refactor(native): extract shared session-adoption parsers from live forwarders#1973
Draft
apple-techie wants to merge 2 commits into
Draft
refactor(native): extract shared session-adoption parsers from live forwarders#1973apple-techie wants to merge 2 commits into
apple-techie wants to merge 2 commits into
Conversation
…orwarders Slice 1 (foundation) of seamless external-session adoption. Pure refactor with ZERO behavior change: factor the native-record -> Omnigent-item translation logic out of the three live forwarders into standalone, side-effect-free modules so a later adoption path can reuse the exact forwarder parsers instead of a second, simplified transcript parser. New modules (no IO / DB / network — pure translation): - omnigent/hermes_native_items.py: HermesMirrorItem, message_to_items, assistant_row_has_tool_calls (from hermes_native_forwarder). - omnigent/claude_transcript_parser.py: ClaudeTranscriptItem, TranscriptReadResult, the JSONL record reader, read_transcript_items_* readers, _transcript_items_from_entry and all subhelpers, and per-entry usage/model extraction (from claude_native_bridge). - omnigent/codex_native_items.py: terminal-error derivation (_terminal_error_from_turn and friends), resume turn -> status derivation, built-in tool-call normalization, turn/source ids, and a new pure iter_resume_items() helper (from codex_native_forwarder). Each live forwarder now imports the extracted helpers and re-exports them under their original names (redundant `as` aliases) so every existing caller and test keeps importing them unchanged. codex _replay_resume_response() now drives the replay loop through iter_resume_items(), the same normalization path adoption will use. Behavior is byte-identical: all existing forwarder/bridge suites pass unmodified (741 across hermes/claude/codex native suites). Adds tests/test_native_adoption_parsers.py characterizing the extracted helpers directly from the new modules. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Cross-review found a missing-re-export gap: `_ATTACHMENT_MARKER_RE` and `_SKILL_INVOKE_RE` were moved from hermes_native_forwarder into hermes_native_items but were not restored as re-exports, so `from omnigent.hermes_native_forwarder import _ATTACHMENT_MARKER_RE` (and attribute access) would break — a violation of the zero-behaviour transition contract that every moved symbol stays importable under its original name. - Re-export both regexes from hermes_native_forwarder via `as`-aliases. - Swept every import site + attribute access of the three original modules (omnigent/ and tests/); confirmed all 5 hermes / 37 claude / 29 codex moved symbols (+ the new iter_resume_items) resolve from their original module. Claude/Codex were already complete (blanket `as`-alias re-export); only the two Hermes regexes were missing. - Add parametrized re-export tests covering all moved symbols to lock the moved-without-alias regression class. Repr note: kept the design's public class name `HermesMirrorItem` with the `_MirrorItem = HermesMirrorItem` alias (no caller/test depends on repr); did not rename the runtime class back. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.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
1 participant
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
Foundation slice for session adoption (importing existing native harness sessions into Omnigent conversations — see
designs/session-adoption.md). This PR is a pure refactor with zero behavior change: it extracts the inline "native record → Omnigent conversation item" translation out of the three live forwarders into shared, side-effect-free modules, so that adoption can reuse the exact same parsers the live forwarders use (rather than reimplementing them).What changed
Extracted pure parsers, rewrote each forwarder to call them, and re-exported every moved symbol under its original name so existing callers/tests are untouched:
omnigent/hermes_native_items.pyhermes_native_forwarder.pyHermesMirrorItem,message_to_items,assistant_row_has_tool_callsomnigent/claude_transcript_parser.pyclaude_native_bridge.pyread_transcript_items_from_offset/_since/_since_with_position,read_assistant_text_since, JSONL record readers, entry→items builders, usage/model extractionomnigent/codex_native_items.pycodex_native_forwarder.pyiter_resume_items(response)codex_native_forwarder._replay_resume_responsenow drives replay throughiter_resume_items(the same normalization path adoption will use). The new modules are pure — no IO/DB/network/polling (Claude's transcript file reads are the one inherent exception).Verification
tests/test_native_adoption_parsers.py).No behavior change; no existing test modified. Later slices (server adoption API, per-harness discover/adopt/resume, web UI) build on this.
🤖 Generated with Claude Code