ConfigNotFound previously used exit code 2 which collides with clap's
usage error code. Remap it to exit 20 to avoid ambiguity. Also add
dedicated NotFound (exit 17) and Ambiguous (exit 18) error codes with
proper ErrorCode variants and Display implementations, replacing the
previous incorrect mapping of these errors to GitLabNotFound.
Co-Authored-By: Claude (us.anthropic.claude-opus-4-5-20251101-v1:0) <noreply@anthropic.com>
Extends the CLI with six new commands that complete the search pipeline:
- lore search <QUERY>: Hybrid search with mode selection (lexical,
hybrid, semantic), rich filtering (--type, --author, --project,
--label, --path, --after, --updated-after), result limits, and
optional explain mode showing RRF score breakdowns. Safe FTS mode
sanitizes user input; raw mode passes through for power users.
- lore stats: Document and index statistics with optional --check
for integrity verification and --repair to fix inconsistencies
(orphaned documents, missing FTS entries, stale dirty queue items).
- lore embed: Generate vector embeddings via Ollama. Supports
--retry-failed to re-attempt previously failed embeddings.
- lore generate-docs: Drain the dirty queue to regenerate documents.
--full seeds all entities for complete rebuild. --project scopes
to a single project.
- lore sync: Full pipeline orchestration (ingest issues + MRs,
generate-docs, embed) with --no-embed and --no-docs flags for
partial runs. Reports per-stage results and total elapsed time.
- lore health: Quick pre-flight check (config exists, DB exists,
schema current). Returns exit code 1 if unhealthy. Designed for
agent pre-flight scripts.
- lore robot-docs: Machine-readable command manifest for agent
self-discovery. Returns all commands, flags, examples, exit codes,
and recommended workflows as structured JSON.
Also enhances lore init with --gitlab-url, --token-env-var, and
--projects flags for fully non-interactive robot-mode initialization.
Fixes init's force/non-interactive precedence logic and adds JSON
output for robot mode.
Updates all command files for the GiError -> LoreError rename.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Integrates the dirty tracking system into all four ingestion paths
(issues, MRs, issue discussions, MR discussions). After each entity
is upserted within its transaction, a corresponding dirty_queue entry
is inserted so the document regenerator knows which documents need
rebuilding.
This ensures that document generation stays transactionally consistent
with data changes: if the ingest transaction rolls back, the dirty
marker rolls back too, preventing stale document regeneration attempts.
Also updates GiError references to LoreError in these files as part
of the codebase-wide rename, and adjusts issue discussion logging
from info to debug level to reduce noise during normal sync runs.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Implements the search module providing three search modes:
- Lexical (FTS5): Full-text search using SQLite FTS5 with safe query
sanitization. User queries are automatically tokenized and wrapped
in proper FTS5 syntax. Supports a "raw" mode for power users who
want direct FTS5 query syntax (NEAR, column filters, etc.).
- Semantic (vector): Embeds the search query via Ollama, then performs
cosine similarity search against stored document embeddings. Results
are deduplicated by doc_id since documents may have multiple chunks.
- Hybrid (default): Executes both lexical and semantic searches in
parallel, then fuses results using Reciprocal Rank Fusion (RRF) with
k=60. This avoids the complexity of score normalization while
producing high-quality merged rankings. Gracefully degrades to
lexical-only when embeddings are unavailable.
Additional components:
- search::filters: Post-retrieval filtering by source_type, author,
project, labels (AND logic), file path prefix, created_after, and
updated_after. Date filters accept relative formats (7d, 2w) and
ISO dates.
- search::rrf: Reciprocal Rank Fusion implementation with configurable
k parameter and optional explain mode that annotates each result
with its component ranks and fusion score breakdown.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Implements the embedding module that generates vector representations
of documents using a local Ollama instance with the nomic-embed-text
model. These embeddings enable semantic (vector) search and the hybrid
search mode that fuses lexical and semantic results via RRF.
Key components:
- embedding::ollama: HTTP client for the Ollama /api/embeddings
endpoint. Handles connection errors with actionable error messages
(OllamaUnavailable, OllamaModelNotFound) and validates response
dimensions.
- embedding::chunking: Splits long documents into overlapping
paragraph-aware chunks for embedding. Uses a configurable max token
estimate (8192 default for nomic-embed-text) with 10% overlap to
preserve cross-chunk context.
- embedding::chunk_ids: Encodes chunk identity as
doc_id * 1000 + chunk_index for the embeddings table rowid. This
allows vector search to map results back to documents and
deduplicate by doc_id efficiently.
- embedding::change_detector: Compares document content_hash against
stored embedding hashes to skip re-embedding unchanged documents,
making incremental embedding runs fast.
- embedding::pipeline: Orchestrates the full embedding flow: detect
changed documents, chunk them, call Ollama in configurable
concurrency (default 4), store results. Supports --retry-failed
to re-attempt previously failed embeddings.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Implements the documents module that transforms raw ingested entities
(issues, MRs, discussions) into searchable document blobs stored in
the documents table. This is the foundation for both FTS5 lexical
search and vector embedding.
Key components:
- documents::extractor: Renders entities into structured text documents.
Issues include title, description, labels, milestone, assignees, and
threaded discussion summaries. MRs additionally include source/target
branches, reviewers, and approval status. Discussions are rendered
with full note threading.
- documents::regenerator: Drains the dirty_queue table to regenerate
only documents whose source entities changed since last sync. Supports
full rebuild mode (seeds all entities into dirty queue first) and
project-scoped regeneration.
- documents::truncation: Safety cap at 2MB per document to prevent
pathological outliers from degrading FTS or embedding performance.
- ingestion::dirty_tracker: Marks entities as dirty inside the
ingestion transaction so document regeneration stays consistent
with data changes. Uses INSERT OR IGNORE to deduplicate.
- ingestion::discussion_queue: Queue-based discussion fetching that
isolates individual discussion failures from the broader ingestion
pipeline, preventing a single corrupt discussion from blocking
an entire project sync.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Two targeted fixes to the GitLab API client:
1. Pagination: When the x-next-page header is missing but the current
page returned a full page of results, heuristically advance to the
next page instead of stopping. This fixes silent data truncation
observed with certain GitLab instances that omit pagination headers
on intermediate pages. The existing early-exit on empty or partial
pages remains as the termination condition.
2. Rate limiter: Refactor the async acquire() method into a synchronous
check_delay() that computes the required sleep duration and updates
last_request time while holding the mutex, then releases the lock
before sleeping. This eliminates holding the Mutex<RateLimiter>
across an await point, which previously could block other request
tasks unnecessarily during the sleep interval.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Mechanical rename of GiError -> LoreError across the core module to
match the project's rebranding from gitlab-inbox to gitlore/lore.
Updates the error enum name, all From impls, and the Result type alias.
Additionally introduces:
- New error variants for embedding pipeline: OllamaUnavailable,
OllamaModelNotFound, EmbeddingFailed, EmbeddingsNotBuilt. Each
includes actionable suggestions (e.g., "ollama serve", "ollama pull
nomic-embed-text") to guide users through recovery.
- New error codes 14-16 for programmatic handling of Ollama failures.
- Savepoint-based migration execution in db.rs: each migration now
runs inside a SQLite SAVEPOINT so a failed migration rolls back
cleanly without corrupting the schema_version tracking. Previously
a partial migration could leave the database in an inconsistent
state.
- core::backoff module: exponential backoff with jitter utility for
retry loops in the embedding pipeline and discussion queues.
- core::project module: helper for resolving project IDs and paths
from the local database, used by the document regenerator and
search filters.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Replaces the verb-first pattern ('lore list issues', 'lore show
issue 42') with noun-first subcommands that feel more natural:
lore issues # list issues
lore issues 42 # show issue #42
lore mrs # list merge requests
lore mrs 99 # show MR #99
lore ingest # ingest everything
lore ingest issues # ingest only issues
lore count issues # count issues
lore status # sync status
lore auth # verify auth
lore doctor # health check
Key changes:
- New IssuesArgs, MrsArgs, IngestArgs, CountArgs structs with
short flags (-n, -s, -p, -a, -l, -o, -f, -J, etc.)
- Global -J/--json flag as shorthand for --robot
- 'lore ingest' with no argument ingests both issues and MRs,
emitting combined JSON summary in robot mode
- --asc flag replaces --order=asc/desc for brevity
- Renamed flags: --has-due-date -> --has-due, --type -> --for,
--confirm -> --yes, target_branch -> --target, etc.
Old commands (list, show, auth-test, sync-status) are preserved
as hidden backward-compat aliases that emit deprecation warnings
to stderr before delegating to the new handlers.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Ingestion counters (discussions_upserted, notes_upserted,
discussions_fetched, diffnotes_count) were incremented before
tx.commit(), meaning a failed commit would report inflated
metrics. Counters now increment only after successful commit
so reported numbers accurately reflect persisted state.
Also simplifies the stale-removal guard in issue discussions:
the received_first_response flag was unnecessary since an empty
seen_discussion_ids list is safe to pass to remove_stale -- if
there were no discussions, stale removal correctly sweeps all
previously-stored discussions. The two separate code paths
(empty vs populated) are collapsed into a single branch.
Derives Default on IngestResult to eliminate verbose zero-init.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Two SQL correctness issues fixed:
1. Project filter used LIKE '%term%' which caused partial matches
(e.g. filtering for "foo" matched "group/foobar"). Now uses
exact match OR suffix match after '/' so "foo" matches
"group/foo" but not "group/foobar".
2. GROUP_CONCAT used comma as delimiter for labels and assignees,
which broke parsing when label names themselves contained commas.
Switched to ASCII unit separator (0x1F) which cannot appear in
GitLab entity names.
Also adds a guard for negative time deltas in format_relative_time
to handle clock skew gracefully instead of panicking.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Error suggestions now include concrete CLI examples so users
(and robot-mode consumers) can act immediately without consulting
docs. For instance, ConfigNotFound now shows the expected path
and the exact command to run, TokenNotSet shows the export syntax,
and Ambiguous shows the -p flag with example project paths.
Also fixes the error code for Ambiguous errors: it now maps to
GitLabNotFound instead of InternalError, since the entity exists
but the user needs to disambiguate -- not an internal failure.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Duplicate ISO 8601 timestamp parsing functions existed in both
discussion.rs and merge_request.rs transformers. This extracts
iso_to_ms_strict() and iso_to_ms_opt_strict() into core::time
as the single source of truth, and updates both transformer
modules to use the shared implementations.
Also removes the private now_ms() from merge_request.rs in
favor of the existing core::time::now_ms(), and replaces the
local parse_timestamp_opt() in discussion.rs with the public
iso_to_ms() from core::time.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Extends all data commands to support merge requests alongside issues,
with consistent patterns and JSON output for robot mode.
List command (gi list mrs):
- MR-specific columns: branches, draft status, reviewers
- Filters: --state (opened|merged|closed|locked|all), --draft,
--no-draft, --reviewer, --target-branch, --source-branch
- Discussion count with unresolved indicator (e.g., "5/2!")
- JSON output includes full MR metadata
Show command (gi show mr <iid>):
- MR details with branches, assignees, reviewers, merge status
- DiffNote positions showing file:line for code review comments
- Full description and discussion bodies (no truncation in JSON)
- --json flag for structured output with ISO timestamps
Count command (gi count mrs):
- MR counting with optional --type filter for discussions/notes
- JSON output with breakdown by state
Ingest command (gi ingest --type mrs):
- Full MR sync with discussion prefetch
- Progress output shows MR-specific metrics (diffnotes count)
- JSON summary with comprehensive sync statistics
All commands respect global --robot mode for auto-JSON output.
The pattern "gi list mrs --json | jq '.mrs[] | .iid'" now works
for scripted MR processing.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Introduces a unified robot mode that enables JSON output across all
commands, designed for AI agent and script consumption.
Robot mode activation (any of):
- --robot flag: Explicit opt-in
- GI_ROBOT=1 env var: For persistent configuration
- Non-TTY stdout: Auto-detect when piped (e.g., gi list issues | jq)
Implementation:
- Cli::is_robot_mode(): Centralized detection logic
- All command handlers receive robot_mode boolean
- Errors emit structured JSON to stderr with exit codes
- Success responses emit JSON to stdout
Behavior changes in robot mode:
- No color/emoji output (no ANSI escapes)
- No progress spinners or interactive prompts
- Timestamps as ISO 8601 strings (not relative "2 hours ago")
- Full content (no truncation of descriptions/notes)
- Structured error objects with code, message, suggestion
This enables reliable parsing by Claude Code, shell scripts, and
automation pipelines. The auto-detect on non-TTY means simple piping
"just works" without explicit flags.
Per-command --json flags remain for explicit control and override
robot mode when needed for human-friendly terminal + JSON file output.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Improves core infrastructure with robot-friendly error output and
faster lock release for better sync behavior.
Error handling improvements (error.rs):
- ErrorCode::exit_code(): Unique exit codes per error type (1-13)
for programmatic error handling in scripts/agents
- GiError::suggestion(): Helpful hints for common error recovery
- GiError::to_robot_error(): Structured JSON error conversion
- RobotError/RobotErrorOutput: Serializable error types with code,
message, and optional suggestion fields
Lock improvements (lock.rs):
- Heartbeat thread now polls every 100ms for release flag, only
updating database heartbeat at full interval (5s default)
- Eliminates 5-10s delay after sync completion when waiting for
heartbeat thread to notice release
- Reduces lock hold time after operation completes
Database (db.rs):
- Bump expected schema version to 6 for MR migration
The exit code mapping enables shell scripts and CI/CD pipelines to
distinguish between configuration errors (2-4), GitLab API errors
(5-8), and database errors (9-11) for appropriate retry/alert logic.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Adds complete merge request ingestion pipeline with a novel two-phase
discussion sync strategy optimized for throughput.
New modules:
- merge_requests.rs: MR upsert with labels/assignees/reviewers handling,
stale MR cleanup, and watermark-based incremental sync
- mr_discussions.rs: Parallel prefetch strategy for MR discussions
Two-phase MR discussion sync:
1. PREFETCH PHASE: Spawn concurrent tasks to fetch discussions for
multiple MRs simultaneously (configurable concurrency, default 8).
Transform and validate in parallel, storing results in memory.
2. WRITE PHASE: Serial database writes to avoid lock contention.
Each MR's discussions written in a single transaction, with
proper stale discussion cleanup.
This approach achieves ~4-8x throughput vs serial fetching while
maintaining database consistency. Transform errors are tracked per-MR
to prevent partial writes from corrupting watermarks.
Orchestrator updates:
- ingest_merge_requests(): Coordinates MR fetch -> discussion sync flow
- Progress callbacks emit MR-specific events for UI feedback
- Respects --full flag to reset discussion watermarks for full resync
The prefetch strategy is critical for MRs which typically have more
discussions than issues, and where API latency dominates sync time.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Introduces NormalizedMergeRequest transformer and updates discussion
normalization to handle both issue and MR discussions polymorphically.
New transformers:
- NormalizedMergeRequest: Transforms API MergeRequest to database row,
extracting labels/assignees/reviewers into separate collections for
junction table insertion. Handles draft detection, detailed_merge_status
preference over deprecated merge_status, and merge_user over merged_by.
Discussion transformer updates:
- NormalizedDiscussion now takes noteable_type ("Issue" | "MergeRequest")
and noteable_id for polymorphic FK binding
- normalize_discussions_for_issue(): Convenience wrapper for issues
- normalize_discussions_for_mr(): Convenience wrapper for MRs
- DiffNote position fields (type, line_range, SHA triplet) now extracted
from API position object for code review context
Design decisions:
- Transformer returns (normalized_item, labels, assignees, reviewers)
tuple for efficient batch insertion without re-querying
- Timestamps converted to ms epoch for SQLite storage consistency
- Optional fields use map() chains for clean null handling
The polymorphic discussion approach allows reusing the same discussions
and notes tables for both issues and MRs, with noteable_type + FK
determining the parent relationship.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Extends GitLabClient with endpoints for fetching merge requests and
their discussions, following the same patterns established for issues.
New methods:
- fetch_merge_requests(): Paginated MR listing with cursor support,
using updated_after filter for incremental sync. Uses 'all' scope
to include MRs where user is author/assignee/reviewer.
- fetch_merge_requests_single_page(): Single page variant for callers
managing their own pagination (used by parallel prefetch)
- fetch_mr_discussions(): Paginated discussion listing for a single MR,
returns full discussion trees with notes
API design notes:
- Uses keyset pagination (order_by=updated_at, keyset=true) for
consistent results during sync operations
- MR endpoint uses /merge_requests (not /mrs) per GitLab API naming
- Discussion endpoint matches issue pattern for consistency
- Per_page defaults to 100 (GitLab max) for efficiency
The fetch_merge_requests_single_page method enables the parallel
prefetch strategy used in mr_discussions.rs, where multiple MRs'
discussions are fetched concurrently during the sweep phase.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Extends GitLab type definitions with comprehensive merge request support,
matching the API response structure for /projects/:id/merge_requests.
New types:
- MergeRequest: Full MR metadata including draft status, branch info,
detailed_merge_status, merge_user (modern API fields replacing
deprecated alternatives), and references for cross-project support
- MrReviewer: Reviewer user info (MR-specific, distinct from assignees)
- MrAssignee: Assignee user info with consistent structure
- MrDiscussion: MR discussion wrapper for polymorphic handling
- DiffNotePosition: Rich position data for code review comments with
line ranges and SHA triplet for commit context
Design decisions:
- Use Option<T> for all nullable API fields to handle partial responses
- Include deprecated fields (merged_by, merge_status) alongside modern
alternatives for backward compatibility with older GitLab instances
- DiffNotePosition uses Option for all fields since different position
types (text/image/file) populate different subsets
These types enable type-safe deserialization of GitLab MR API responses
with full coverage of the fields needed for CP2 ingestion.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This is a P1 fix from the CP1-CP2 alignment audit. The --full flag was
designed to enable complete data re-synchronization, but it only reset
sync_cursors for issues—it failed to reset the per-issue
discussions_synced_for_updated_at watermark.
The result was an inconsistent state: issues would be re-fetched from
GitLab (because sync_cursors were cleared), but their discussions would
NOT be re-synced (because the watermark comparison prevented it). This
was a subtle bug because the watermark check uses:
WHERE updated_at > COALESCE(discussions_synced_for_updated_at, 0)
When discussions_synced_for_updated_at is already set to the issue's
updated_at, the comparison fails and discussions are skipped.
Fix: Before clearing sync_cursors, set discussions_synced_for_updated_at
to NULL for all issues in the project. This makes COALESCE return 0,
ensuring all issues become eligible for discussion sync.
The ordering is important: watermarks must be reset BEFORE cursors to
ensure the full sync behaves consistently.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This is a P0 fix from the CP1-CP2 alignment audit. The original
NormalizedDiscussion struct had issue_id as a non-optional i64 and
hardcoded noteable_type to "Issue", making it incompatible with merge
request discussions even though the database schema already supports
both via nullable columns and a CHECK constraint.
Changes:
- Add NoteableRef enum with Issue(i64) and MergeRequest(i64) variants
to provide compile-time safety against mixing up issue vs MR IDs
- Change NormalizedDiscussion.issue_id from i64 to Option<i64>
- Add NormalizedDiscussion.merge_request_id: Option<i64>
- Update transform_discussion() signature to take NoteableRef instead
of local_issue_id, deriving issue_id/merge_request_id/noteable_type
from the enum variant
- Update upsert_discussion() SQL to include merge_request_id column
(now 12 parameters instead of 11)
- Export NoteableRef from transformers module
- Add test for MergeRequest discussion transformation
- Update all existing tests to use NoteableRef::Issue(id)
The database schema (migration 002) was forward-thinking and already
supports both issue_id and merge_request_id as nullable columns with
a CHECK constraint. This change prepares the application layer for
CP2 merge request support without requiring any migrations.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Provides a typed interface to the GitLab API with pagination support.
src/gitlab/types.rs - API response type definitions:
- GitLabIssue: Full issue payload with author, assignees, labels
- GitLabDiscussion: Discussion thread with notes array
- GitLabNote: Individual note with author, timestamps, body
- GitLabAuthor/GitLabUser: User information with avatar URLs
- GitLabProject: Project metadata from /api/v4/projects
- GitLabVersion: GitLab instance version from /api/v4/version
- GitLabNotePosition: Line-level position for diff notes
- All types derive Deserialize for JSON parsing
src/gitlab/client.rs - HTTP client with authentication:
- Bearer token authentication from config
- Base URL configuration for self-hosted instances
- Paginated iteration via keyset or offset pagination
- Automatic Link header parsing for next page URLs
- Per-page limit control (default 100)
- Methods: get_user(), get_version(), get_project()
- Async stream for issues: list_issues_paginated()
- Async stream for discussions: list_issue_discussions_paginated()
- Respects GitLab rate limiting via response headers
src/gitlab/transformers/ - API to database mapping:
transformers/issue.rs - Issue transformation:
- Maps GitLabIssue to IssueRow for database insert
- Extracts milestone ID and due date
- Normalizes author/assignee usernames
- Preserves label IDs for junction table
- Returns IssueWithMetadata including label/assignee lists
transformers/discussion.rs - Discussion transformation:
- Maps GitLabDiscussion to NormalizedDiscussion
- Extracts thread metadata (resolvable, resolved)
- Flattens notes to NormalizedNote with foreign keys
- Handles system notes vs user notes
- Preserves note position for diff discussions
transformers/mod.rs - Re-exports all transformer types
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
