Replace direct indexing (self.projects[self.selected_index - 1]) with
.get() to prevent panic if selected_index is somehow out of bounds.
Falls back to "All Projects" scope when the index is invalid instead
of panicking.
Introduce breakpoint-aware helper functions in layout.rs that
centralize per-screen responsive decisions. Each function maps a
Breakpoint to a screen-specific value, replacing scattered
hardcoded checks across view modules:
- detail_side_panel: show discussions side panel at Lg+
- info_screen_columns: 1 column on Xs/Sm, 2 on Md+
- search_show_project: hide project path column on narrow terminals
- timeline_time_width: compact time on Xs (5), full on Md+ (12)
- who_abbreviated_tabs: shorten tab labels on Xs/Sm
- sync_progress_bar_width: scale progress bar 15→50 with width
All functions are const fn with exhaustive match arms.
Includes 6 unit tests covering every breakpoint variant.
Four view modules (search, command_palette, file_history, trace) each had
their own copy of cursor_cell_offset / text_cell_width for converting a
byte-offset cursor position to a display-column offset. Phase 5 introduced
a proper text_width module with these functions; this commit removes the
duplicates and rewires all call sites to use crate::text_width.
- search.rs: removed local text_cell_width + cursor_cell_offset definitions
- command_palette.rs: removed local cursor_cell_offset definition
- file_history.rs: replaced inline chars().count() cursor calc with import
- trace.rs: replaced inline chars().count() cursor calc with import
Complete TUI Phase 3 implementation with all 5 power feature screens:
- Who screen: 5 modes (expert/workload/reviews/active/overlap) with
mode tabs, input bar, result rendering, and hint bar
- Search screen: full-text search with result list and scoring display
- Timeline screen: chronological event feed with time-relative display
- Trace screen: file provenance chains with expand/collapse, rename
tracking, and linked issues/discussions
- File History screen: per-file MR timeline with rename chain display
and discussion snippets
Also includes:
- Command palette overlay (fuzzy search)
- Bootstrap screen (initial sync flow)
- Action layer split from monolithic action.rs to per-screen modules
- Entity and render cache infrastructure
- Shared who_types module in core crate
- All screens wired into view/mod.rs dispatch
- 597 tests passing, clippy clean (pedantic + nursery), fmt clean
- drift.rs: Replace `sorted.sort_by(|a, b| b.1.cmp(&a.1))` with idiomatic
`sort_by_key(|entry| std::cmp::Reverse(entry.1))` (clippy::use_sort_by_key).
- list.rs: Change `query_issues` and `query_mrs` from `fn` to `pub fn` so the
TUI crate can call them directly instead of duplicating the query logic.
- vector.rs: Reformat multi-line assert! macro to satisfy rustfmt check
(body moved onto separate lines with trailing message string).
Move the 14 result structs and enums (WhoResult, ExpertResult, Expert,
ScoreComponents, ExpertMrDetail, WorkloadResult, WorkloadIssue, WorkloadMr,
WorkloadDiscussion, ReviewsResult, ReviewCategory, ActiveResult,
ActiveDiscussion, OverlapResult, OverlapUser) from cli::commands::who into
a new core::who_types module.
The TUI Who screen needs these types to render results, but importing from
the CLI layer would create a circular dependency (TUI -> CLI -> core). By
placing them in core, both the CLI and TUI can depend on them cleanly.
The CLI module re-exports all types via `pub use crate::core::who_types::*`
so existing consumers are unaffected.
Workload and active modes now exclude discussions on closed issues and
merged/closed MRs by default. Adds --include-closed flag to restore
the previous behavior when needed.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When a bare filename like 'operators.ts' matches multiple full paths,
check if they are the same file connected by renames (via BFS on
mr_file_changes). If so, auto-resolve to the newest path instead of
erroring. Also wires path resolution into file-history and trace
commands so bare filenames work everywhere.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Overhaul the sync command's human output to use semantic colors and a
cleaner rendering architecture. The changes fall into four areas:
Stage lines: Replace direct finish_stage() calls with an
emit_stage_line/emit_stage_block pattern that clears the spinner first,
then prints static lines via MultiProgress::suspend. Stage icons are
now color-coded green (success) or yellow (warning) via color_icon().
A separate "Status" stage line now appears after Issues, summarizing
work-item status enrichment across all projects.
Sub-rows: Replace the imperative print_issue_sub_rows/print_mr_sub_rows
functions with functional issue_sub_rows(), mr_sub_rows(), and new
status_sub_rows() that return Vec<String>. Project paths use
Theme::muted(), error/failure counts use Theme::warning(), and
separators use the dim middle-dot style. Sub-rows are printed atomically
with their parent stage line to avoid interleaving with spinners.
Summary: In print_sync(), counts now use Theme::info().bold() for visual
pop, detail-line separators are individually styled (dim middle-dot),
and a new "Sync completed with issues" headline appears when any stage
had failures. Document errors and embedding failures are surfaced in
both the doc-parts line and the errors line.
Tests: Full coverage for append_failures, summarize_status_enrichment,
should_print_timings, issue_sub_rows, mr_sub_rows, and status_sub_rows.
Add -t/--timings flag to the sync subcommand, allowing users to opt
into a per-stage timing breakdown after the sync summary. Wire the flag
through main.rs into print_sync() which passes it to the new
should_print_timings() gate.
Enrich the data structures that flow through the sync pipeline so
downstream renderers have full error visibility:
- ProjectSummary gains status_errors (issue-side status enrichment
failures per project)
- ProjectStatusEnrichment gains path (project path for sub-row display)
- SyncResult gains documents_errored and embedding_failed so the
summary can surface doc-gen and embed failures separately
- Autocorrect table updated with --timings for fuzzy flag matching
Pull the line-formatting logic out of finish_stage() into a standalone
public format_stage_line() so that sync.rs can build stage lines without
needing a live ProgressBar (e.g. for static multi-line blocks printed
after the spinner is cleared).
The new function applies Theme::info().bold() to the label and
Theme::timing() to the elapsed column, giving every stage line
consistent color treatment. finish_stage() now delegates to it.
Includes a unit test asserting the formatted output contains the
expected icon, label, summary, and elapsed components.
Miscellaneous fixes across CLI and core modules:
- Timeline: widen TAG_WIDTH from 10 to 11 to accommodate longer event
type labels without truncation
- render.rs: save and restore LORE_ICONS env var in glyph_mode test to
prevent interference from the test environment leaking into or from
other tests that set LORE_ICONS
- logging.rs: adjust verbose=1 to info level (was debug), verbose=2 to
debug — this reduces noise at -v while keeping -vv as the full debug
experience
- issues.rs, merge_requests.rs: use infodebug! macro consistently for
ingestion summary logging
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add per-project detail rows beneath stage completion lines during multi-project
syncs, showing itemized counts (issues/MRs, discussions, events, statuses, diffs)
for each project. Previously, only aggregate totals were visible, making it hard
to diagnose which project contributed what during a sync.
Status enrichment gets proper progress bars replacing the old spinner-only
display: StatusEnrichmentStarted now carries a total count so the CLI can
render a determinate bar with rate and ETA. The enrichment SQL is tightened
to use IS NOT comparisons for diff-only UPDATEs (skip rows where values
haven't changed), and a follow-up touch_stmt ensures status_synced_at is
updated even for unchanged rows so staleness detection works correctly.
Other improvements:
- New ProjectSummary struct aggregates per-project metrics during ingestion
- SyncResult gains statuses_enriched + per-project summary vectors
- "Already up to date" message when sync finds zero changes
- Remove Arc<AtomicBool> tick_started pattern from docs/embed stages
(enable_steady_tick is idempotent, the guard was unnecessary)
- Progress bar styling: dim spinner, dark_gray track, per_sec + eta display
- Tick intervals tightened from 100ms to 60ms for smoother animation
- statuses_without_widget calculation uses fetch_result.statuses.len()
instead of subtracting enriched (more accurate when some statuses lack
work item widgets)
- Status enrichment completion log downgraded from info to debug
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Extract parse_retry_after_value(header, now) as a pure function to enable
deterministic testing of Retry-After header parsing. The previous
implementation used let-chains with SystemTime::now() inline, which made
it untestable and would panic on negative durations when the server
clock was behind or the header contained a date in the past.
Changes:
- Extract parse_retry_after_value() taking an explicit `now` parameter
- Handle past HTTP dates by returning 1 second instead of panicking on
negative Duration (date.duration_since(now) returns Err for past dates)
- Trim whitespace from header values before parsing
- Add test for past HTTP date returning 1 second minimum
- Add test for delta-seconds with surrounding whitespace
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Phase 7 cleanup: migrate timeline.rs and main.rs search spinner
from stage_spinner() to stage_spinner_v2() with proper icon labels,
then remove the now-unused stage_spinner() function and its tests.
No external callers remain for the old numbered-stage API.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Phase 6 of the UX overhaul. Applies consistent visual treatment across
the remaining command outputs: stats, doctor, timeline, who, count,
and drift.
Stats (stats.rs):
- Apply render::format_number() to all numeric values (documents,
FTS indexed, embedding counts, chunks) for thousand-separator
formatting in large databases
Doctor (doctor.rs):
- Replace Unicode check/warning/cross symbols with Icons::success(),
Icons::warning(), Icons::error() for glyph-mode awareness
- Add summary line after checks showing "Ready/Not ready" with counts
of passed, warnings, and failed checks separated by middle dots
- Remove "lore doctor" title header for cleaner output
Count (count.rs):
- Right-align numeric values with {:>10} format for columnar output
in count and state breakdown displays
Timeline (timeline.rs):
- Add entity icons (issue/MR) before entity references in event rows
- Refactor format_event_tag to pad plain text before applying style,
preventing ANSI codes from breaking column alignment
- Extract style_padded() helper for width-then-style pattern
Who (who.rs):
- Add Icons::user() before usernames in expert, workload, reviews,
and overlap displays
- Replace manual bold section headers with render::section_divider()
in workload view (Assigned Issues, Authored MRs, Reviewing MRs,
Unresolved Discussions)
Drift (drift.rs):
- Add Icons::error()/success() before drift detection status line
- Replace '#' bar character with Unicode full block for similarity
curve visualization
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Phase 4 of the UX overhaul. Restructures the show issue and show MR
detail displays with consistent section layout, state icons, and
improved typography.
Issue detail changes:
- Replace bold header + box-drawing underline with indented title using
Theme::bold() for the title text only
- Organize fields into named sections using render::section_divider():
Details, Development, Description, Discussions
- Add state icons (Icons::issue_opened/closed) alongside text labels
- Add relative time in parentheses next to Created/Updated dates
- Switch labels from "Labels: (none)" to only showing when present,
using format_labels_bare for clean comma-separated output
- Move URL and confidential indicator into Details section
- Closing MRs show state-colored icons (merged/opened/closed)
- Discussions use section_divider instead of bold text, remove colons
from author lines, adjust wrap widths for consistent indentation
MR detail changes:
- Same section-divider layout: Details, Description, Discussions
- State icons for opened/merged/closed using Icons::mr_* helpers
- Draft indicator uses Icons::mr_draft() instead of [Draft] text prefix
- Relative times added to Created, Updated, Merged, Closed dates
- Reviewers and Assignees fields aligned with fixed-width labels
- Labels shown only when present, using format_labels_bare
- Discussion formatting matches issue detail style
Both views use 5-space left indent for field alignment and consistent
wrap widths (72 for descriptions, 68/66 for discussion notes/replies).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Phase 3 of the UX overhaul. Enhances the issues, merge requests, and
notes list displays with visual indicators and improved formatting.
List display changes (src/cli/commands/list.rs):
- Add state icons to issues (opened/closed) and merge requests
(opened/merged/closed) using Icons:: helpers alongside text labels
- Replace [DRAFT] prefix with Icons::mr_draft() glyph for draft MRs
- Switch from format_relative_time to format_relative_time_compact for
tighter column widths in tabular output
- Switch from format_labels to format_labels_bare for unlabeled style
- Change format_discussions() return type from String to StyledCell so
unresolved counts render with Theme::warning() color inline
- Bold the section headers ("Issues", "Merge Requests", "Notes")
with count separated from the label for cleaner scanning
- Import Icons from render module
Test updates (src/cli/commands/list_tests.rs):
- Update format_discussions tests to assert on StyledCell.text field
instead of raw String, since the function now returns styled output
- The unresolved-count test checks starts_with/contains to handle
embedded ANSI escape codes from Theme::warning()
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Phase 2 of the UX overhaul. Replaces the old numbered-stage progress
system (1/4, 2/4...) and manual indicatif ProgressBar/ProgressStyle
setup with the new centralized progress helpers.
Sync command changes (src/cli/commands/sync.rs):
- Replace stage_spinner(n, total, msg) with stage_spinner_v2(icon, label, status)
removing the rigid numbered-stage counter in favor of named stages
- Replace manual ProgressBar::new + ProgressStyle::default_bar for docs
and embed sub-progress with nested_progress(label, len, robot_mode)
- Add finish_stage() calls that display a completion summary with
elapsed time, e.g. "Issues 42 issues from 3 projects 1.2s"
- Each stage (Issues, MRs, Docs, Embed) now reports what it did on
completion rather than just clearing the spinner silently
- Embed failure path uses Icons::warning() instead of inline Theme
formatting, keeping error display consistent with success path
- Remove indicatif direct dependency from sync.rs (now handled by
progress module)
Main entry point changes (src/main.rs):
- Add GlyphMode detection: auto-detect Unicode/Nerd Font support or
fall back to ASCII based on --icons flag, --color=never, NO_COLOR,
or robot mode
- Update all LoreRenderer::init() calls to pass GlyphMode alongside
ColorMode for icon-aware rendering throughout the CLI
- Overhaul handle_error() formatting: use Icons::error() glyph,
bold error text, arrow prefixed action suggestions, and breathing
room with blank lines for scannability
- Migrate handle_embed() progress bar from manual ProgressBar +
ProgressStyle to nested_progress() helper, matching sync command
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Phase 5 of the UX overhaul. Migrates search result display from raw
console styling to the centralized Theme system with semantic methods,
improving visual consistency and readability.
Search result changes:
- Type badges now use semantic styles (issue_ref, mr_ref) with
fixed-width alignment for clean columnar layout
- Snippet rendering uses Theme::highlight() for matched terms and
Theme::muted() for surrounding context, replacing bold+underline
- Metadata line uses Theme::username() for authors and per-part
styling with middle-dot separators instead of a single dim line
- Result numbering uses muted style with right-aligned width
- Consistent 8-space indent for metadata, snippets, and explain lines
- Header line uses muted style for search mode instead of dim+parens
- Trailing blank line moved after the result loop instead of per-result
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Update test_raw_mode_leading_wildcard_falls_back_to_safe to match the
actual Safe mode behavior: OR is a recognized FTS5 boolean operator and
passes through unquoted, so the expected output is '"*" OR "auth"' not
'"*" "OR" "auth"'. The previous assertion was incorrect since the Safe
mode operator-passthrough logic was added.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Invert the timeline mention-expansion flag semantics. Previously, mention
edges were excluded by default and --expand-mentions opted in. Now mention
edges are included by default (matching the more common use case) and
--no-mentions opts out to reduce fan-out when needed.
This is a breaking CLI change but aligns with the principle that the
default behavior should produce the most useful output. Users who were
passing --expand-mentions get the same behavior without any flag. Users
who want reduced output can pass --no-mentions.
Updated: CLI args (TimelineArgs), autocorrect flag list, robot-docs
schema, README documentation and flag reference table.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Wire the LoreRenderer singleton initialization into main.rs color mode
handling, replacing the console::style import with Theme throughout.
Key changes:
- Color initialization: LoreRenderer::init() called for all code paths
(NO_COLOR, --color never/always/auto, unknown mode fallback) alongside
the existing console::set_colors_enabled() calls. Both systems must
agree since some transitive code still uses console (e.g. dialoguer).
- Tracing: Replace .with_target(false) with .event_format(CompactHumanFormat)
for the stderr layer, producing the clean 'HH:MM:SS LEVEL message' format.
- Error handling: handle_error() now shows machine-actionable recovery
commands from gi_error.actions() below the hint, formatted with dim '$'
prefix and bold command text.
- Deprecation warnings: All 'lore list', 'lore show', 'lore auth-test',
'lore sync-status' warnings migrated to Theme::warning().
- Init wizard: All success/info/error messages migrated. Unicode check
marks use explicit \u{2713} escapes instead of literal symbols.
- Embed command: Added progress bar with indicatif for embedding stage,
showing position/total with steady tick. Elapsed time shown on completion.
- Generate-docs and ingest commands: Added 'Done in Xs' elapsed time and
next-step hints (run embed after generate-docs, run generate-docs after
ingest) for better workflow guidance.
- Sync output: Interrupt message and lock release migrated to Theme.
- Health command: Status labels and overall healthy/unhealthy styled.
- Robot-docs: Added drift command schema, updated sync flags to include
--no-file-changes, updated who flags with new options.
- Timeline --expand-mentions -> --no-mentions flag rename wired through
params and robot-docs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace all console::style() calls in command modules with the centralized
Theme API and render:: utility functions. This ensures consistent color
behavior across the entire CLI, proper NO_COLOR/--color never support via
the LoreRenderer singleton, and eliminates duplicated formatting code.
Changes per module:
- count.rs: Theme for table headers, render::format_number replacing local
duplicate. Removed local format_number implementation.
- doctor.rs: Theme::success/warning/error for check status symbols and
messages. Unicode escapes for check/warning/cross symbols.
- drift.rs: Theme::bold/error/success for drift detection headers and
status messages.
- embed.rs: Compact output format — headline with count, zero-suppressed
detail lines, 'nothing to embed' short-circuit for no-op runs.
- generate_docs.rs: Same compact pattern — headline + detail + hint for
next step. No-op short-circuit when regenerated==0.
- ingest.rs: Theme for project summaries, sync status, dry-run preview.
All console::style -> Theme replacements.
- list.rs: Replace comfy-table with render::LoreTable for issue/MR listing.
Remove local colored_cell, colored_cell_hex, format_relative_time,
truncate_with_ellipsis, and format_labels (all moved to render.rs).
- list_tests.rs: Update test assertions to use render:: functions.
- search.rs: Add render_snippet() for FTS5 <mark> tag highlighting via
Theme::bold().underline(). Compact result layout with type badges.
- show.rs: Theme for entity detail views, delegate format_date and
wrap_text to render module.
- stats.rs: Section-based layout using render::section_divider. Compact
middle-dot format for document counts. Color-coded embedding coverage
percentage (green >=95%, yellow >=50%, red <50%).
- sync.rs: Compact sync summary — headline with counts and elapsed time,
zero-suppressed detail lines, visually prominent error-only section.
- sync_status.rs: Theme for run history headers, removed local
format_number duplicate.
- timeline.rs: Theme for headers/footers, render:: for date/truncate,
standard format! padding replacing console::pad_str.
- who.rs: Theme for all expert/workload/active/overlap/review output
modes, render:: for relative time and truncation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Migrate all ingestion completion logs to use nonzero_summary() for compact,
zero-suppressed output. Before: 8-14 individual key=value structured fields
per completion message. After: a single summary field like
'42 fetched · 3 labels · 12 notes' that only shows non-zero counters.
Also downgrade all 'Shutdown requested...' messages from info! to debug!.
These are emitted on every Ctrl+C and add noise to the partial results
output that immediately follows. They remain visible at -vv for debugging
graceful shutdown behavior.
Affected modules:
- issues.rs: issue ingestion completion
- merge_requests.rs: MR ingestion completion, full-sync cursor reset
- mr_discussions.rs: discussion ingestion completion
- orchestrator.rs: project-level issue and MR completion summaries,
all shutdown-requested checkpoints across discussion sync, resource
events drain, closes-issues drain, and MR diffs drain
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Three quality-of-life improvements to reduce log noise and improve readability:
1. logging.rs: Add CompactHumanFormat for stderr tracing output. Replaces the
default format with a minimal 'HH:MM:SS LEVEL message key=value' layout —
no span context, no full timestamps, no target module. The JSON file log
layer is unaffected. This makes watching 'lore sync' output much cleaner.
2. lock.rs: Downgrade AppLock acquire/release messages from info! to debug!.
Lock lifecycle events (acquired new, acquired existing, released) are
operational bookkeeping that clutters normal output. They remain visible
at -vv verbosity for troubleshooting.
3. ingestion/mod.rs: Add nonzero_summary() utility that formats named counters
as a compact middle-dot-separated string, suppressing zero values. Produces
output like '42 fetched · 3 labels · 12 notes' instead of verbose key=value
structured fields. Returns 'nothing to update' when all values are zero.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Introduce src/cli/render.rs as the single source of truth for all terminal
output styling and formatting utilities. Key components:
- LoreRenderer: global singleton initialized once at startup, resolving
color mode (Auto/Always/Never) against TTY state and NO_COLOR env var.
This fixes lipgloss's limitation of hardcoded TrueColor rendering by
gating all style application through a colors_on() check.
- Theme: semantic style constants (success/warning/error/info/accent,
entity refs, state colors, structural styles) that return plain
Style::new() when colors are disabled. Replaces ad-hoc console::style()
calls scattered across 15+ command modules.
- Shared formatting utilities consolidated from duplicated implementations:
format_relative_time (was in list.rs and who.rs), format_number (was in
count.rs and sync_status.rs), truncate (was truncate_with_ellipsis in
list.rs and truncate_summary in timeline.rs), format_labels, format_date,
wrap_indent, section_divider.
- LoreTable: lightweight table renderer replacing comfy-table with simple
column alignment (Left/Right/Center), adaptive terminal width, and
NO_COLOR-safe output.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Switch from comfy-table to the lipgloss Rust port for terminal styling.
lipgloss provides a composable Style API better suited to our new semantic
theming approach (Theme::success(), Theme::error(), etc.) where we apply
styles to individual text spans rather than constructing styled table cells.
The comfy-table dependency was only used by the list command's human output
and is no longer needed.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Two related improvements to agent ergonomics in main.rs:
1. suggest_similar_command now matches against aliases (issue->issues,
mr->mrs, find->search, stat->stats, note->notes, etc.) and provides
contextual usage examples via a new command_example() helper, so
agents get actionable recovery hints like "Did you mean 'lore mrs'?
Example: lore --robot mrs -n 10" instead of just the command name.
2. robot-docs now includes an error_tolerance section documenting every
auto-correction the CLI performs: types (single_dash_long_flag,
case_normalization, flag_prefix, fuzzy_flag, subcommand_alias,
value_normalization, value_fuzzy, prefix_matching), examples, and
mode behavior (threshold differences). Also expands the aliases
section with command_aliases and pre_clap_aliases maps for complete
agent self-discovery.
Together these ensure agents can programmatically discover and recover
from any CLI input error without human intervention.
Three-phase pipeline replacing the single-pass correction:
- Phase A: Subcommand alias correction — handles forms clap can't
express (merge_requests, mergerequests, robotdocs, generatedocs,
gen-docs, etc.) via case-insensitive alias map lookup.
- Phase B: Per-arg flag corrections — adds unambiguous prefix expansion
(--proj -> --project) alongside existing single-dash, case, and fuzzy
rules. New FlagPrefix rule with 0.95 confidence.
- Phase C: Enum value normalization — auto-corrects casing, prefixes,
and typos for flags with known valid values. Handles both --flag value
and --flag=value forms. Respects POSIX -- option terminator.
Changes strict/robot mode from disabling fuzzy matching entirely to using
a higher threshold (0.9 vs 0.8), still catching obvious typos like
--projct while avoiding speculative corrections that mislead agents.
New CorrectionRule variants: SubcommandAlias, ValueNormalization,
ValueFuzzy, FlagPrefix. Each has a corresponding teaching note.
Comprehensive test coverage for all new correction types including
subcommand aliases, value normalization (case, prefix, fuzzy, eq-form),
flag prefix (ambiguous rejection, eq-value preservation), and updated
strict mode behavior.
- Fix MENTIONED_RE/CLOSED_BY_RE to match real GitLab format
('mentioned in issue #N' / 'mentioned in merge request !N')
- Add GITLAB_URL_RE + parse_url_refs() for full URL extraction
- Add extract_refs_from_descriptions() -> source_method='description_parse'
- Add extract_refs_from_user_notes() -> source_method='note_parse'
- Wire both into orchestrator after system note extraction
- 36 tests: regex fix, URL parsing, integration, idempotency
Adds issue:N / i:N / mr:N / m:N query syntax to bypass hybrid search
and seed the timeline directly from a known entity. All discussions for
the entity are gathered without needing Ollama.
- parse_timeline_query() detects entity-direct patterns
- resolve_entity_by_iid() resolves IID to EntityRef with ambiguity handling
- seed_timeline_direct() gathers all discussions for the entity
- 20 new tests (5 resolve, 6 direct seed, 9 parse)
- Updated CLI help text and robot-docs manifest
Moves the conn.prepare() call for fetching discussion notes outside the
per-discussion loop in collect_discussion_threads(). The SQL is identical
for every iteration, so preparing it once and rebinding parameters avoids
redundant statement compilation on each matched discussion.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
FTS5 boolean operators (AND, OR, NOT, NEAR) are case-sensitive uppercase
keywords that must appear unquoted in the query string. Previously, the
user-friendly query builder would double-quote every token, causing
queries like "switch AND health" to search for the literal word "AND"
instead of using it as a boolean conjunction.
Adds a FTS5_OPERATORS constant and checks each token against it before
quoting, allowing natural boolean search syntax to work as expected.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
