Merge MCP server into the tempyr binary#2
Merged
Conversation
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: ASSERTIVE Plan: Pro Run ID: ⛔ Files ignored due to path filters (1)
📒 Files selected for processing (10)
💤 Files with no reviewable changes (1)
📝 WalkthroughWalkthroughThe changes convert Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 2 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Comment |
This was referenced Apr 24, 2026
cleak
added a commit
that referenced
this pull request
Apr 28, 2026
…on-miss in journal_get Three findings, all valid: 1. **`ingest_open_file` slurped the whole file**: `std::fs::read(path)` on a JSONL that's been growing across many agent invocations could OOM the indexer on very large sessions. Switched to `File::open` + `BufReader` + `seek(SeekFrom::Start(start))` and `read_until(b'\n')` to stream line-by-line, keeping memory bounded to one line buffer at a time. Partial-trailing-line behavior is preserved exactly: if the last line is missing its `\n` (writer mid-append), we stop *before* it and don't advance the offset, so the next refresh picks it up once the writer adds the newline. 2. **`read_last_offset` / `read_last_sha` swallowed all DB errors** via `.ok().flatten()`, hiding real problems (corrupt `indexer_state` table, locked db, etc.) behind silent re-ingestion. Switched to explicit match: `QueryReturnedNoRows` → `Ok(0)` / `Ok(None)` for first-time scans; everything else propagates. Distinguishes "no row yet" from real errors so a corrupted state surfaces instead of looking like a healthy fresh scan. 3. **`journal_get` returned null on cache miss without retrying**: an agent that called `journal_log` and then immediately called `journal_get(id)` would get null, because the indexer is the only writer and `journal_log` only appends to the JSONL. Fixed by adding a single refresh-on-miss retry. Re-opening the db alone wouldn't help — the index is the same. Calling `refresh_index` first ingests the new line, then the second `get_entry` hits. Bounded to exactly one retry; a still-missing id after refresh is a legitimate result (cross-machine entry not yet fetched). Two new regression tests: - `partial_trailing_line_is_left_for_next_refresh`: writes a complete line + a partial line (no `\n`), verifies refresh #1 ingests only the complete one and `j-partial` is absent. Then appends `\n` and verifies refresh #2 picks up the now-complete line. - `read_last_offset_returns_zero_for_first_time_scan`: pins the happy-path behavior of the new error-handling path so a future regression can't silently swallow errors again. Tests: 545 workspace-wide (+2 net). Clippy + fmt clean. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
4 tasks
cleak
added a commit
that referenced
this pull request
Apr 29, 2026
feat: cross-encoder reranking on journal search (v2 backlog #2)
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
tempyrbinary behind an earlytempyr --mcplaunch pathtempyr-mcpfrom a standalone binary crate into an internal library withserve_stdio()Details
tempyr-clinow inspects raw args before clap parsing and routes--mcpinto MCP stdio mode--mcpare rejected instead of being ignoredtempyr-mcpexposes library entrypoints andTempyrServernow implementsDefaultcrates/tempyr-mcp/src/main.rsbinary entrypoint is removedtempyr-mcpas a library launched viatempyr --mcpVerification
cargo clippy -p tempyr-cli -p tempyr-mcp --all-targetscargo test --workspacecargo build --release -p tempyr-cli -F tempyr-index/local-embeddingsPackaging result
tempyr.exetempyr.exe + tempyr-mcp.exefootprint of64,291,328bytes to a singletempyr.exeat37,658,624bytesSummary by CodeRabbit
New Features
--mcpstartup mode to launch the MCP server over standard I/O.Documentation
tempyr --mcpoption.