docs: Restructure checkpoint-3 PRD with gated milestones

Reorganizes the Search & Sync MVP plan into three independently
verifiable gates (A: Lexical MVP, B: Hybrid MVP, C: Sync MVP)
to reduce integration risk. Each gate has explicit deliverables,
acceptance criteria, and can ship on its own.

Expands the specification with additional detail on document
generation, search API surface, sync orchestration, and
integrity repair paths. Removes the outdated rename note since
the project is now fully migrated to gitlore/lore naming.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

docs/prd/checkpoint-3.md

@@ -1,7 +1,5 @@
 # Checkpoint 3: Search & Sync MVP
 
-> **Note:** The project was renamed from "gitlab-inbox" to "gitlore" and the CLI from "gi" to "lore". References to "gi" in this document should be read as "lore".
-
 > **Status:** Planning
 > **Prerequisite:** Checkpoints 0, 1, 2 complete (issues, MRs, discussions ingested)
 > **Goal:** Deliver working semantic + lexical hybrid search with efficient incremental sync
@@ -17,13 +15,27 @@ All code integrates with existing `gitlore` infrastructure:
 
 ---
 
-## Executive Summary
+## Executive Summary (Gated Milestones)
+
+This checkpoint ships in three gates to reduce integration risk. Each gate is independently verifiable and shippable:
+
+**Gate A (Lexical MVP):** documents + FTS + filters + `lore search --mode=lexical` + `lore stats`
+**Gate B (Hybrid MVP):** embeddings + vector + RRF fusion + graceful degradation
+**Gate C (Sync MVP):** `lore sync` orchestration + queues/backoff + integrity check/repair
 
 **Deliverables:**
+
+**Gate A**
 1. Document generation from issues/MRs/discussions with FTS5 indexing
-2. Ollama-powered embedding pipeline with sqlite-vec storage
-3. Hybrid search (RRF-ranked vector + lexical) with rich filtering
-4. Orchestrated `gi sync` command with incremental re-embedding
+2. Lexical search + filters + snippets + `lore stats`
+
+**Gate B**
+3. Ollama-powered embedding pipeline with sqlite-vec storage
+4. Hybrid search (RRF-ranked vector + lexical) with rich filtering + graceful degradation
+
+**Gate C**
+5. Orchestrated `lore sync` command with incremental doc regen + re-embedding
+6. Integrity checks + repair paths for FTS/embeddings consistency
 
 **Key Design Decisions:**
 - Documents are the search unit (not raw entities)
@@ -144,15 +156,19 @@ CREATE VIRTUAL TABLE documents_fts USING fts5(
   prefix='2 3 4'
 );
 
--- Keep FTS in sync via triggers
+-- Keep FTS in sync via triggers.
+-- IMPORTANT: COALESCE(title, '') ensures FTS5 external-content table never
+-- receives NULL values, which can cause inconsistencies with delete operations.
+-- FTS5 delete requires exact match of original values; NULL != NULL in SQL,
+-- so a NULL title on insert would make the delete trigger fail silently.
 CREATE TRIGGER documents_ai AFTER INSERT ON documents BEGIN
   INSERT INTO documents_fts(rowid, title, content_text)
-  VALUES (new.id, new.title, new.content_text);
+  VALUES (new.id, COALESCE(new.title, ''), new.content_text);
 END;
 
 CREATE TRIGGER documents_ad AFTER DELETE ON documents BEGIN
   INSERT INTO documents_fts(documents_fts, rowid, title, content_text)
-  VALUES('delete', old.id, old.title, old.content_text);
+  VALUES('delete', old.id, COALESCE(old.title, ''), old.content_text);
 END;
 
 -- Only rebuild FTS when searchable text actually changes (not metadata-only updates)
@@ -160,9 +176,9 @@ CREATE TRIGGER documents_au AFTER UPDATE ON documents
 WHEN old.title IS NOT new.title OR old.content_text != new.content_text
 BEGIN
   INSERT INTO documents_fts(documents_fts, rowid, title, content_text)
-  VALUES('delete', old.id, old.title, old.content_text);
+  VALUES('delete', old.id, COALESCE(old.title, ''), old.content_text);
   INSERT INTO documents_fts(rowid, title, content_text)
-  VALUES (new.id, new.title, new.content_text);
+  VALUES (new.id, COALESCE(new.title, ''), new.content_text);
 END;
 ```
 
@@ -490,7 +506,7 @@ pub struct NoteContent {
 
 ---
 
-### 2.4 CLI: `gi generate-docs` (Incremental by Default)
+### 2.4 CLI: `lore generate-docs` (Incremental by Default)
 
 **File:** `src/cli/commands/generate_docs.rs`
 
@@ -518,84 +534,83 @@ pub struct GenerateDocsResult {
     pub skipped: usize,  // Unchanged documents
 }
 
-/// Chunk size for --full mode transactions.
+/// Chunk size for --full mode dirty queue seeding.
 /// Balances throughput against WAL file growth and memory pressure.
 const FULL_MODE_CHUNK_SIZE: usize = 2000;
 
 /// Run document generation (incremental by default).
 ///
+/// IMPORTANT: Both modes use the same regenerator codepath to avoid
+/// logic divergence in label/path hashing, deletion semantics, and
+/// write-optimization behavior. The only difference is how dirty_sources
+/// gets populated.
+///
 /// Incremental mode (default):
-/// - Processes only items in dirty_sources queue
+/// - Processes only items already in dirty_sources queue
 /// - Fast for routine syncs
 ///
 /// Full mode (--full):
-/// - Regenerates ALL documents from scratch
-/// - Uses chunked transactions (2k docs/tx) to bound WAL growth
+/// - Seeds dirty_sources with ALL source entities in chunks
+/// - Drains through the same regenerator pipeline
+/// - Uses keyset pagination (WHERE id > last_id) to avoid OFFSET degradation
+/// - Final FTS optimize after all chunks complete
 /// - Use when schema changes or after migration
 pub fn run_generate_docs(
     config: &Config,
     full: bool,
     project_filter: Option<&str>,
 ) -> Result<GenerateDocsResult> {
+    let conn = open_db(config)?;
+
     if full {
-        // Full mode: regenerate everything using chunked transactions
+        // Full mode: seed dirty_sources with all source entities, then drain.
+        // Uses keyset pagination to avoid O(n²) OFFSET degradation on large tables.
         //
-        // Using chunked transactions instead of a single giant transaction:
-        // - Bounds WAL file growth (single 50k-doc tx could balloon WAL)
-        // - Reduces memory pressure from statement caches
-        // - Allows progress reporting between chunks
-        // - Crash partway through leaves partial but consistent state
+        // Seeding is chunked to bound WAL growth:
+        // 1. For each source type (issues, MRs, discussions):
+        //    a. Query next chunk WHERE id > last_id ORDER BY id LIMIT chunk_size
+        //    b. INSERT OR IGNORE each into dirty_sources
+        //    c. Advance last_id = chunk.last().id
+        //    d. Loop until chunk is empty
+        // 2. Drain dirty_sources through regenerator (same as incremental)
+        // 3. Final FTS optimize (not full rebuild — triggers handle consistency)
         //
-        // Steps per chunk:
-        // 1. BEGIN IMMEDIATE transaction
-        // 2. Query next batch of sources (issues/MRs/discussions)
-        // 3. For each: generate document, compute hash
-        // 4. Upsert into `documents` table (FTS triggers auto-fire)
-        // 5. Populate `document_labels` and `document_paths`
-        // 6. COMMIT
-        // 7. Report progress, loop to next chunk
-        //
-        // After all chunks:
-        // 8. Single final transaction for FTS rebuild:
-        //    INSERT INTO documents_fts(documents_fts) VALUES('rebuild')
-        //
-        // Example implementation:
-        let conn = open_db(config)?;
-        let mut result = GenerateDocsResult::default();
-        let mut offset = 0;
+        // Benefits of unified codepath:
+        // - No divergence in label/path hash behavior
+        // - No divergence in deletion semantics
+        // - No divergence in write-optimization logic (labels_hash, paths_hash)
+        // - FTS triggers fire identically in both modes
 
+        // Seed issues
+        let mut last_id: i64 = 0;
         loop {
-            // Process issues in chunks
-            let issues: Vec<Issue> = query_issues(&conn, project_filter, FULL_MODE_CHUNK_SIZE, offset)?;
-            if issues.is_empty() { break; }
-
+            let chunk = query_issue_ids_after(&conn, project_filter, FULL_MODE_CHUNK_SIZE, last_id)?;
+            if chunk.is_empty() { break; }
             let tx = conn.transaction()?;
-            for issue in &issues {
-                let doc = generate_issue_document(issue)?;
-                upsert_document(&tx, &doc)?;
-                result.issues += 1;
+            for id in &chunk {
+                mark_dirty(&tx, SourceType::Issue, *id)?;
             }
             tx.commit()?;
-
-            offset += issues.len();
-            // Report progress here if using indicatif
+            last_id = *chunk.last().unwrap();
         }
 
-        // Similar chunked loops for MRs and discussions...
+        // Similar keyset-paginated seeding for MRs and discussions...
 
-        // Final FTS rebuild in its own transaction
-        let tx = conn.transaction()?;
-        tx.execute(
-            "INSERT INTO documents_fts(documents_fts) VALUES('rebuild')",
+        // Report: seeding complete, now regenerating
+    }
+
+    // Both modes: drain dirty_sources through the regenerator
+    let regen = regenerate_dirty_documents(&conn)?;
+
+    if full {
+        // FTS optimize after bulk operations (compacts index segments)
+        conn.execute(
+            "INSERT INTO documents_fts(documents_fts) VALUES('optimize')",
             [],
         )?;
-        tx.commit()?;
-    } else {
-        // Incremental mode: process dirty_sources only
-        // 1. Query dirty_sources (bounded by LIMIT)
-        // 2. Regenerate only those documents
-        // 3. Clear from dirty_sources after processing
     }
+
+    // Map regen -> GenerateDocsResult stats
     todo!()
 }
 
@@ -849,7 +864,7 @@ pub fn search_fts(
     )?;
 
     let results = stmt
-        .query_map([&safe_query, &limit.to_string()], |row| {
+        .query_map(rusqlite::params![safe_query, limit as i64], |row| {
             Ok(FtsResult {
                 document_id: row.get(0)?,
                 rank: row.get(1)?,
@@ -897,10 +912,11 @@ pub struct SearchFilters {
     pub source_type: Option<SourceType>,
     pub author: Option<String>,
     pub project_id: Option<i64>,
-    pub after: Option<i64>,        // ms epoch
-    pub labels: Vec<String>,       // AND logic
+    pub after: Option<i64>,           // ms epoch (created_at >=)
+    pub updated_after: Option<i64>,   // ms epoch (updated_at >=)
+    pub labels: Vec<String>,          // AND logic
     pub path: Option<PathFilter>,
-    pub limit: usize,              // Default 20, max 100
+    pub limit: usize,                 // Default 20, max 100
 }
 
 impl SearchFilters {
@@ -910,6 +926,7 @@ impl SearchFilters {
             || self.author.is_some()
             || self.project_id.is_some()
             || self.after.is_some()
+            || self.updated_after.is_some()
             || !self.labels.is_empty()
             || self.path.is_some()
     }
@@ -990,6 +1007,11 @@ pub fn apply_filters(
         params.push(Box::new(after));
     }
 
+    if let Some(updated_after) = filters.updated_after {
+        conditions.push("d.updated_at >= ?".into());
+        params.push(Box::new(updated_after));
+    }
+
     // Labels: AND logic - all labels must be present
     for label in &filters.labels {
         conditions.push(
@@ -1064,6 +1086,7 @@ pub fn apply_filters(
 | `--author` | `author_username` | Exact match |
 | `--project` | `project_id` | Resolve path to ID |
 | `--after` | `created_at` | `>= date` (ms epoch) |
+| `--updated-after` | `updated_at` | `>= date` (ms epoch), common triage filter |
 | `--label` | `document_labels` | JOIN, multiple = AND |
 | `--path` | `document_paths` | JOIN, trailing `/` = prefix |
 | `--limit` | N/A | Default 20, max 100 |
@@ -1072,6 +1095,7 @@ pub fn apply_filters(
 - [ ] Each filter correctly restricts results
 - [ ] Multiple `--label` flags use AND logic
 - [ ] Path prefix vs exact match works correctly
+- [ ] `--updated-after` filters on updated_at (not created_at)
 - [ ] Filters compose (all applied together)
 - [ ] Ranking order preserved after filtering (ORDER BY position)
 - [ ] Limit clamped to valid range [1, 100]
@@ -1080,7 +1104,7 @@ pub fn apply_filters(
 
 ---
 
-### 3.4 CLI: `gi search --mode=lexical`
+### 3.4 CLI: `lore search --mode=lexical`
 
 **File:** `src/cli/commands/search.rs`
 
@@ -1141,9 +1165,49 @@ pub fn run_search(
     explain: bool,
 ) -> Result<SearchResponse> {
     // 1. Parse query and filters
-    // 2. Execute search based on mode
-    // 3. Apply post-retrieval filters
-    // 4. Format and return results
+    // 2. Execute search based on mode -> ranked doc_ids (+ explain ranks)
+    // 3. Apply post-retrieval filters preserving ranking order
+    // 4. HYDRATE in one DB round-trip (see hydration query below):
+    //    - documents fields (title, url, created_at, updated_at, content_text)
+    //    - project_path via JOIN projects
+    //    - labels aggregated via json_group_array
+    //    - paths aggregated via json_group_array (optional)
+    // 5. Attach snippet:
+    //    - prefer FTS snippet when doc hit FTS
+    //    - fallback: truncated content_text via generate_fallback_snippet()
+    // 6. For --mode=semantic with 0% embedding coverage:
+    //    return early with actionable error message (distinct from "Ollama down")
+    todo!()
+}
+
+/// Hydration query: fetch all display fields for ranked doc IDs in a single round-trip.
+///
+/// Uses json_each(?) to preserve ranking order from the search pipeline.
+/// Aggregates labels and paths inline to avoid N+1 queries.
+///
+/// ```sql
+/// SELECT d.id, d.source_type, d.title, d.url, d.author_username,
+///        d.created_at, d.updated_at, d.content_text,
+///        p.path AS project_path,
+///        (SELECT json_group_array(dl.label_name)
+///         FROM document_labels dl WHERE dl.document_id = d.id) AS labels,
+///        (SELECT json_group_array(dp.path)
+///         FROM document_paths dp WHERE dp.document_id = d.id) AS paths
+/// FROM json_each(?) AS j
+/// JOIN documents d ON d.id = j.value
+/// JOIN projects p ON p.id = d.project_id
+/// ORDER BY j.key
+/// ```
+///
+/// This single query replaces what would otherwise be:
+/// - 1 query per document for metadata
+/// - 1 query per document for labels
+/// - 1 query per document for paths
+/// For 20 results, that's 60 queries reduced to 1.
+fn hydrate_results(
+    conn: &Connection,
+    doc_ids: &[i64],
+) -> Result<Vec<HydratedDocument>> {
     todo!()
 }
 
@@ -1240,6 +1304,10 @@ pub struct SearchArgs {
     #[arg(long)]
     after: Option<String>,
 
+    /// Filter by updated date (recently active items)
+    #[arg(long)]
+    updated_after: Option<String>,
+
     /// Filter by label (can specify multiple)
     #[arg(long, action = clap::ArgAction::Append)]
     label: Vec<String>,
@@ -1266,12 +1334,15 @@ pub struct SearchArgs {
 
 **Acceptance Criteria:**
 - [ ] Works without Ollama running
-- [ ] All filters functional
+- [ ] All filters functional (including `--updated-after`)
 - [ ] Human-readable output with snippets
 - [ ] Semantic-only results get fallback snippets from content_text
+- [ ] Results hydrated in single DB round-trip (no N+1 queries)
 - [ ] JSON output matches schema
 - [ ] Empty results show helpful message
 - [ ] "No data indexed" message if documents table empty
+- [ ] `--mode=semantic` with 0% embedding coverage returns actionable error
+  (distinct from "Ollama unavailable" — tells user to run `lore embed` first)
 - [ ] `--fts-mode=safe` (default) preserves prefix `*` while escaping special chars
 - [ ] `--fts-mode=raw` passes FTS5 MATCH syntax through unchanged
 
@@ -1535,7 +1606,7 @@ impl GiError {
             // ... existing mappings ...
             Self::OllamaUnavailable { .. } => Some("Start Ollama: ollama serve"),
             Self::OllamaModelNotFound { model } => Some("Pull the model: ollama pull nomic-embed-text"),
-            Self::EmbeddingFailed { .. } => Some("Check Ollama logs or retry with 'gi embed --retry-failed'"),
+            Self::EmbeddingFailed { .. } => Some("Check Ollama logs or retry with 'lore embed --retry-failed'"),
         }
     }
 }
@@ -1558,6 +1629,7 @@ use crate::embedding::OllamaClient;
 const BATCH_SIZE: usize = 32;
 
 /// SQLite page size for paging through pending documents.
+/// Uses keyset paging (id > last_id) to avoid rescanning previously-processed rows.
 const DB_PAGE_SIZE: usize = 500;
 
 /// Expected embedding dimensions for nomic-embed-text model.
@@ -1584,11 +1656,16 @@ pub struct EmbedResult {
 /// Embed documents that need embedding.
 ///
 /// Process:
-/// 1. Query dirty_sources ordered by queued_at
-/// 2. For each: regenerate document, compute new hash
-/// 3. ALWAYS upsert document (labels/paths may change even if content_hash unchanged)
-/// 4. Track whether content_hash changed (for stats)
-/// 5. Delete from dirty_sources (or record error on failure)
+/// 1. Select documents needing embeddings:
+///    - Pending: missing embedding_metadata row OR content_hash mismatch
+///    - RetryFailed: embedding_metadata.last_error IS NOT NULL
+/// 2. Page through candidates using keyset pagination (id > last_id)
+///    to avoid rescanning already-processed rows
+/// 3. Batch texts -> Ollama `/api/embed` with concurrent HTTP requests
+/// 4. Write embeddings + embedding_metadata in per-batch transactions
+/// 5. Failed batches record `last_error` in embedding_metadata
+///    (excluded from Pending selection; retried via RetryFailed)
+/// 6. Progress reported as (embedded + failed) vs total_pending
 pub async fn embed_documents(
     conn: &Connection,
     client: &OllamaClient,
@@ -1605,9 +1682,11 @@ pub async fn embed_documents(
         return Ok(result);
     }
 
-    // Page through pending documents to avoid loading all into memory
+    // Page through pending documents using keyset pagination to avoid
+    // both memory pressure and OFFSET performance degradation.
+    let mut last_id: i64 = 0;
     loop {
-        let pending = find_pending_documents(conn, DB_PAGE_SIZE, selection)?;
+        let pending = find_pending_documents(conn, DB_PAGE_SIZE, last_id, selection)?;
         if pending.is_empty() {
             break;
         }
@@ -1640,6 +1719,11 @@ pub async fn embed_documents(
             collect_writes(conn, &meta, res, &mut result)?;
         }
 
+        // Advance keyset cursor for next page
+        if let Some(last) = pending.last() {
+            last_id = last.id;
+        }
+
         if let Some(ref cb) = progress_callback {
             cb(result.embedded + result.failed, total_pending);
         }
@@ -1718,14 +1802,16 @@ fn count_pending_documents(conn: &Connection, selection: EmbedSelection) -> Resu
     Ok(count)
 }
 
-/// Find pending documents for embedding.
+/// Find pending documents for embedding using keyset pagination.
 ///
-/// IMPORTANT: Uses deterministic ORDER BY d.id to ensure consistent
-/// paging behavior. Without ordering, SQLite may return rows in
-/// different orders across calls, causing missed or duplicate documents.
+/// IMPORTANT: Uses keyset pagination (d.id > last_id) instead of OFFSET.
+/// OFFSET degrades O(n²) on large result sets because SQLite must scan
+/// and discard all rows before the offset. Keyset pagination is O(1) per page
+/// since the index seek goes directly to the starting row.
 fn find_pending_documents(
     conn: &Connection,
     limit: usize,
+    last_id: i64,
     selection: EmbedSelection,
 ) -> Result<Vec<PendingDocument>> {
     let sql = match selection {
@@ -1733,8 +1819,9 @@ fn find_pending_documents(
             "SELECT d.id, d.content_text, d.content_hash
              FROM documents d
              LEFT JOIN embedding_metadata em ON d.id = em.document_id
-             WHERE em.document_id IS NULL
-                OR em.content_hash != d.content_hash
+             WHERE (em.document_id IS NULL
+                OR em.content_hash != d.content_hash)
+               AND d.id > ?
              ORDER BY d.id
              LIMIT ?",
         EmbedSelection::RetryFailed =>
@@ -1742,13 +1829,14 @@ fn find_pending_documents(
              FROM documents d
              JOIN embedding_metadata em ON d.id = em.document_id
              WHERE em.last_error IS NOT NULL
+               AND d.id > ?
              ORDER BY d.id
              LIMIT ?",
     };
     let mut stmt = conn.prepare(sql)?;
 
     let docs = stmt
-        .query_map([limit], |row| {
+        .query_map(rusqlite::params![last_id, limit as i64], |row| {
             Ok(PendingDocument {
                 id: row.get(0)?,
                 content: row.get(1)?,
@@ -1827,7 +1915,7 @@ fn record_embedding_error(
 
 ---
 
-### 4.5 CLI: `gi embed`
+### 4.5 CLI: `lore embed`
 
 **File:** `src/cli/commands/embed.rs`
 
@@ -1928,7 +2016,7 @@ pub struct EmbedArgs {
 
 ---
 
-### 4.6 CLI: `gi stats`
+### 4.6 CLI: `lore stats`
 
 **File:** `src/cli/commands/stats.rs`
 
@@ -2034,7 +2122,13 @@ pub struct RepairResult {
 /// Fixes:
 /// - Deletes orphaned embeddings (embedding_metadata rows with no matching document)
 /// - Clears stale embedding_metadata (hash mismatch) so they get re-embedded
-/// - Repopulates FTS for documents missing from documents_fts
+/// - Rebuilds FTS index from scratch (correct-by-construction)
+///
+/// NOTE: FTS repair uses `rebuild` rather than partial row insertion.
+/// With `content='documents'` (external-content FTS), partial repopulation
+/// via INSERT of missing rows is fragile — if the external content table
+/// and FTS content diverge in any way, partial fixes can leave the index
+/// in an inconsistent state. A full rebuild is slower but guaranteed correct.
 pub fn run_repair(config: &Config) -> Result<RepairResult> {
     let conn = open_db(config)?;
 
@@ -2061,19 +2155,19 @@ pub fn run_repair(config: &Config) -> Result<RepairResult> {
         [],
     )?;
 
-    // Repopulate FTS for missing documents
-    let fts_repopulated = conn.execute(
-        "INSERT INTO documents_fts(rowid, title, content_text)
-         SELECT id, COALESCE(title, ''), content_text
-         FROM documents
-         WHERE id NOT IN (SELECT rowid FROM documents_fts)",
+    // Rebuild FTS index from scratch — correct-by-construction.
+    // This re-reads all rows from the external content table (documents)
+    // and rebuilds the index. Slower than partial fix but guaranteed consistent.
+    conn.execute(
+        "INSERT INTO documents_fts(documents_fts) VALUES('rebuild')",
         [],
     )?;
+    let fts_rebuilt = 1; // rebuild is all-or-nothing
 
     Ok(RepairResult {
         orphaned_embeddings_deleted: orphaned_deleted,
         stale_embeddings_cleared: stale_cleared,
-        missing_fts_repopulated: fts_repopulated,
+        missing_fts_repopulated: fts_rebuilt,
     })
 }
 
@@ -2772,7 +2866,7 @@ pub fn record_fetch_error(
     current_attempt: i64,
 ) -> Result<()> {
     let now = now_ms();
-    let next_attempt = compute_next_attempt_at(now, current_attempt + 1);
+    let next_attempt = crate::core::backoff::compute_next_attempt_at(now, current_attempt + 1);
 
     conn.execute(
         "UPDATE pending_discussion_fetches
@@ -2786,14 +2880,44 @@ pub fn record_fetch_error(
     Ok(())
 }
 
+// NOTE: Backoff computation uses the shared utility in `src/core/backoff.rs`.
+// See Phase 6.X below for the shared implementation.
+```
+
+**Acceptance Criteria:**
+- [ ] Updated entities queued for discussion fetch
+- [ ] Success removes from queue
+- [ ] Failure increments attempt_count and sets next_attempt_at
+- [ ] Processing bounded per run (max 100)
+- [ ] Exponential backoff uses `next_attempt_at` (index-friendly, no overflow)
+- [ ] Backoff computed with jitter to prevent thundering herd
+
+---
+
+### 6.X Shared Backoff Utility
+
+**File:** `src/core/backoff.rs`
+
+Single implementation of exponential backoff with jitter, used by both
+`dirty_sources` and `pending_discussion_fetches` queues. Living in `src/core/`
+because it's a cross-cutting concern used by multiple modules.
+
+```rust
+use rand::Rng;
+
 /// Compute next_attempt_at with exponential backoff and jitter.
 ///
 /// Formula: now + min(3600000, 1000 * 2^attempt_count) * (0.9 to 1.1)
 /// - Capped at 1 hour to prevent runaway delays
 /// - ±10% jitter prevents synchronized retries after outages
+///
+/// Used by:
+/// - `dirty_sources` retry scheduling (document regeneration failures)
+/// - `pending_discussion_fetches` retry scheduling (API fetch failures)
+///
+/// Having one implementation prevents subtle divergence between queues
+/// (e.g., different caps or jitter ranges).
 pub fn compute_next_attempt_at(now: i64, attempt_count: i64) -> i64 {
-    use rand::Rng;
-
     // Cap attempt_count to prevent overflow (2^30 > 1 hour anyway)
     let capped_attempts = attempt_count.min(30) as u32;
     let base_delay_ms = 1000_i64.saturating_mul(1 << capped_attempts);
@@ -2807,13 +2931,16 @@ pub fn compute_next_attempt_at(now: i64, attempt_count: i64) -> i64 {
 }
 ```
 
+**Update `src/core/mod.rs`:**
+```rust
+pub mod backoff;  // Add to existing modules
+```
+
 **Acceptance Criteria:**
-- [ ] Updated entities queued for discussion fetch
-- [ ] Success removes from queue
-- [ ] Failure increments attempt_count and sets next_attempt_at
-- [ ] Processing bounded per run (max 100)
-- [ ] Exponential backoff uses `next_attempt_at` (index-friendly, no overflow)
-- [ ] Backoff computed with jitter to prevent thundering herd
+- [ ] Single implementation shared by both queue retry paths
+- [ ] Cap at 1 hour prevents runaway delays
+- [ ] Jitter prevents thundering herd after outage recovery
+- [ ] Unit tests verify backoff curve and cap behavior
 
 ---
 
@@ -2917,19 +3044,36 @@ fn delete_document(
 }
 
 /// Record a regeneration error on a dirty source for retry.
+///
+/// IMPORTANT: Sets `next_attempt_at` using the shared backoff utility.
+/// Without this, failed items would retry every run (hot-loop), defeating
+/// the backoff design documented in the schema.
 fn record_dirty_error(
     conn: &Connection,
     source_type: SourceType,
     source_id: i64,
     error: &str,
 ) -> Result<()> {
+    let now = now_ms();
+
+    // Read current attempt_count from DB to compute backoff
+    let attempt_count: i64 = conn.query_row(
+        "SELECT attempt_count FROM dirty_sources WHERE source_type = ? AND source_id = ?",
+        rusqlite::params![source_type.as_str(), source_id],
+        |row| row.get(0),
+    )?;
+
+    // Use shared backoff utility (same as pending_discussion_fetches)
+    let next_attempt_at = crate::core::backoff::compute_next_attempt_at(now, attempt_count + 1);
+
     conn.execute(
         "UPDATE dirty_sources
          SET attempt_count = attempt_count + 1,
              last_attempt_at = ?,
-             last_error = ?
+             last_error = ?,
+             next_attempt_at = ?
          WHERE source_type = ? AND source_id = ?",
-        rusqlite::params![now_ms(), error, source_type.as_str(), source_id],
+        rusqlite::params![now, error, next_attempt_at, source_type.as_str(), source_id],
     )?;
     Ok(())
 }
@@ -3080,7 +3224,7 @@ fn get_document_id(
 
 ---
 
-### 6.4 CLI: `gi sync`
+### 6.4 CLI: `lore sync`
 
 **File:** `src/cli/commands/sync.rs`
 
@@ -3198,7 +3342,8 @@ pub struct SyncArgs {
 | FTS query sanitization | `src/search/fts.rs` (mod tests) | `to_fts_query()` edge cases: `-`, `"`, `:`, `*`, `C++` |
 | SourceType parsing | `src/documents/extractor.rs` (mod tests) | `parse()` accepts aliases: `mr`, `mrs`, `issue`, etc. |
 | SearchFilters | `src/search/filters.rs` (mod tests) | `has_any_filter()`, `clamp_limit()` |
-| Backoff logic | `src/ingestion/dirty_tracker.rs` (mod tests) | Exponential backoff query timing |
+| Backoff logic | `src/core/backoff.rs` (mod tests) | Shared exponential backoff curve, cap, jitter |
+| Hydration | `src/cli/commands/search.rs` (mod tests) | Single round-trip, label/path aggregation |
 
 ### Integration Tests
 
@@ -3232,25 +3377,29 @@ Each query must have at least one expected URL in top 10 results.
 
 | Command | Expected | Pass Criteria |
 |---------|----------|---------------|
-| `gi generate-docs` | Progress, count | Completes, count > 0 |
-| `gi generate-docs` (re-run) | 0 regenerated | Hash comparison works |
-| `gi embed` | Progress, count | Completes, count matches docs |
-| `gi embed` (re-run) | 0 embedded | Skips unchanged |
-| `gi embed --retry-failed` | Processes failed | Only failed docs processed |
-| `gi stats` | Coverage stats | Shows 100% after embed |
-| `gi stats` | Queue depths | Shows dirty_sources and pending_discussion_fetches counts |
-| `gi search "auth" --mode=lexical` | Results | Works without Ollama |
-| `gi search "auth"` | Hybrid results | Vector + FTS combined |
-| `gi search "auth"` (Ollama down) | FTS results + warning | Graceful degradation, warning in response |
-| `gi search "auth" --explain` | Rank breakdown | Shows vector/FTS/RRF |
-| `gi search "auth" --type=mr` | Filtered results | Only MRs |
-| `gi search "auth" --type=mrs` | Filtered results | Alias works |
-| `gi search "auth" --label=bug` | Filtered results | Only labeled docs |
-| `gi search "-DWITH_SSL"` | Results | Leading dash doesn't cause FTS error |
-| `gi search 'C++'` | Results | Special chars in query work |
-| `gi search "nonexistent123"` | No results | Graceful empty state |
-| `gi sync` | Full pipeline | All steps complete |
-| `gi sync --no-embed` | Skip embedding | Docs generated, not embedded |
+| `lore generate-docs` | Progress, count | Completes, count > 0 |
+| `lore generate-docs` (re-run) | 0 regenerated | Hash comparison works |
+| `lore embed` | Progress, count | Completes, count matches docs |
+| `lore embed` (re-run) | 0 embedded | Skips unchanged |
+| `lore embed --retry-failed` | Processes failed | Only failed docs processed |
+| `lore stats` | Coverage stats | Shows 100% after embed |
+| `lore stats` | Queue depths | Shows dirty_sources and pending_discussion_fetches counts |
+| `lore search "auth" --mode=lexical` | Results | Works without Ollama |
+| `lore search "auth"` | Hybrid results | Vector + FTS combined |
+| `lore search "auth"` (Ollama down) | FTS results + warning | Graceful degradation, warning in response |
+| `lore search "auth" --explain` | Rank breakdown | Shows vector/FTS/RRF |
+| `lore search "auth" --type=mr` | Filtered results | Only MRs |
+| `lore search "auth" --type=mrs` | Filtered results | Alias works |
+| `lore search "auth" --label=bug` | Filtered results | Only labeled docs |
+| `lore search "-DWITH_SSL"` | Results | Leading dash doesn't cause FTS error |
+| `lore search 'C++'` | Results | Special chars in query work |
+| `lore search "auth" --updated-after 2024-01-01` | Filtered results | Only recently updated docs |
+| `lore search "nonexistent123"` | No results | Graceful empty state |
+| `lore search "auth" --mode=semantic` (no embeddings) | Actionable error | Tells user to run `lore embed` first |
+| `lore sync` | Full pipeline | All steps complete |
+| `lore sync --no-embed` | Skip embedding | Docs generated, not embedded |
+| `lore generate-docs --full` | Progress, count | Keyset pagination completes without OFFSET degradation |
+| `lore stats --check --repair` | Repair results | FTS rebuilt, orphans cleaned |
 
 ---
 
@@ -3274,43 +3423,59 @@ Each query must have at least one expected URL in top 10 results.
 
 ## Success Criteria
 
-Checkpoint 3 is complete when:
+Checkpoint 3 is complete when all three gates pass:
+
+### Gate A: Lexical MVP
 
 1. **Lexical search works without Ollama**
-   - `gi search "query" --mode=lexical` returns relevant results
-   - All filters functional
+   - `lore search "query" --mode=lexical` returns relevant results
+   - All filters functional (including `--updated-after`)
    - FTS5 syntax errors prevented by query sanitization
    - Special characters in queries work correctly (`-DWITH_SSL`, `C++`)
+   - Search results hydrated in single DB round-trip (no N+1)
 
-2. **Semantic search works with Ollama**
-   - `gi embed` completes successfully
-   - `gi search "query"` returns semantically relevant results
+2. **Document generation is correct**
+   - Full and incremental modes use the same regenerator codepath
+   - `--full` uses keyset pagination (no OFFSET degradation)
+   - FTS triggers use COALESCE for NULL-safe operation
+
+### Gate B: Hybrid MVP
+
+3. **Semantic search works with Ollama**
+   - `lore embed` completes successfully
+   - `lore search "query"` returns semantically relevant results
    - `--explain` shows ranking breakdown
+   - `--mode=semantic` with 0% embedding coverage returns actionable error
 
-3. **Hybrid search combines both**
+4. **Hybrid search combines both**
    - Documents appearing in both retrievers rank higher
    - Graceful degradation when Ollama unavailable (falls back to FTS)
    - Transient embed failures don't fail the entire search
    - Warning message included in response on degradation
+   - Embedding pipeline uses keyset pagination for consistent paging
 
-4. **Incremental sync is efficient**
-   - `gi sync` only processes changed entities
+### Gate C: Sync MVP
+
+5. **Incremental sync is efficient**
+   - `lore sync` only processes changed entities
    - Re-embedding only happens for changed documents
    - Progress visible during long syncs
-   - Queue backoff prevents hot-loop retries on persistent failures
+   - Queue backoff actually prevents hot-loop retries (both queues set `next_attempt_at`)
+   - Shared backoff utility ensures consistent behavior across queues
 
-5. **Data integrity maintained**
+6. **Data integrity maintained**
    - All counts match between tables
    - No orphaned records
    - Hashes consistent
    - `get_existing_hash()` properly distinguishes "not found" from DB errors
+   - `--repair` uses FTS `rebuild` for correct-by-construction repair
 
-6. **Observability**
-   - `gi stats` shows queue depths and failed item counts
+7. **Observability**
+   - `lore stats` shows queue depths and failed item counts
    - Failed items visible for operator intervention
    - Deterministic ordering ensures consistent paging
 
-7. **Tests pass**
-   - Unit tests for core algorithms (including FTS sanitization, backoff)
+8. **Tests pass**
+   - Unit tests for core algorithms (including FTS sanitization, shared backoff, hydration)
    - Integration tests for pipelines
    - Golden queries return expected results

src/cli/commands/ingest.rs

@@ -17,6 +17,7 @@ use crate::ingestion::{
 };
 
 /// Result of ingest command for display.
+#[derive(Default)]
 pub struct IngestResult {
     pub resource_type: String,
     pub projects_synced: usize,
@@ -130,24 +131,7 @@ pub async fn run_ingest(
 
     let mut total = IngestResult {
         resource_type: resource_type.to_string(),
-        projects_synced: 0,
-        // Issue fields
-        issues_fetched: 0,
-        issues_upserted: 0,
-        issues_synced_discussions: 0,
-        issues_skipped_discussion_sync: 0,
-        // MR fields
-        mrs_fetched: 0,
-        mrs_upserted: 0,
-        mrs_synced_discussions: 0,
-        mrs_skipped_discussion_sync: 0,
-        assignees_linked: 0,
-        reviewers_linked: 0,
-        diffnotes_count: 0,
-        // Shared fields
-        labels_created: 0,
-        discussions_fetched: 0,
-        notes_upserted: 0,
+        ..Default::default()
     };
 
     let type_label = if resource_type == "issues" {

src/cli/commands/list.rs

@@ -232,8 +232,11 @@ fn query_issues(conn: &Connection, filters: &ListFilters) -> Result<ListResult>
     let mut params: Vec<Box<dyn rusqlite::ToSql>> = Vec::new();
 
     if let Some(project) = filters.project {
-        where_clauses.push("p.path_with_namespace LIKE ?");
-        params.push(Box::new(format!("%{project}%")));
+        // Exact match or suffix match after '/' to avoid partial matches
+        // e.g. "foo" matches "group/foo" but NOT "group/foobar"
+        where_clauses.push("(p.path_with_namespace = ? OR p.path_with_namespace LIKE ?)");
+        params.push(Box::new(project.to_string()));
+        params.push(Box::new(format!("%/{project}")));
     }
 
     if let Some(state) = filters.state
@@ -337,11 +340,11 @@ fn query_issues(conn: &Connection, filters: &ListFilters) -> Result<ListResult>
             i.updated_at,
             i.web_url,
             p.path_with_namespace,
-            (SELECT GROUP_CONCAT(l.name, ',')
+            (SELECT GROUP_CONCAT(l.name, X'1F')
              FROM issue_labels il
              JOIN labels l ON il.label_id = l.id
              WHERE il.issue_id = i.id) AS labels_csv,
-            (SELECT GROUP_CONCAT(ia.username, ',')
+            (SELECT GROUP_CONCAT(ia.username, X'1F')
              FROM issue_assignees ia
              WHERE ia.issue_id = i.id) AS assignees_csv,
             COALESCE(d.total, 0) AS discussion_count,
@@ -369,12 +372,12 @@ fn query_issues(conn: &Connection, filters: &ListFilters) -> Result<ListResult>
         .query_map(param_refs.as_slice(), |row| {
             let labels_csv: Option<String> = row.get(8)?;
             let labels = labels_csv
-                .map(|s| s.split(',').map(String::from).collect())
+                .map(|s| s.split('\x1F').map(String::from).collect())
                 .unwrap_or_default();
 
             let assignees_csv: Option<String> = row.get(9)?;
             let assignees = assignees_csv
-                .map(|s| s.split(',').map(String::from).collect())
+                .map(|s| s.split('\x1F').map(String::from).collect())
                 .unwrap_or_default();
 
             Ok(IssueListRow {
@@ -416,8 +419,11 @@ fn query_mrs(conn: &Connection, filters: &MrListFilters) -> Result<MrListResult>
     let mut params: Vec<Box<dyn rusqlite::ToSql>> = Vec::new();
 
     if let Some(project) = filters.project {
-        where_clauses.push("p.path_with_namespace LIKE ?");
-        params.push(Box::new(format!("%{project}%")));
+        // Exact match or suffix match after '/' to avoid partial matches
+        // e.g. "foo" matches "group/foo" but NOT "group/foobar"
+        where_clauses.push("(p.path_with_namespace = ? OR p.path_with_namespace LIKE ?)");
+        params.push(Box::new(project.to_string()));
+        params.push(Box::new(format!("%/{project}")));
     }
 
     if let Some(state) = filters.state
@@ -536,14 +542,14 @@ fn query_mrs(conn: &Connection, filters: &MrListFilters) -> Result<MrListResult>
             m.updated_at,
             m.web_url,
             p.path_with_namespace,
-            (SELECT GROUP_CONCAT(l.name, ',')
+            (SELECT GROUP_CONCAT(l.name, X'1F')
              FROM mr_labels ml
              JOIN labels l ON ml.label_id = l.id
              WHERE ml.merge_request_id = m.id) AS labels_csv,
-            (SELECT GROUP_CONCAT(ma.username, ',')
+            (SELECT GROUP_CONCAT(ma.username, X'1F')
              FROM mr_assignees ma
              WHERE ma.merge_request_id = m.id) AS assignees_csv,
-            (SELECT GROUP_CONCAT(mr.username, ',')
+            (SELECT GROUP_CONCAT(mr.username, X'1F')
              FROM mr_reviewers mr
              WHERE mr.merge_request_id = m.id) AS reviewers_csv,
             COALESCE(d.total, 0) AS discussion_count,
@@ -571,17 +577,17 @@ fn query_mrs(conn: &Connection, filters: &MrListFilters) -> Result<MrListResult>
         .query_map(param_refs.as_slice(), |row| {
             let labels_csv: Option<String> = row.get(11)?;
             let labels = labels_csv
-                .map(|s| s.split(',').map(String::from).collect())
+                .map(|s| s.split('\x1F').map(String::from).collect())
                 .unwrap_or_default();
 
             let assignees_csv: Option<String> = row.get(12)?;
             let assignees = assignees_csv
-                .map(|s| s.split(',').map(String::from).collect())
+                .map(|s| s.split('\x1F').map(String::from).collect())
                 .unwrap_or_default();
 
             let reviewers_csv: Option<String> = row.get(13)?;
             let reviewers = reviewers_csv
-                .map(|s| s.split(',').map(String::from).collect())
+                .map(|s| s.split('\x1F').map(String::from).collect())
                 .unwrap_or_default();
 
             let draft_int: i64 = row.get(3)?;
@@ -615,6 +621,10 @@ fn format_relative_time(ms_epoch: i64) -> String {
     let now = now_ms();
     let diff = now - ms_epoch;
 
+    if diff < 0 {
+        return "in the future".to_string();
+    }
+
     match diff {
         d if d < 60_000 => "just now".to_string(),
         d if d < 3_600_000 => format!("{} min ago", d / 60_000),

src/cli/commands/show.rs

@@ -150,8 +150,12 @@ fn find_issue(conn: &Connection, iid: i64, project_filter: Option<&str>) -> Resu
                     i.created_at, i.updated_at, i.web_url, p.path_with_namespace
              FROM issues i
              JOIN projects p ON i.project_id = p.id
-             WHERE i.iid = ? AND p.path_with_namespace LIKE ?",
-            vec![Box::new(iid), Box::new(format!("%{}%", project))],
+             WHERE i.iid = ? AND (p.path_with_namespace = ? OR p.path_with_namespace LIKE ?)",
+            vec![
+                Box::new(iid),
+                Box::new(project.to_string()),
+                Box::new(format!("%/{}", project)),
+            ],
         ),
         None => (
             "SELECT i.id, i.iid, i.title, i.description, i.state, i.author_username,
@@ -336,8 +340,12 @@ fn find_mr(conn: &Connection, iid: i64, project_filter: Option<&str>) -> Result<
                     m.web_url, p.path_with_namespace
              FROM merge_requests m
              JOIN projects p ON m.project_id = p.id
-             WHERE m.iid = ? AND p.path_with_namespace LIKE ?",
-            vec![Box::new(iid), Box::new(format!("%{}%", project))],
+             WHERE m.iid = ? AND (p.path_with_namespace = ? OR p.path_with_namespace LIKE ?)",
+            vec![
+                Box::new(iid),
+                Box::new(project.to_string()),
+                Box::new(format!("%/{}", project)),
+            ],
         ),
         None => (
             "SELECT m.id, m.iid, m.title, m.description, m.state, m.draft,

src/cli/mod.rs

@@ -11,13 +11,17 @@ use std::io::IsTerminal;
 #[command(version, about, long_about = None)]
 pub struct Cli {
     /// Path to config file
-    #[arg(short, long, global = true)]
+    #[arg(short = 'c', long, global = true)]
     pub config: Option<String>,
 
     /// Machine-readable JSON output (auto-enabled when piped)
     #[arg(long, global = true, env = "LORE_ROBOT")]
     pub robot: bool,
 
+    /// JSON output (global shorthand)
+    #[arg(short = 'J', long = "json", global = true)]
+    pub json: bool,
+
     #[command(subcommand)]
     pub command: Commands,
 }
@@ -25,17 +29,41 @@ pub struct Cli {
 impl Cli {
     /// Check if robot mode is active (explicit flag, env var, or non-TTY stdout)
     pub fn is_robot_mode(&self) -> bool {
-        self.robot || !std::io::stdout().is_terminal()
+        self.robot || self.json || !std::io::stdout().is_terminal()
     }
 }
 
 #[derive(Subcommand)]
 #[allow(clippy::large_enum_variant)]
 pub enum Commands {
+    /// List or show issues
+    Issues(IssuesArgs),
+
+    /// List or show merge requests
+    Mrs(MrsArgs),
+
+    /// Ingest data from GitLab
+    Ingest(IngestArgs),
+
+    /// Count entities in local database
+    Count(CountArgs),
+
+    /// Show sync state
+    Status,
+
+    /// Verify GitLab authentication
+    Auth,
+
+    /// Check environment health
+    Doctor,
+
+    /// Show version information
+    Version,
+
     /// Initialize configuration and database
     Init {
         /// Skip overwrite confirmation
-        #[arg(long)]
+        #[arg(short = 'f', long)]
         force: bool,
 
         /// Fail if prompts would be shown
@@ -43,149 +71,67 @@ pub enum Commands {
         non_interactive: bool,
     },
 
-    /// Verify GitLab authentication
-    AuthTest,
-
-    /// Check environment health
-    Doctor {
-        /// Output as JSON
-        #[arg(long)]
-        json: bool,
-    },
-
-    /// Show version information
-    Version,
-
     /// Create timestamped database backup
     Backup,
 
     /// Delete database and reset all state
     Reset {
         /// Skip confirmation prompt
-        #[arg(long)]
-        confirm: bool,
+        #[arg(short = 'y', long)]
+        yes: bool,
     },
 
     /// Run pending database migrations
     Migrate,
 
-    /// Show sync state
-    SyncStatus,
-
-    /// Ingest data from GitLab
-    Ingest {
-        /// Resource type to ingest
-        #[arg(long, value_parser = ["issues", "mrs"])]
-        r#type: String,
-
-        /// Filter to single project
-        #[arg(long)]
-        project: Option<String>,
-
-        /// Override stale sync lock
-        #[arg(long)]
-        force: bool,
-
-        /// Full re-sync: reset cursors and fetch all data from scratch
-        #[arg(long)]
-        full: bool,
-    },
-
-    /// List issues or MRs from local database
+    // --- Hidden backward-compat aliases ---
+    /// List issues or MRs (deprecated: use 'lore issues' or 'lore mrs')
+    #[command(hide = true)]
     List {
         /// Entity type to list
         #[arg(value_parser = ["issues", "mrs"])]
         entity: String,
 
-        /// Maximum results
         #[arg(long, default_value = "50")]
         limit: usize,
-
-        /// Filter by project path
         #[arg(long)]
         project: Option<String>,
-
-        /// Filter by state (opened|closed|all for issues; opened|merged|closed|locked|all for MRs)
         #[arg(long)]
         state: Option<String>,
-
-        /// Filter by author username
         #[arg(long)]
         author: Option<String>,
-
-        /// Filter by assignee username
         #[arg(long)]
         assignee: Option<String>,
-
-        /// Filter by label (repeatable, AND logic)
         #[arg(long)]
         label: Option<Vec<String>>,
-
-        /// Filter by milestone title (issues only)
         #[arg(long)]
         milestone: Option<String>,
-
-        /// Filter by time (7d, 2w, 1m, or YYYY-MM-DD)
         #[arg(long)]
         since: Option<String>,
-
-        /// Filter by due date (before this date, YYYY-MM-DD) (issues only)
         #[arg(long)]
         due_before: Option<String>,
-
-        /// Show only issues with a due date (issues only)
         #[arg(long)]
         has_due_date: bool,
-
-        /// Sort field
         #[arg(long, value_parser = ["updated", "created", "iid"], default_value = "updated")]
         sort: String,
-
-        /// Sort order
         #[arg(long, value_parser = ["desc", "asc"], default_value = "desc")]
         order: String,
-
-        /// Open first matching item in browser
         #[arg(long)]
         open: bool,
-
-        /// Output as JSON
-        #[arg(long)]
-        json: bool,
-
-        /// Show only draft MRs (MRs only)
         #[arg(long, conflicts_with = "no_draft")]
         draft: bool,
-
-        /// Exclude draft MRs (MRs only)
         #[arg(long, conflicts_with = "draft")]
         no_draft: bool,
-
-        /// Filter by reviewer username (MRs only)
         #[arg(long)]
         reviewer: Option<String>,
-
-        /// Filter by target branch (MRs only)
         #[arg(long)]
         target_branch: Option<String>,
-
-        /// Filter by source branch (MRs only)
         #[arg(long)]
         source_branch: Option<String>,
     },
 
-    /// Count entities in local database
-    Count {
-        /// Entity type to count
-        #[arg(value_parser = ["issues", "mrs", "discussions", "notes"])]
-        entity: String,
-
-        /// Filter by noteable type (for discussions/notes)
-        #[arg(long, value_parser = ["issue", "mr"])]
-        r#type: Option<String>,
-    },
-
-    /// Show detailed entity information
+    /// Show detailed entity information (deprecated: use 'lore issues <IID>' or 'lore mrs <IID>')
+    #[command(hide = true)]
     Show {
         /// Entity type to show
         #[arg(value_parser = ["issue", "mr"])]
@@ -194,12 +140,173 @@ pub enum Commands {
         /// Entity IID
         iid: i64,
 
-        /// Filter by project path (required if iid is ambiguous)
         #[arg(long)]
         project: Option<String>,
-
-        /// Output as JSON
-        #[arg(long)]
-        json: bool,
     },
+
+    /// Verify GitLab authentication (deprecated: use 'lore auth')
+    #[command(hide = true, name = "auth-test")]
+    AuthTest,
+
+    /// Show sync state (deprecated: use 'lore status')
+    #[command(hide = true, name = "sync-status")]
+    SyncStatus,
+}
+
+/// Arguments for `lore issues [IID]`
+#[derive(Parser)]
+pub struct IssuesArgs {
+    /// Issue IID (omit to list, provide to show details)
+    pub iid: Option<i64>,
+
+    /// Maximum results
+    #[arg(short = 'n', long = "limit", default_value = "50")]
+    pub limit: usize,
+
+    /// Filter by state (opened, closed, all)
+    #[arg(short = 's', long)]
+    pub state: Option<String>,
+
+    /// Filter by project path
+    #[arg(short = 'p', long)]
+    pub project: Option<String>,
+
+    /// Filter by author username
+    #[arg(short = 'a', long)]
+    pub author: Option<String>,
+
+    /// Filter by assignee username
+    #[arg(short = 'A', long)]
+    pub assignee: Option<String>,
+
+    /// Filter by label (repeatable, AND logic)
+    #[arg(short = 'l', long)]
+    pub label: Option<Vec<String>>,
+
+    /// Filter by milestone title
+    #[arg(short = 'm', long)]
+    pub milestone: Option<String>,
+
+    /// Filter by time (7d, 2w, 1m, or YYYY-MM-DD)
+    #[arg(long)]
+    pub since: Option<String>,
+
+    /// Filter by due date (before this date, YYYY-MM-DD)
+    #[arg(long = "due-before")]
+    pub due_before: Option<String>,
+
+    /// Show only issues with a due date
+    #[arg(long = "has-due")]
+    pub has_due: bool,
+
+    /// Sort field (updated, created, iid)
+    #[arg(long, value_parser = ["updated", "created", "iid"], default_value = "updated")]
+    pub sort: String,
+
+    /// Sort ascending (default: descending)
+    #[arg(long)]
+    pub asc: bool,
+
+    /// Open first matching item in browser
+    #[arg(short = 'o', long)]
+    pub open: bool,
+}
+
+/// Arguments for `lore mrs [IID]`
+#[derive(Parser)]
+pub struct MrsArgs {
+    /// MR IID (omit to list, provide to show details)
+    pub iid: Option<i64>,
+
+    /// Maximum results
+    #[arg(short = 'n', long = "limit", default_value = "50")]
+    pub limit: usize,
+
+    /// Filter by state (opened, merged, closed, locked, all)
+    #[arg(short = 's', long)]
+    pub state: Option<String>,
+
+    /// Filter by project path
+    #[arg(short = 'p', long)]
+    pub project: Option<String>,
+
+    /// Filter by author username
+    #[arg(short = 'a', long)]
+    pub author: Option<String>,
+
+    /// Filter by assignee username
+    #[arg(short = 'A', long)]
+    pub assignee: Option<String>,
+
+    /// Filter by reviewer username
+    #[arg(short = 'r', long)]
+    pub reviewer: Option<String>,
+
+    /// Filter by label (repeatable, AND logic)
+    #[arg(short = 'l', long)]
+    pub label: Option<Vec<String>>,
+
+    /// Filter by time (7d, 2w, 1m, or YYYY-MM-DD)
+    #[arg(long)]
+    pub since: Option<String>,
+
+    /// Show only draft MRs
+    #[arg(short = 'd', long, conflicts_with = "no_draft")]
+    pub draft: bool,
+
+    /// Exclude draft MRs
+    #[arg(short = 'D', long = "no-draft", conflicts_with = "draft")]
+    pub no_draft: bool,
+
+    /// Filter by target branch
+    #[arg(long)]
+    pub target: Option<String>,
+
+    /// Filter by source branch
+    #[arg(long)]
+    pub source: Option<String>,
+
+    /// Sort field (updated, created, iid)
+    #[arg(long, value_parser = ["updated", "created", "iid"], default_value = "updated")]
+    pub sort: String,
+
+    /// Sort ascending (default: descending)
+    #[arg(long)]
+    pub asc: bool,
+
+    /// Open first matching item in browser
+    #[arg(short = 'o', long)]
+    pub open: bool,
+}
+
+/// Arguments for `lore ingest [ENTITY]`
+#[derive(Parser)]
+pub struct IngestArgs {
+    /// Entity to ingest (issues, mrs). Omit to ingest everything.
+    #[arg(value_parser = ["issues", "mrs"])]
+    pub entity: Option<String>,
+
+    /// Filter to single project
+    #[arg(short = 'p', long)]
+    pub project: Option<String>,
+
+    /// Override stale sync lock
+    #[arg(short = 'f', long)]
+    pub force: bool,
+
+    /// Full re-sync: reset cursors and fetch all data from scratch
+    #[arg(long)]
+    pub full: bool,
+}
+
+/// Arguments for `lore count <ENTITY>`
+#[derive(Parser)]
+pub struct CountArgs {
+    /// Entity type to count (issues, mrs, discussions, notes)
+    #[arg(value_parser = ["issues", "mrs", "discussions", "notes"])]
+    pub entity: String,
+
+    /// Parent type filter: issue or mr (for discussions/notes)
+    #[arg(short = 'f', long = "for", value_parser = ["issue", "mr"])]
+    pub for_entity: Option<String>,
 }

src/core/error.rs

@@ -150,27 +150,49 @@ impl GiError {
             Self::Io(_) => ErrorCode::IoError,
             Self::Transform(_) => ErrorCode::TransformError,
             Self::NotFound(_) => ErrorCode::GitLabNotFound,
-            Self::Ambiguous(_) => ErrorCode::InternalError,
+            Self::Ambiguous(_) => ErrorCode::GitLabNotFound,
             Self::Other(_) => ErrorCode::InternalError,
         }
     }
 
-    /// Get a suggestion for how to fix this error.
+    /// Get a suggestion for how to fix this error, including inline examples.
     pub fn suggestion(&self) -> Option<&'static str> {
         match self {
-            Self::ConfigNotFound { .. } => Some("Run 'lore init' to create configuration"),
-            Self::ConfigInvalid { .. } => Some("Check config file syntax or run 'lore init' to recreate"),
-            Self::GitLabAuthFailed => Some("Verify token has read_api scope and is not expired"),
-            Self::GitLabNotFound { .. } => Some("Check the resource path exists and you have access"),
+            Self::ConfigNotFound { .. } => Some(
+                "Run 'lore init' to set up your GitLab connection.\n\n  Expected: ~/.config/lore/config.json",
+            ),
+            Self::ConfigInvalid { .. } => Some(
+                "Check config file syntax or run 'lore init' to recreate.\n\n  Example:\n    lore init\n    lore init --force",
+            ),
+            Self::GitLabAuthFailed => Some(
+                "Verify token has read_api scope and is not expired.\n\n  Example:\n    export GITLAB_TOKEN=glpat-xxxxxxxxxxxx\n    lore auth",
+            ),
+            Self::GitLabNotFound { .. } => Some(
+                "Check the resource path exists and you have access.\n\n  Example:\n    lore issues -p group/project\n    lore mrs -p group/project",
+            ),
             Self::GitLabRateLimited { .. } => Some("Wait and retry, or reduce request frequency"),
-            Self::GitLabNetworkError { .. } => Some("Check network connection and GitLab URL"),
-            Self::DatabaseLocked { .. } => Some("Wait for other sync to complete or use --force"),
-            Self::MigrationFailed { .. } => Some("Check database file permissions or reset with 'lore reset'"),
-            Self::TokenNotSet { .. } => Some("Export the token environment variable"),
-            Self::Database(_) => Some("Check database file permissions or reset with 'lore reset'"),
+            Self::GitLabNetworkError { .. } => Some(
+                "Check network connection and GitLab URL.\n\n  Example:\n    lore doctor\n    lore auth",
+            ),
+            Self::DatabaseLocked { .. } => Some(
+                "Wait for other sync to complete or use --force.\n\n  Example:\n    lore ingest --force\n    lore ingest issues --force",
+            ),
+            Self::MigrationFailed { .. } => Some(
+                "Check database file permissions or reset with 'lore reset'.\n\n  Example:\n    lore migrate\n    lore reset --yes",
+            ),
+            Self::TokenNotSet { .. } => Some(
+                "Export the token to your shell:\n\n  export GITLAB_TOKEN=glpat-xxxxxxxxxxxx\n\n  Your token needs the read_api scope.",
+            ),
+            Self::Database(_) => Some(
+                "Check database file permissions or reset with 'lore reset'.\n\n  Example:\n    lore doctor\n    lore reset --yes",
+            ),
             Self::Http(_) => Some("Check network connection"),
-            Self::NotFound(_) => Some("Verify the entity exists using 'lore list'"),
-            Self::Ambiguous(_) => Some("Use --project flag to disambiguate"),
+            Self::NotFound(_) => Some(
+                "Verify the entity exists.\n\n  Example:\n    lore issues\n    lore mrs",
+            ),
+            Self::Ambiguous(_) => Some(
+                "Use -p to choose a specific project.\n\n  Example:\n    lore issues 42 -p group/project-a\n    lore mrs 99 -p group/project-b",
+            ),
             _ => None,
         }
     }

src/core/time.rs

@@ -59,6 +59,22 @@ pub fn parse_since(input: &str) -> Option<i64> {
     iso_to_ms(input)
 }
 
+/// Convert ISO 8601 timestamp to milliseconds with strict error handling.
+/// Returns Err with a descriptive message if the timestamp is invalid.
+pub fn iso_to_ms_strict(iso_string: &str) -> Result<i64, String> {
+    DateTime::parse_from_rfc3339(iso_string)
+        .map(|dt| dt.timestamp_millis())
+        .map_err(|_| format!("Invalid timestamp: {}", iso_string))
+}
+
+/// Convert optional ISO 8601 timestamp to optional milliseconds (strict).
+pub fn iso_to_ms_opt_strict(iso_string: &Option<String>) -> Result<Option<i64>, String> {
+    match iso_string {
+        Some(s) => iso_to_ms_strict(s).map(Some),
+        None => Ok(None),
+    }
+}
+
 /// Format milliseconds epoch to human-readable full datetime.
 pub fn format_full_datetime(ms: i64) -> String {
     DateTime::from_timestamp_millis(ms)

src/gitlab/transformers/discussion.rs

@@ -1,8 +1,6 @@
 //! Discussion and note transformers: convert GitLab discussions to local schema.
 
-use chrono::DateTime;
-
-use crate::core::time::now_ms;
+use crate::core::time::{iso_to_ms, iso_to_ms_strict, now_ms};
 use crate::gitlab::types::{GitLabDiscussion, GitLabNote};
 
 /// Reference to the parent noteable (Issue or MergeRequest).
@@ -60,16 +58,9 @@ pub struct NormalizedNote {
     pub position_head_sha: Option<String>, // Head commit SHA for diff
 }
 
-/// Parse ISO 8601 timestamp to milliseconds, returning None on failure.
-fn parse_timestamp_opt(ts: &str) -> Option<i64> {
-    DateTime::parse_from_rfc3339(ts)
-        .ok()
-        .map(|dt| dt.timestamp_millis())
-}
-
 /// Parse ISO 8601 timestamp to milliseconds, defaulting to 0 on failure.
 fn parse_timestamp(ts: &str) -> i64 {
-    parse_timestamp_opt(ts).unwrap_or(0)
+    iso_to_ms(ts).unwrap_or(0)
 }
 
 /// Transform a GitLab discussion into normalized schema.
@@ -90,7 +81,7 @@ pub fn transform_discussion(
     let note_timestamps: Vec<i64> = gitlab_discussion
         .notes
         .iter()
-        .filter_map(|n| parse_timestamp_opt(&n.created_at))
+        .filter_map(|n| iso_to_ms(&n.created_at))
         .collect();
 
     let first_note_at = note_timestamps.iter().min().copied();
@@ -191,7 +182,7 @@ fn transform_single_note(
         resolved_at: note
             .resolved_at
             .as_ref()
-            .and_then(|ts| parse_timestamp_opt(ts)),
+            .and_then(|ts| iso_to_ms(ts)),
         position_old_path,
         position_new_path,
         position_old_line,
@@ -244,13 +235,6 @@ fn extract_position_fields(
     }
 }
 
-/// Parse ISO 8601 timestamp to milliseconds with strict error handling.
-/// Returns Err with the invalid timestamp in the error message.
-fn parse_timestamp_strict(ts: &str) -> Result<i64, String> {
-    DateTime::parse_from_rfc3339(ts)
-        .map(|dt| dt.timestamp_millis())
-        .map_err(|_| format!("Invalid timestamp: {}", ts))
-}
 
 /// Transform notes from a GitLab discussion with strict timestamp parsing.
 /// Returns Err if any timestamp is invalid - no silent fallback to 0.
@@ -275,10 +259,10 @@ fn transform_single_note_strict(
     now: i64,
 ) -> Result<NormalizedNote, String> {
     // Parse timestamps with strict error handling
-    let created_at = parse_timestamp_strict(&note.created_at)?;
-    let updated_at = parse_timestamp_strict(&note.updated_at)?;
+    let created_at = iso_to_ms_strict(&note.created_at)?;
+    let updated_at = iso_to_ms_strict(&note.updated_at)?;
     let resolved_at = match &note.resolved_at {
-        Some(ts) => Some(parse_timestamp_strict(ts)?),
+        Some(ts) => Some(iso_to_ms_strict(ts)?),
         None => None,
     };
 

src/gitlab/transformers/merge_request.rs

@@ -1,33 +1,8 @@
 //! Merge request transformer: converts GitLabMergeRequest to local schema.
 
-use chrono::DateTime;
-use std::time::{SystemTime, UNIX_EPOCH};
-
+use crate::core::time::{iso_to_ms_opt_strict, iso_to_ms_strict, now_ms};
 use crate::gitlab::types::GitLabMergeRequest;
 
-/// Get current time in milliseconds since Unix epoch.
-fn now_ms() -> i64 {
-    SystemTime::now()
-        .duration_since(UNIX_EPOCH)
-        .expect("Time went backwards")
-        .as_millis() as i64
-}
-
-/// Parse ISO 8601 timestamp to milliseconds since Unix epoch.
-fn iso_to_ms(ts: &str) -> Result<i64, String> {
-    DateTime::parse_from_rfc3339(ts)
-        .map(|dt| dt.timestamp_millis())
-        .map_err(|e| format!("Failed to parse timestamp '{}': {}", ts, e))
-}
-
-/// Parse optional ISO 8601 timestamp to optional milliseconds since Unix epoch.
-fn iso_to_ms_opt(ts: &Option<String>) -> Result<Option<i64>, String> {
-    match ts {
-        Some(s) => iso_to_ms(s).map(Some),
-        None => Ok(None),
-    }
-}
-
 /// Local schema representation of a merge request row.
 #[derive(Debug, Clone)]
 pub struct NormalizedMergeRequest {
@@ -77,12 +52,12 @@ pub fn transform_merge_request(
     local_project_id: i64,
 ) -> Result<MergeRequestWithMetadata, String> {
     // Parse required timestamps
-    let created_at = iso_to_ms(&gitlab_mr.created_at)?;
-    let updated_at = iso_to_ms(&gitlab_mr.updated_at)?;
+    let created_at = iso_to_ms_strict(&gitlab_mr.created_at)?;
+    let updated_at = iso_to_ms_strict(&gitlab_mr.updated_at)?;
 
     // Parse optional timestamps
-    let merged_at = iso_to_ms_opt(&gitlab_mr.merged_at)?;
-    let closed_at = iso_to_ms_opt(&gitlab_mr.closed_at)?;
+    let merged_at = iso_to_ms_opt_strict(&gitlab_mr.merged_at)?;
+    let closed_at = iso_to_ms_opt_strict(&gitlab_mr.closed_at)?;
 
     // Draft: prefer draft, fallback to work_in_progress
     let is_draft = gitlab_mr.draft || gitlab_mr.work_in_progress;

src/ingestion/discussions.rs

@@ -89,16 +89,10 @@ async fn ingest_discussions_for_issue(
 
     // Track discussions we've seen for stale removal
     let mut seen_discussion_ids: Vec<String> = Vec::new();
-    // Track if we've started receiving data (to distinguish empty result from failure)
-    let mut received_first_response = false;
     // Track if any error occurred during pagination
     let mut pagination_error: Option<crate::core::error::GiError> = None;
 
     while let Some(disc_result) = discussions_stream.next().await {
-        // Mark that we've received at least one response from the API
-        if !received_first_response {
-            received_first_response = true;
-        }
 
         // Handle errors - record but don't delete stale data
         let gitlab_discussion = match disc_result {
@@ -139,8 +133,6 @@ async fn ingest_discussions_for_issue(
         let tx = conn.unchecked_transaction()?;
 
         upsert_discussion(&tx, &normalized, payload_id)?;
-        result.discussions_upserted += 1;
-        seen_discussion_ids.push(normalized.gitlab_discussion_id.clone());
 
         // Get local discussion ID
         let local_discussion_id: i64 = tx.query_row(
@@ -151,6 +143,7 @@ async fn ingest_discussions_for_issue(
 
         // Transform and store notes
         let notes = transform_notes(&gitlab_discussion, local_project_id);
+        let notes_count = notes.len();
 
         // Delete existing notes for this discussion (full refresh)
         tx.execute(
@@ -178,26 +171,19 @@ async fn ingest_discussions_for_issue(
             )?;
 
             insert_note(&tx, local_discussion_id, &note, note_payload_id)?;
-            result.notes_upserted += 1;
         }
 
         tx.commit()?;
+
+        // Increment counters AFTER successful commit to keep metrics honest
+        result.discussions_upserted += 1;
+        result.notes_upserted += notes_count;
+        seen_discussion_ids.push(normalized.gitlab_discussion_id.clone());
     }
 
-    // Only remove stale discussions if pagination completed without errors
-    // AND we actually received a response (empty or not)
-    if pagination_error.is_none() && received_first_response {
-        let removed = remove_stale_discussions(conn, issue.local_issue_id, &seen_discussion_ids)?;
-        result.stale_discussions_removed = removed;
-
-        // Update discussions_synced_for_updated_at on the issue
-        update_issue_sync_timestamp(conn, issue.local_issue_id, issue.updated_at)?;
-    } else if pagination_error.is_none()
-        && !received_first_response
-        && seen_discussion_ids.is_empty()
-    {
-        // Stream was empty but no error - issue genuinely has no discussions
-        // This is safe to remove stale discussions (if any exist from before)
+    // Only remove stale discussions and advance watermark if pagination completed
+    // without errors. Safe for both empty results and populated results.
+    if pagination_error.is_none() {
         let removed = remove_stale_discussions(conn, issue.local_issue_id, &seen_discussion_ids)?;
         result.stale_discussions_removed = removed;
 
@@ -208,7 +194,6 @@ async fn ingest_discussions_for_issue(
             discussions_seen = seen_discussion_ids.len(),
             "Skipping stale removal due to pagination error"
         );
-        // Return the error to signal incomplete sync
         return Err(err);
     }
 

src/ingestion/mr_discussions.rs

@@ -155,14 +155,13 @@ pub fn write_prefetched_mr_discussions(
 
     // Write each discussion
     for disc in &prefetched.discussions {
-        result.discussions_fetched += 1;
-
-        // Count DiffNotes
-        result.diffnotes_count += disc
+        // Count DiffNotes upfront (independent of transaction)
+        let diffnotes_in_disc = disc
             .notes
             .iter()
             .filter(|n| n.position_new_path.is_some() || n.position_old_path.is_some())
             .count();
+        let notes_in_disc = disc.notes.len();
 
         // Start transaction
         let tx = conn.unchecked_transaction()?;
@@ -182,7 +181,6 @@ pub fn write_prefetched_mr_discussions(
 
         // Upsert discussion
         upsert_discussion(&tx, &disc.normalized, run_seen_at, payload_id)?;
-        result.discussions_upserted += 1;
 
         // Get local discussion ID
         let local_discussion_id: i64 = tx.query_row(
@@ -219,10 +217,15 @@ pub fn write_prefetched_mr_discussions(
             };
 
             upsert_note(&tx, local_discussion_id, note, run_seen_at, note_payload_id)?;
-            result.notes_upserted += 1;
         }
 
         tx.commit()?;
+
+        // Increment counters AFTER successful commit to keep metrics honest
+        result.discussions_fetched += 1;
+        result.discussions_upserted += 1;
+        result.notes_upserted += notes_in_disc;
+        result.diffnotes_count += diffnotes_in_disc;
     }
 
     // Only sweep stale data and advance watermark on full success
@@ -343,8 +346,6 @@ async fn ingest_discussions_for_mr(
                 break;
             }
         };
-        result.discussions_fetched += 1;
-
         // CRITICAL: Parse notes BEFORE any destructive DB operations
         let notes = match transform_notes_with_diff_position(&gitlab_discussion, local_project_id) {
             Ok(notes) => notes,
@@ -361,11 +362,12 @@ async fn ingest_discussions_for_mr(
             }
         };
 
-        // Count DiffNotes
-        result.diffnotes_count += notes
+        // Count DiffNotes upfront (independent of transaction)
+        let diffnotes_in_disc = notes
             .iter()
             .filter(|n| n.position_new_path.is_some() || n.position_old_path.is_some())
             .count();
+        let notes_count = notes.len();
 
         // Transform discussion
         let normalized_discussion =
@@ -389,7 +391,6 @@ async fn ingest_discussions_for_mr(
 
         // Upsert discussion with run_seen_at
         upsert_discussion(&tx, &normalized_discussion, run_seen_at, payload_id)?;
-        result.discussions_upserted += 1;
 
         // Get local discussion ID
         let local_discussion_id: i64 = tx.query_row(
@@ -433,10 +434,15 @@ async fn ingest_discussions_for_mr(
             };
 
             upsert_note(&tx, local_discussion_id, note, run_seen_at, note_payload_id)?;
-            result.notes_upserted += 1;
         }
 
         tx.commit()?;
+
+        // Increment counters AFTER successful commit to keep metrics honest
+        result.discussions_fetched += 1;
+        result.discussions_upserted += 1;
+        result.notes_upserted += notes_count;
+        result.diffnotes_count += diffnotes_in_disc;
     }
 
     // Only sweep stale data and advance watermark on full success

src/main.rs

@@ -18,7 +18,7 @@ use lore::cli::commands::{
     run_doctor, run_ingest, run_init, run_list_issues, run_list_mrs, run_show_issue, run_show_mr,
     run_sync_status,
 };
-use lore::cli::{Cli, Commands};
+use lore::cli::{Cli, Commands, CountArgs, IngestArgs, IssuesArgs, MrsArgs};
 use lore::core::db::{create_connection, get_schema_version, run_migrations};
 use lore::core::error::{GiError, RobotErrorOutput};
 use lore::core::paths::get_config_path;
@@ -47,33 +47,25 @@ async fn main() {
     let robot_mode = cli.is_robot_mode();
 
     let result = match cli.command {
+        Commands::Issues(args) => handle_issues(cli.config.as_deref(), args, robot_mode).await,
+        Commands::Mrs(args) => handle_mrs(cli.config.as_deref(), args, robot_mode).await,
+        Commands::Ingest(args) => handle_ingest(cli.config.as_deref(), args, robot_mode).await,
+        Commands::Count(args) => {
+            handle_count(cli.config.as_deref(), args, robot_mode).await
+        }
+        Commands::Status => handle_sync_status_cmd(cli.config.as_deref(), robot_mode).await,
+        Commands::Auth => handle_auth_test(cli.config.as_deref(), robot_mode).await,
+        Commands::Doctor => handle_doctor(cli.config.as_deref(), robot_mode).await,
+        Commands::Version => handle_version(robot_mode),
         Commands::Init {
             force,
             non_interactive,
         } => handle_init(cli.config.as_deref(), force, non_interactive, robot_mode).await,
-        Commands::AuthTest => handle_auth_test(cli.config.as_deref(), robot_mode).await,
-        Commands::Doctor { json } => handle_doctor(cli.config.as_deref(), json || robot_mode).await,
-        Commands::Version => handle_version(robot_mode),
         Commands::Backup => handle_backup(robot_mode),
-        Commands::Reset { confirm: _ } => handle_reset(robot_mode),
+        Commands::Reset { yes: _ } => handle_reset(robot_mode),
         Commands::Migrate => handle_migrate(cli.config.as_deref(), robot_mode).await,
-        Commands::SyncStatus => handle_sync_status(cli.config.as_deref(), robot_mode).await,
-        Commands::Ingest {
-            r#type,
-            project,
-            force,
-            full,
-        } => {
-            handle_ingest(
-                cli.config.as_deref(),
-                &r#type,
-                project.as_deref(),
-                force,
-                full,
-                robot_mode,
-            )
-            .await
-        }
+
+        // --- Backward-compat: deprecated aliases ---
         Commands::List {
             entity,
             limit,
@@ -89,14 +81,17 @@ async fn main() {
             sort,
             order,
             open,
-            json,
             draft,
             no_draft,
             reviewer,
             target_branch,
             source_branch,
         } => {
-            handle_list(
+            eprintln!(
+                "{}",
+                style("warning: 'lore list' is deprecated, use 'lore issues' or 'lore mrs'").yellow()
+            );
+            handle_list_compat(
                 cli.config.as_deref(),
                 &entity,
                 limit,
@@ -112,7 +107,7 @@ async fn main() {
                 &sort,
                 &order,
                 open,
-                json || robot_mode,
+                robot_mode,
                 draft,
                 no_draft,
                 reviewer.as_deref(),
@@ -121,24 +116,42 @@ async fn main() {
             )
             .await
         }
-        Commands::Count { entity, r#type } => {
-            handle_count(cli.config.as_deref(), &entity, r#type.as_deref(), robot_mode).await
-        }
         Commands::Show {
             entity,
             iid,
             project,
-            json,
         } => {
-            handle_show(
+            eprintln!(
+                "{}",
+                style(format!(
+                    "warning: 'lore show' is deprecated, use 'lore {}s {}'",
+                    entity, iid
+                ))
+                .yellow()
+            );
+            handle_show_compat(
                 cli.config.as_deref(),
                 &entity,
                 iid,
                 project.as_deref(),
-                json || robot_mode,
+                robot_mode,
             )
             .await
         }
+        Commands::AuthTest => {
+            eprintln!(
+                "{}",
+                style("warning: 'lore auth-test' is deprecated, use 'lore auth'").yellow()
+            );
+            handle_auth_test(cli.config.as_deref(), robot_mode).await
+        }
+        Commands::SyncStatus => {
+            eprintln!(
+                "{}",
+                style("warning: 'lore sync-status' is deprecated, use 'lore status'").yellow()
+            );
+            handle_sync_status_cmd(cli.config.as_deref(), robot_mode).await
+        }
     };
 
     if let Err(e) = result {
@@ -207,6 +220,259 @@ fn handle_error(e: Box<dyn std::error::Error>, robot_mode: bool) -> ! {
     std::process::exit(1);
 }
 
+// ============================================================================
+// Primary command handlers
+// ============================================================================
+
+async fn handle_issues(
+    config_override: Option<&str>,
+    args: IssuesArgs,
+    robot_mode: bool,
+) -> Result<(), Box<dyn std::error::Error>> {
+    let config = Config::load(config_override)?;
+    let order = if args.asc { "asc" } else { "desc" };
+
+    if let Some(iid) = args.iid {
+        // Show mode
+        let result = run_show_issue(&config, iid, args.project.as_deref())?;
+        if robot_mode {
+            print_show_issue_json(&result);
+        } else {
+            print_show_issue(&result);
+        }
+    } else {
+        // List mode
+        let filters = ListFilters {
+            limit: args.limit,
+            project: args.project.as_deref(),
+            state: args.state.as_deref(),
+            author: args.author.as_deref(),
+            assignee: args.assignee.as_deref(),
+            labels: args.label.as_deref(),
+            milestone: args.milestone.as_deref(),
+            since: args.since.as_deref(),
+            due_before: args.due_before.as_deref(),
+            has_due_date: args.has_due,
+            sort: &args.sort,
+            order,
+        };
+
+        let result = run_list_issues(&config, filters)?;
+
+        if args.open {
+            open_issue_in_browser(&result);
+        } else if robot_mode {
+            print_list_issues_json(&result);
+        } else {
+            print_list_issues(&result);
+        }
+    }
+
+    Ok(())
+}
+
+async fn handle_mrs(
+    config_override: Option<&str>,
+    args: MrsArgs,
+    robot_mode: bool,
+) -> Result<(), Box<dyn std::error::Error>> {
+    let config = Config::load(config_override)?;
+    let order = if args.asc { "asc" } else { "desc" };
+
+    if let Some(iid) = args.iid {
+        // Show mode
+        let result = run_show_mr(&config, iid, args.project.as_deref())?;
+        if robot_mode {
+            print_show_mr_json(&result);
+        } else {
+            print_show_mr(&result);
+        }
+    } else {
+        // List mode
+        let filters = MrListFilters {
+            limit: args.limit,
+            project: args.project.as_deref(),
+            state: args.state.as_deref(),
+            author: args.author.as_deref(),
+            assignee: args.assignee.as_deref(),
+            reviewer: args.reviewer.as_deref(),
+            labels: args.label.as_deref(),
+            since: args.since.as_deref(),
+            draft: args.draft,
+            no_draft: args.no_draft,
+            target_branch: args.target.as_deref(),
+            source_branch: args.source.as_deref(),
+            sort: &args.sort,
+            order,
+        };
+
+        let result = run_list_mrs(&config, filters)?;
+
+        if args.open {
+            open_mr_in_browser(&result);
+        } else if robot_mode {
+            print_list_mrs_json(&result);
+        } else {
+            print_list_mrs(&result);
+        }
+    }
+
+    Ok(())
+}
+
+async fn handle_ingest(
+    config_override: Option<&str>,
+    args: IngestArgs,
+    robot_mode: bool,
+) -> Result<(), Box<dyn std::error::Error>> {
+    let config = Config::load(config_override)?;
+
+    match args.entity.as_deref() {
+        Some(resource_type) => {
+            // Single entity ingest
+            let result = run_ingest(
+                &config,
+                resource_type,
+                args.project.as_deref(),
+                args.force,
+                args.full,
+                robot_mode,
+            )
+            .await?;
+
+            if robot_mode {
+                print_ingest_summary_json(&result);
+            } else {
+                print_ingest_summary(&result);
+            }
+        }
+        None => {
+            // Ingest everything: issues then MRs
+            if !robot_mode {
+                println!(
+                    "{}",
+                    style("Ingesting all content (issues + merge requests)...").blue()
+                );
+                println!();
+            }
+
+            let issues_result = run_ingest(
+                &config,
+                "issues",
+                args.project.as_deref(),
+                args.force,
+                args.full,
+                robot_mode,
+            )
+            .await?;
+
+            let mrs_result = run_ingest(
+                &config,
+                "mrs",
+                args.project.as_deref(),
+                args.force,
+                args.full,
+                robot_mode,
+            )
+            .await?;
+
+            if robot_mode {
+                print_combined_ingest_json(&issues_result, &mrs_result);
+            } else {
+                print_ingest_summary(&issues_result);
+                print_ingest_summary(&mrs_result);
+            }
+        }
+    }
+
+    Ok(())
+}
+
+/// JSON output for combined ingest (issues + mrs).
+#[derive(Serialize)]
+struct CombinedIngestOutput {
+    ok: bool,
+    data: CombinedIngestData,
+}
+
+#[derive(Serialize)]
+struct CombinedIngestData {
+    resource_type: String,
+    issues: CombinedIngestEntityStats,
+    merge_requests: CombinedIngestEntityStats,
+}
+
+#[derive(Serialize)]
+struct CombinedIngestEntityStats {
+    projects_synced: usize,
+    fetched: usize,
+    upserted: usize,
+    labels_created: usize,
+    discussions_fetched: usize,
+    notes_upserted: usize,
+}
+
+fn print_combined_ingest_json(
+    issues: &lore::cli::commands::ingest::IngestResult,
+    mrs: &lore::cli::commands::ingest::IngestResult,
+) {
+    let output = CombinedIngestOutput {
+        ok: true,
+        data: CombinedIngestData {
+            resource_type: "all".to_string(),
+            issues: CombinedIngestEntityStats {
+                projects_synced: issues.projects_synced,
+                fetched: issues.issues_fetched,
+                upserted: issues.issues_upserted,
+                labels_created: issues.labels_created,
+                discussions_fetched: issues.discussions_fetched,
+                notes_upserted: issues.notes_upserted,
+            },
+            merge_requests: CombinedIngestEntityStats {
+                projects_synced: mrs.projects_synced,
+                fetched: mrs.mrs_fetched,
+                upserted: mrs.mrs_upserted,
+                labels_created: mrs.labels_created,
+                discussions_fetched: mrs.discussions_fetched,
+                notes_upserted: mrs.notes_upserted,
+            },
+        },
+    };
+
+    println!("{}", serde_json::to_string(&output).unwrap());
+}
+
+async fn handle_count(
+    config_override: Option<&str>,
+    args: CountArgs,
+    robot_mode: bool,
+) -> Result<(), Box<dyn std::error::Error>> {
+    let config = Config::load(config_override)?;
+
+    let result = run_count(&config, &args.entity, args.for_entity.as_deref())?;
+    if robot_mode {
+        print_count_json(&result);
+    } else {
+        print_count(&result);
+    }
+    Ok(())
+}
+
+async fn handle_sync_status_cmd(
+    config_override: Option<&str>,
+    robot_mode: bool,
+) -> Result<(), Box<dyn std::error::Error>> {
+    let config = Config::load(config_override)?;
+
+    let result = run_sync_status(&config)?;
+    if robot_mode {
+        print_sync_status_json(&result);
+    } else {
+        print_sync_status(&result);
+    }
+    Ok(())
+}
+
 async fn handle_init(
     config_override: Option<&str>,
     force: bool,
@@ -389,11 +655,11 @@ async fn handle_auth_test(
 
 async fn handle_doctor(
     config_override: Option<&str>,
-    json: bool,
+    robot_mode: bool,
 ) -> Result<(), Box<dyn std::error::Error>> {
     let result = run_doctor(config_override).await;
 
-    if json {
+    if robot_mode {
         println!("{}", serde_json::to_string_pretty(&result)?);
     } else {
         print_doctor_results(&result);
@@ -406,191 +672,6 @@ async fn handle_doctor(
     Ok(())
 }
 
-async fn handle_ingest(
-    config_override: Option<&str>,
-    resource_type: &str,
-    project_filter: Option<&str>,
-    force: bool,
-    full: bool,
-    robot_mode: bool,
-) -> Result<(), Box<dyn std::error::Error>> {
-    let config = Config::load(config_override)?;
-
-    match run_ingest(&config, resource_type, project_filter, force, full, robot_mode).await {
-        Ok(result) => {
-            if robot_mode {
-                print_ingest_summary_json(&result);
-            } else {
-                print_ingest_summary(&result);
-            }
-            Ok(())
-        }
-        Err(e) => {
-            eprintln!("{}", style(format!("Error: {e}")).red());
-            std::process::exit(1);
-        }
-    }
-}
-
-#[allow(clippy::too_many_arguments)]
-async fn handle_list(
-    config_override: Option<&str>,
-    entity: &str,
-    limit: usize,
-    project_filter: Option<&str>,
-    state_filter: Option<&str>,
-    author_filter: Option<&str>,
-    assignee_filter: Option<&str>,
-    label_filter: Option<&[String]>,
-    milestone_filter: Option<&str>,
-    since_filter: Option<&str>,
-    due_before_filter: Option<&str>,
-    has_due_date: bool,
-    sort: &str,
-    order: &str,
-    open_browser: bool,
-    json_output: bool,
-    draft: bool,
-    no_draft: bool,
-    reviewer_filter: Option<&str>,
-    target_branch_filter: Option<&str>,
-    source_branch_filter: Option<&str>,
-) -> Result<(), Box<dyn std::error::Error>> {
-    let config = Config::load(config_override)?;
-
-    match entity {
-        "issues" => {
-            let filters = ListFilters {
-                limit,
-                project: project_filter,
-                state: state_filter,
-                author: author_filter,
-                assignee: assignee_filter,
-                labels: label_filter,
-                milestone: milestone_filter,
-                since: since_filter,
-                due_before: due_before_filter,
-                has_due_date,
-                sort,
-                order,
-            };
-
-            let result = run_list_issues(&config, filters)?;
-
-            if open_browser {
-                open_issue_in_browser(&result);
-            } else if json_output {
-                print_list_issues_json(&result);
-            } else {
-                print_list_issues(&result);
-            }
-
-            Ok(())
-        }
-        "mrs" => {
-            let filters = MrListFilters {
-                limit,
-                project: project_filter,
-                state: state_filter,
-                author: author_filter,
-                assignee: assignee_filter,
-                reviewer: reviewer_filter,
-                labels: label_filter,
-                since: since_filter,
-                draft,
-                no_draft,
-                target_branch: target_branch_filter,
-                source_branch: source_branch_filter,
-                sort,
-                order,
-            };
-
-            let result = run_list_mrs(&config, filters)?;
-
-            if open_browser {
-                open_mr_in_browser(&result);
-            } else if json_output {
-                print_list_mrs_json(&result);
-            } else {
-                print_list_mrs(&result);
-            }
-
-            Ok(())
-        }
-        _ => {
-            eprintln!("{}", style(format!("Unknown entity: {entity}")).red());
-            std::process::exit(1);
-        }
-    }
-}
-
-async fn handle_count(
-    config_override: Option<&str>,
-    entity: &str,
-    type_filter: Option<&str>,
-    robot_mode: bool,
-) -> Result<(), Box<dyn std::error::Error>> {
-    let config = Config::load(config_override)?;
-
-    let result = run_count(&config, entity, type_filter)?;
-    if robot_mode {
-        print_count_json(&result);
-    } else {
-        print_count(&result);
-    }
-    Ok(())
-}
-
-async fn handle_sync_status(
-    config_override: Option<&str>,
-    robot_mode: bool,
-) -> Result<(), Box<dyn std::error::Error>> {
-    let config = Config::load(config_override)?;
-
-    let result = run_sync_status(&config)?;
-    if robot_mode {
-        print_sync_status_json(&result);
-    } else {
-        print_sync_status(&result);
-    }
-    Ok(())
-}
-
-async fn handle_show(
-    config_override: Option<&str>,
-    entity: &str,
-    iid: i64,
-    project_filter: Option<&str>,
-    json: bool,
-) -> Result<(), Box<dyn std::error::Error>> {
-    let config = Config::load(config_override)?;
-
-    match entity {
-        "issue" => {
-            let result = run_show_issue(&config, iid, project_filter)?;
-            if json {
-                print_show_issue_json(&result);
-            } else {
-                print_show_issue(&result);
-            }
-            Ok(())
-        }
-        "mr" => {
-            let result = run_show_mr(&config, iid, project_filter)?;
-            if json {
-                print_show_mr_json(&result);
-            } else {
-                print_show_mr(&result);
-            }
-            Ok(())
-        }
-        _ => {
-            eprintln!("{}", style(format!("Unknown entity: {entity}")).red());
-            std::process::exit(1);
-        }
-    }
-}
-
 /// JSON output for version command.
 #[derive(Serialize)]
 struct VersionOutput {
@@ -758,3 +839,134 @@ async fn handle_migrate(
 
     Ok(())
 }
+
+// ============================================================================
+// Backward-compat handlers (deprecated, delegate to new handlers)
+// ============================================================================
+
+#[allow(clippy::too_many_arguments)]
+async fn handle_list_compat(
+    config_override: Option<&str>,
+    entity: &str,
+    limit: usize,
+    project_filter: Option<&str>,
+    state_filter: Option<&str>,
+    author_filter: Option<&str>,
+    assignee_filter: Option<&str>,
+    label_filter: Option<&[String]>,
+    milestone_filter: Option<&str>,
+    since_filter: Option<&str>,
+    due_before_filter: Option<&str>,
+    has_due_date: bool,
+    sort: &str,
+    order: &str,
+    open_browser: bool,
+    json_output: bool,
+    draft: bool,
+    no_draft: bool,
+    reviewer_filter: Option<&str>,
+    target_branch_filter: Option<&str>,
+    source_branch_filter: Option<&str>,
+) -> Result<(), Box<dyn std::error::Error>> {
+    let config = Config::load(config_override)?;
+
+    match entity {
+        "issues" => {
+            let filters = ListFilters {
+                limit,
+                project: project_filter,
+                state: state_filter,
+                author: author_filter,
+                assignee: assignee_filter,
+                labels: label_filter,
+                milestone: milestone_filter,
+                since: since_filter,
+                due_before: due_before_filter,
+                has_due_date,
+                sort,
+                order,
+            };
+
+            let result = run_list_issues(&config, filters)?;
+
+            if open_browser {
+                open_issue_in_browser(&result);
+            } else if json_output {
+                print_list_issues_json(&result);
+            } else {
+                print_list_issues(&result);
+            }
+
+            Ok(())
+        }
+        "mrs" => {
+            let filters = MrListFilters {
+                limit,
+                project: project_filter,
+                state: state_filter,
+                author: author_filter,
+                assignee: assignee_filter,
+                reviewer: reviewer_filter,
+                labels: label_filter,
+                since: since_filter,
+                draft,
+                no_draft,
+                target_branch: target_branch_filter,
+                source_branch: source_branch_filter,
+                sort,
+                order,
+            };
+
+            let result = run_list_mrs(&config, filters)?;
+
+            if open_browser {
+                open_mr_in_browser(&result);
+            } else if json_output {
+                print_list_mrs_json(&result);
+            } else {
+                print_list_mrs(&result);
+            }
+
+            Ok(())
+        }
+        _ => {
+            eprintln!("{}", style(format!("Unknown entity: {entity}")).red());
+            std::process::exit(1);
+        }
+    }
+}
+
+async fn handle_show_compat(
+    config_override: Option<&str>,
+    entity: &str,
+    iid: i64,
+    project_filter: Option<&str>,
+    json: bool,
+) -> Result<(), Box<dyn std::error::Error>> {
+    let config = Config::load(config_override)?;
+
+    match entity {
+        "issue" => {
+            let result = run_show_issue(&config, iid, project_filter)?;
+            if json {
+                print_show_issue_json(&result);
+            } else {
+                print_show_issue(&result);
+            }
+            Ok(())
+        }
+        "mr" => {
+            let result = run_show_mr(&config, iid, project_filter)?;
+            if json {
+                print_show_mr_json(&result);
+            } else {
+                print_show_mr(&result);
+            }
+            Ok(())
+        }
+        _ => {
+            eprintln!("{}", style(format!("Unknown entity: {entity}")).red());
+            std::process::exit(1);
+        }
+    }
+}