aiwcli 0.12.6 → 0.12.7

package/dist/templates/cc-native/.claude/commands/cc-native/rlm/ask.md

@@ -1,136 +1,136 @@
----
-description: Answer a question using past session transcripts
----
-# RLM Ask — Intelligent Q&A About Past Work
-
-Ask any question about past Claude Code sessions. Auto-selects the best search method (semantic or keyword) based on available indexes.
-
-## Input
-
-A natural language question. Examples:
-- "How did we implement the plan review system?"
-- "What approach did we use for hook error handling?"
-- "When did we last work on the bridge app?"
-
-## Workflow
-
-### Step 1: Auto-Index Check
-Check if indexes exist, build if missing (one-time setup):
-```bash
-INDEX_DIR="$HOME/.claude/rlm-index"
-if [ ! -d "$INDEX_DIR" ] || [ -z "$(ls -A $INDEX_DIR 2>/dev/null)" ]; then
-  echo "🔄 Building indexes (one-time setup, ~10-30s)..."
-
-  # Run indexer with explicit error handling
-  if ! bun .aiwcli/_cc-native/lib-ts/rlm/transcript-indexer.ts --batch; then
-    echo ""
-    echo "❌ Index build failed"
-    echo ""
-    echo "Troubleshooting:"
-    echo "- Check file permissions in ~/.claude/projects/"
-    echo "- Look for corrupted .jsonl files"
-    echo "- Try manual rebuild: /rlm:index"
-    echo ""
-    exit 1
-  fi
-
-  echo "✅ Indexes built successfully"
-fi
-```
-
-### Step 2: Search Strategy Selection
-Auto-detect best search method with auto-model-pull:
-
-```bash
-VECTOR_DB="$HOME/.claude/rlm-vectors.db"
-OLLAMA_RUNNING=false
-MODEL_PRESENT=false
-
-# Check if Ollama is running
-if curl -s http://localhost:11434/api/tags >/dev/null 2>&1; then
-  OLLAMA_RUNNING=true
-
-  # Check if nomic-embed-text model is installed
-  if ollama list | grep -q "nomic-embed-text"; then
-    MODEL_PRESENT=true
-  else
-    echo "📥 Ollama running but nomic-embed-text model missing"
-    echo "🔄 Downloading model (~400MB, this may take 1-2 minutes)..."
-
-    # Auto-pull model with error handling
-    if ollama pull nomic-embed-text; then
-      echo "✅ Model downloaded successfully"
-      MODEL_PRESENT=true
-    else
-      echo "⚠️  Model download failed, falling back to keyword search"
-      OLLAMA_RUNNING=false
-    fi
-  fi
-fi
-
-# Use semantic search if vector DB + Ollama + model available
-if [ -f "$VECTOR_DB" ] && [ "$OLLAMA_RUNNING" = true ] && [ "$MODEL_PRESENT" = true ]; then
-  echo "Using semantic search (vector similarity)..."
-  SEARCH_RESULTS=$(bun .aiwcli/_cc-native/lib-ts/rlm/retrieval-pipeline.ts "$ARGUMENTS")
-else
-  echo "Using keyword search (weighted scoring)..."
-  SEARCH_RESULTS=$(bun .aiwcli/_cc-native/lib-ts/rlm/transcript-searcher.ts "$ARGUMENTS" --top=10)
-fi
-```
-
-If zero results, tell the user and suggest refining the query or running `/rlm:index`.
-
-### Step 3: Load Transcript Segments
-Parse search results and load top N segments (N = result count, capped at 5):
-
-For each of the top 1-5 results, load the relevant transcript segments:
-
-```bash
-bun .aiwcli/_cc-native/lib-ts/rlm/transcript-loader.ts "<source_path>" --lines=<start>-<end>
-```
-
-Use the `matching_segments` from search results to target specific line ranges. If no segments matched, load the first 200 lines.
-
-### Step 4: Parallel Analysis
-Spawn N parallel sub-agents (Task tool, subagent_type=general-purpose) where N = min(result_count, 5):
-- 1 result → 1 agent
-- 3 results → 3 agents
-- 10 results → 5 agents (capped)
-
-Each agent analyzes one segment:
-
-> Given this session transcript segment, extract all information relevant to: **{user's question}**
->
-> Include: specific details, code snippets, decisions made, exact quotes, and file paths mentioned.
-> Cite line numbers where possible.
->
-> Transcript:
-> {loaded segment content}
-
-### Step 5: Synthesis
-Combine findings into a coherent answer:
-1. Group findings by theme/topic
-2. Add citations: "In session {date} (project: {project}): {finding}"
-3. Note contradictions or evolution across sessions
-4. Highlight most recent/relevant information
-
-### Output Format
-```
-## Answer: {user's question}
-
-{Synthesized answer with inline citations}
-
-### Sources
-| Date | Project | Session | Relevance |
-|------|---------|---------|-----------|
-| YYYY-MM-DD | project-name | session-id | Brief match reason |
-```
-
-Always include the sources table. If findings are sparse, say so honestly rather than speculating.
-
-## Notes
-
-- **First run:** Auto-builds indexes (10-30s delay)
-- **Semantic search:** Uses vector similarity (higher quality, requires Ollama)
-- **Keyword search:** Uses weighted scoring (faster, always available)
-- **Graceful fallback:** No Ollama? Automatically uses keyword search
+---
+description: Answer a question using past session transcripts
+---
+# RLM Ask — Intelligent Q&A About Past Work
+
+Ask any question about past Claude Code sessions. Auto-selects the best search method (semantic or keyword) based on available indexes.
+
+## Input
+
+A natural language question. Examples:
+- "How did we implement the plan review system?"
+- "What approach did we use for hook error handling?"
+- "When did we last work on the bridge app?"
+
+## Workflow
+
+### Step 1: Auto-Index Check
+Check if indexes exist, build if missing (one-time setup):
+```bash
+INDEX_DIR="$HOME/.claude/rlm-index"
+if [ ! -d "$INDEX_DIR" ] || [ -z "$(ls -A $INDEX_DIR 2>/dev/null)" ]; then
+  echo "🔄 Building indexes (one-time setup, ~10-30s)..."
+
+  # Run indexer with explicit error handling
+  if ! bun .aiwcli/_cc-native/lib-ts/rlm/transcript-indexer.ts --batch; then
+    echo ""
+    echo "❌ Index build failed"
+    echo ""
+    echo "Troubleshooting:"
+    echo "- Check file permissions in ~/.claude/projects/"
+    echo "- Look for corrupted .jsonl files"
+    echo "- Try manual rebuild: /rlm:index"
+    echo ""
+    exit 1
+  fi
+
+  echo "✅ Indexes built successfully"
+fi
+```
+
+### Step 2: Search Strategy Selection
+Auto-detect best search method with auto-model-pull:
+
+```bash
+VECTOR_DB="$HOME/.claude/rlm-vectors.db"
+OLLAMA_RUNNING=false
+MODEL_PRESENT=false
+
+# Check if Ollama is running
+if curl -s http://localhost:11434/api/tags >/dev/null 2>&1; then
+  OLLAMA_RUNNING=true
+
+  # Check if nomic-embed-text model is installed
+  if ollama list | grep -q "nomic-embed-text"; then
+    MODEL_PRESENT=true
+  else
+    echo "📥 Ollama running but nomic-embed-text model missing"
+    echo "🔄 Downloading model (~400MB, this may take 1-2 minutes)..."
+
+    # Auto-pull model with error handling
+    if ollama pull nomic-embed-text; then
+      echo "✅ Model downloaded successfully"
+      MODEL_PRESENT=true
+    else
+      echo "⚠️  Model download failed, falling back to keyword search"
+      OLLAMA_RUNNING=false
+    fi
+  fi
+fi
+
+# Use semantic search if vector DB + Ollama + model available
+if [ -f "$VECTOR_DB" ] && [ "$OLLAMA_RUNNING" = true ] && [ "$MODEL_PRESENT" = true ]; then
+  echo "Using semantic search (vector similarity)..."
+  SEARCH_RESULTS=$(bun .aiwcli/_cc-native/lib-ts/rlm/retrieval-pipeline.ts "$ARGUMENTS")
+else
+  echo "Using keyword search (weighted scoring)..."
+  SEARCH_RESULTS=$(bun .aiwcli/_cc-native/lib-ts/rlm/transcript-searcher.ts "$ARGUMENTS" --top=10)
+fi
+```
+
+If zero results, tell the user and suggest refining the query or running `/rlm:index`.
+
+### Step 3: Load Transcript Segments
+Parse search results and load top N segments (N = result count, capped at 5):
+
+For each of the top 1-5 results, load the relevant transcript segments:
+
+```bash
+bun .aiwcli/_cc-native/lib-ts/rlm/transcript-loader.ts "<source_path>" --lines=<start>-<end>
+```
+
+Use the `matching_segments` from search results to target specific line ranges. If no segments matched, load the first 200 lines.
+
+### Step 4: Parallel Analysis
+Spawn N parallel sub-agents (Task tool, subagent_type=general-purpose) where N = min(result_count, 5):
+- 1 result → 1 agent
+- 3 results → 3 agents
+- 10 results → 5 agents (capped)
+
+Each agent analyzes one segment:
+
+> Given this session transcript segment, extract all information relevant to: **{user's question}**
+>
+> Include: specific details, code snippets, decisions made, exact quotes, and file paths mentioned.
+> Cite line numbers where possible.
+>
+> Transcript:
+> {loaded segment content}
+
+### Step 5: Synthesis
+Combine findings into a coherent answer:
+1. Group findings by theme/topic
+2. Add citations: "In session {date} (project: {project}): {finding}"
+3. Note contradictions or evolution across sessions
+4. Highlight most recent/relevant information
+
+### Output Format
+```
+## Answer: {user's question}
+
+{Synthesized answer with inline citations}
+
+### Sources
+| Date | Project | Session | Relevance |
+|------|---------|---------|-----------|
+| YYYY-MM-DD | project-name | session-id | Brief match reason |
+```
+
+Always include the sources table. If findings are sparse, say so honestly rather than speculating.
+
+## Notes
+
+- **First run:** Auto-builds indexes (10-30s delay)
+- **Semantic search:** Uses vector similarity (higher quality, requires Ollama)
+- **Keyword search:** Uses weighted scoring (faster, always available)
+- **Graceful fallback:** No Ollama? Automatically uses keyword search

package/dist/templates/cc-native/.claude/commands/cc-native/rlm/index.md

@@ -1,21 +1,21 @@
----
-description: Build/refresh the RLM transcript index across all Claude Code sessions
----
-# RLM Index — Build Transcript Index
-
-Run the TranscriptIndexer to scan all Claude Code session transcripts and build searchable indexes.
-
-## Steps
-
-1. Run the indexer:
-```bash
-bun .aiwcli/_cc-native/lib-ts/rlm/transcript-indexer.ts --batch $ARGUMENTS
-```
-
-The indexer accepts optional flags:
-- `--limit=N` — Index only the first N unindexed sessions
-- `--project=<name>` — Only index sessions for projects matching this name (partial match)
-
-2. Report the results from the JSON output (indexed count, skipped, errors, total).
-
-3. If this is the first run, suggest: "Index is ready. Try `/rlm:search <query>` or `/rlm:recall <question>` to search your session history."
+---
+description: Build/refresh the RLM transcript index across all Claude Code sessions
+---
+# RLM Index — Build Transcript Index
+
+Run the TranscriptIndexer to scan all Claude Code session transcripts and build searchable indexes.
+
+## Steps
+
+1. Run the indexer:
+```bash
+bun .aiwcli/_cc-native/lib-ts/rlm/transcript-indexer.ts --batch $ARGUMENTS
+```
+
+The indexer accepts optional flags:
+- `--limit=N` — Index only the first N unindexed sessions
+- `--project=<name>` — Only index sessions for projects matching this name (partial match)
+
+2. Report the results from the JSON output (indexed count, skipped, errors, total).
+
+3. If this is the first run, suggest: "Index is ready. Try `/rlm:search <query>` or `/rlm:recall <question>` to search your session history."

package/dist/templates/cc-native/.claude/commands/cc-native/rlm/overview.md

@@ -1,56 +1,56 @@
----
-description: Get a timeline overview of sessions by topic or time range
----
-# RLM Overview — Timeline Summary of Past Work
-
-Get a grouped timeline summary of sessions matching a topic or time range.
-
-## Input
-
-A topic, project name, or time range. Examples:
-- "hook development this week"
-- "bridge-practice-app"
-- "last 3 days"
-
-## Workflow
-
-### Step 1: Search
-Find matching sessions:
-```bash
-bun .aiwcli/_cc-native/lib-ts/rlm/transcript-searcher.ts "$ARGUMENTS" --top=20
-```
-
-For time-range queries ("this week", "last 3 days"), search with broad terms and filter by date in output.
-
-### Step 2: Group
-Group results by:
-- **Date** (if sessions span multiple days)
-- **Project** (if sessions span multiple projects)
-- **Theme** (cluster by keywords/summary similarity)
-
-### Step 3: Summarize per Group
-For each group of 3+ sessions, load excerpts from 2-3 representative sessions:
-```bash
-bun .aiwcli/_cc-native/lib-ts/rlm/transcript-loader.ts "<source_path>" --lines=1-100
-```
-
-Spawn parallel sub-agents (Task tool, subagent_type=general-purpose, model=haiku):
-
-> Summarize these session excerpts in 2-3 sentences. Focus on: what was built, what decisions were made, what problems were solved.
-
-### Step 4: Output
-Present a timeline-style summary:
-
-```
-## Session Overview: {topic/range}
-
-### {Date or Group 1}
-- {2-3 sentence summary}
-- Sessions: {count} | Project: {name}
-
-### {Date or Group 2}
-- {2-3 sentence summary}
-- Sessions: {count} | Project: {name}
-
-**Total: {N} sessions across {M} projects**
-```
+---
+description: Get a timeline overview of sessions by topic or time range
+---
+# RLM Overview — Timeline Summary of Past Work
+
+Get a grouped timeline summary of sessions matching a topic or time range.
+
+## Input
+
+A topic, project name, or time range. Examples:
+- "hook development this week"
+- "bridge-practice-app"
+- "last 3 days"
+
+## Workflow
+
+### Step 1: Search
+Find matching sessions:
+```bash
+bun .aiwcli/_cc-native/lib-ts/rlm/transcript-searcher.ts "$ARGUMENTS" --top=20
+```
+
+For time-range queries ("this week", "last 3 days"), search with broad terms and filter by date in output.
+
+### Step 2: Group
+Group results by:
+- **Date** (if sessions span multiple days)
+- **Project** (if sessions span multiple projects)
+- **Theme** (cluster by keywords/summary similarity)
+
+### Step 3: Summarize per Group
+For each group of 3+ sessions, load excerpts from 2-3 representative sessions:
+```bash
+bun .aiwcli/_cc-native/lib-ts/rlm/transcript-loader.ts "<source_path>" --lines=1-100
+```
+
+Spawn parallel sub-agents (Task tool, subagent_type=general-purpose, model=haiku):
+
+> Summarize these session excerpts in 2-3 sentences. Focus on: what was built, what decisions were made, what problems were solved.
+
+### Step 4: Output
+Present a timeline-style summary:
+
+```
+## Session Overview: {topic/range}
+
+### {Date or Group 1}
+- {2-3 sentence summary}
+- Sessions: {count} | Project: {name}
+
+### {Date or Group 2}
+- {2-3 sentence summary}
+- Sessions: {count} | Project: {name}
+
+**Total: {N} sessions across {M} projects**
+```

package/dist/templates/cc-native/.claude/commands/cc-native/specdev.md

@@ -1,10 +1,10 @@
----
-description: Interview user about a spec file and develop detailed requirements
----
-# Spec Development
-
-First, ask the user for the path to the spec file they want to develop.
-
-Then load and execute `_cc-native/workflows/specdev.md`.
-
-Follow the Clarify → Explore → Write pattern using native tools.
+---
+description: Interview user about a spec file and develop detailed requirements
+---
+# Spec Development
+
+First, ask the user for the path to the spec file they want to develop.
+
+Then load and execute `_cc-native/workflows/specdev.md`.
+
+Follow the Clarify → Explore → Write pattern using native tools.

package/dist/templates/cc-native/.windsurf/workflows/cc-native/fix.md

@@ -1,8 +1,8 @@
----
-description: Fix an issue with clarification, exploration, and planning
----
-# Fix Issue
-
-Load and execute `_cc-native/workflows/fix.md`.
-
-Follow the Clarify → Explore → Plan → Execute pattern.
+---
+description: Fix an issue with clarification, exploration, and planning
+---
+# Fix Issue
+
+Load and execute `_cc-native/workflows/fix.md`.
+
+Follow the Clarify → Explore → Plan → Execute pattern.

package/dist/templates/cc-native/.windsurf/workflows/cc-native/implement.md

@@ -1,8 +1,8 @@
----
-description: Implement a new feature with discovery and planning
----
-# Implement Feature
-
-Load and execute `_cc-native/workflows/implement.md`.
-
-Follow the Clarify → Explore → Plan → Execute pattern.
+---
+description: Implement a new feature with discovery and planning
+---
+# Implement Feature
+
+Load and execute `_cc-native/workflows/implement.md`.
+
+Follow the Clarify → Explore → Plan → Execute pattern.

package/dist/templates/cc-native/.windsurf/workflows/cc-native/research.md

@@ -1,8 +1,8 @@
----
-description: Research a topic and persist findings
----
-# Research
-
-Load and execute `_cc-native/workflows/research.md`.
-
-Follow the Clarify → Explore → Write pattern.
+---
+description: Research a topic and persist findings
+---
+# Research
+
+Load and execute `_cc-native/workflows/research.md`.
+
+Follow the Clarify → Explore → Write pattern.