a partial trailing line on a live session; drops meta lines and pure tool_result turns - --no-sidechains to exclude subagent sidechain turns - forward-compatible extended-thinking capture (Claude Code persists thinking blocks with empty text + a signature, so the reasoning text is stripped) 6 tests added; full suite green. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (2)
🚧 Files skipped from review as they are similar to previous changes (1)
📝 WalkthroughWalkthroughThis PR introduces a Claude Code session adapter for SkillLoop that parses JSONL session transcripts into structured ChangesClaude Code Session Adapter
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~22 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)
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: 1
🧹 Nitpick comments (1)
tests/test_claude_code.py (1)
55-63: ⚡ Quick winAdd a regression test for malformed non-trailing JSON lines.
Line 55 onward covers empty sessions, but there’s no assertion that a malformed middle JSONL line raises an error instead of being silently dropped. This protects the parsing contract and prevents future silent data loss regressions.
🧪 Suggested test
+import pytest + def test_claude_code_adapter_rejects_empty_session(tmp_path): @@ raise AssertionError("expected ValueError for a session with no usable messages") + + +def test_claude_code_adapter_rejects_malformed_non_trailing_line(tmp_path): + session = tmp_path / "malformed_mid.jsonl" + session.write_text( + "\n".join( + [ + json.dumps({"message": {"role": "user", "content": [{"type": "text", "text": "hi"}]}}), + '{"message": ', # malformed non-trailing line + json.dumps({"message": {"role": "assistant", "content": [{"type": "text", "text": "bye"}]}}), + ] + ) + + "\n", + encoding="utf-8", + ) + with pytest.raises(ValueError): + load_claude_code_session(session)🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tests/test_claude_code.py` around lines 55 - 63, Add a regression test that ensures load_claude_code_session raises on a malformed non-trailing JSONL line: create a temp session file via _write_session containing a valid message, then a malformed JSON line (e.g., truncated or invalid JSON), then another valid message, and assert that calling load_claude_code_session(session) raises ValueError (similar to test_claude_code_adapter_rejects_empty_session but specifically exercising a malformed middle line); reference the test name test_claude_code_adapter_rejects_empty_session and the loader function load_claude_code_session when locating where to add the new test.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@skillloop/adapters/claude_code.py`:
- Around line 69-76: The loop currently swallows every JSONDecodeError for any
line (iterating raw_text -> raw), but we should only tolerate a malformed JSON
on the final non-empty line; change the loop to detect whether the current `raw`
is the last non-empty line (e.g., pre-split `raw_text.splitlines()` into a list
and use index/enumeration or compute remaining non-empty lines) and in the
except json.JSONDecodeError block only `continue` when there are no further
non-empty lines; otherwise re-raise (or propagate) the error so corrupted middle
lines are not silently dropped. Reference the variables and loop using
`raw_text`, `raw`, `parsed` and the JSON parse try/except surrounding
`json.loads(raw)`.
---
Nitpick comments:
In `@tests/test_claude_code.py`:
- Around line 55-63: Add a regression test that ensures load_claude_code_session
raises on a malformed non-trailing JSONL line: create a temp session file via
_write_session containing a valid message, then a malformed JSON line (e.g.,
truncated or invalid JSON), then another valid message, and assert that calling
load_claude_code_session(session) raises ValueError (similar to
test_claude_code_adapter_rejects_empty_session but specifically exercising a
malformed middle line); reference the test name
test_claude_code_adapter_rejects_empty_session and the loader function
load_claude_code_session when locating where to add the new test.
🪄 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: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: 774ee182-7695-4351-af9b-b5977d7a6f3f
📒 Files selected for processing (3)
skillloop/adapters/claude_code.pyskillloop/cli.pytests/test_claude_code.py
|
Nice adapter. One preservation detail I would keep even if The most useful fields are It also helps render subagent sidechains later without losing the original transcript topology. Generated with ax. |
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.
Claude Code session-transcript adapter
Adds an adapter so SkillLoop can ingest Claude Code sessions, alongside the existing
genericandhermesadapters.Claude Code persists sessions as JSONL under
~/.claude/projects/<slug>/*.jsonlwith a nested envelope ({message: {role, content: [blocks]}}), block-array content (text/tool_use/tool_result), and interleaved non-message meta lines, so thegenericadapter cannot read it. This normalizes that intoAgentTrace:tool_resultblocks and matches them back to the originatingtool_useacross turns, populatingToolCallname / arguments / result / status /started_at/ended_at/duration_ms.textblocks intocontent; tolerates a partial trailing JSON line on a live session; drops meta lines and pure-tool_resultturns.--no-sidechainsflag to exclude subagent sidechain turns.thinkingfield plus asignature, so the reasoning text is stripped from the transcript and there is nothing to recover from a Claude Code session (documented in code).Usage
Tests
6 new tests in
tests/test_claude_code.py, verified end to end against a real ~860-message session.Heads-up (separate from this PR): on Windows,
test_store.py::test_store_preserves_raw_trace_and_hashesfails becauseraw_artifact_refis stored withos.sep(backslashes) while the test asserts forward slashes. Happy to send a one-line fix as a follow-up.🤖 Generated with Claude Code
Summary by CodeRabbit
ingestcommand to load either the latest session (scoped by project) or a provided input file, with options to control sidechain inclusion.