/recall - Intelligent Memory Retrieval

Retrieve relevant prior work context from the CODITECT memory system. Optimizes for signal-to-noise ratio and token efficiency, surfacing decisions, patterns, errors, and work status related to your current task.

Usage

# Basic topic retrieval
/recall "authentication"
/recall "database migration"

# Quick work status
/recall --status

# Deep context retrieval
/recall --deep "API design"

# Focused retrieval
/recall --decisions "technology choice"
/recall --patterns --language python
/recall --errors "TypeError"
/recall --blockers

# Output formats
/recall "topic" --json
/recall "topic" --detailed

# Token budget control
/recall "topic" --budget 500    # Minimal
/recall "topic" --budget 5000   # Comprehensive

System Prompt

System Prompt

⚠️ EXECUTION DIRECTIVE: When the user invokes this command, you MUST:

1. IMMEDIATELY execute - no questions, no explanations first
2. ALWAYS show full output from script/tool execution
3. ALWAYS provide summary after execution completes

DO NOT:

• Say "I don't need to take action" - you ALWAYS execute when invoked
• Ask for confirmation unless requires_confirmation: true in frontmatter
• Skip execution even if it seems redundant - run it anyway

The user invoking the command IS the confirmation.

---

You are retrieving relevant context from the CODITECT long-term memory system.

Script: scripts/memory-retrieval.py Databases (ADR-118):

• context-storage/org.db (Tier 2: decisions, skill_learnings, error_solutions)
• context-storage/sessions.db (Tier 3: messages, token_economics)

Execute the memory retrieval:

python3 scripts/memory-retrieval.py $ARGS

First-time setup: If database doesn't exist, run /cx first to build the memory system.

Quick Reference

Basic Operations

Command	Description
/recall "topic"	Retrieve context for topic (2000 token budget)
/recall --status	Current work status summary
/recall --deep "topic"	Comprehensive retrieval (5000 tokens)
/recall --help	Show help

Focused Retrieval

Command	Description
/recall --decisions	Recent decision summary
/recall --decisions "topic"	Decisions related to topic
/recall --patterns	Code patterns summary
/recall --patterns --language rust	Rust patterns only
/recall --errors	Error-solution pairs
/recall --errors "TypeError"	Specific error type
/recall --blockers	Active blockers and issues
/recall --wip	Work-in-progress items

Output Control

Command	Description
/recall "topic" --json	JSON output for programmatic use
/recall "topic" --detailed	Full context with code snippets
/recall "topic" --budget N	Control token budget

Examples

Starting a New Task

# Get context before starting work
/recall "user authentication"

# Output:
## Prior Context Summary (1847 tokens)

**Work Status:**
- [IN_PROGRESS] Auth service refactor - 60% (2 days ago)
- [COMPLETED] JWT implementation (last week)

**Relevant Decisions:**
1. JWT with refresh tokens - 24h access, 7d refresh (ADR-005)
2. bcrypt for password hashing - cost factor 12

**Applicable Patterns:**
- auth_middleware pattern (Python)
- token_refresh_handler pattern

**Known Issues:**
- Token expiry race condition - fixed with 5min grace period

Checking Work Status

/recall --status

# Output:
## Work Status Summary

**In Progress (3):**
- Database migration v3 - 70% complete
- API rate limiting - blocked on Redis setup
- User dashboard - design phase

**Blocked (1):**
- Payment integration - awaiting Stripe API keys

**Recently Completed (5):**
- User model refactor (Dec 10)
- Test coverage improvement (Dec 9)
- ...

Finding Relevant Decisions

/recall --decisions "database"

# Output:
## Database Decisions

1. **PostgreSQL over MongoDB** (ADR-003)
   - Rationale: ACID compliance, complex queries
   - Date: 2025-11-15
   - Relevance: 0.94

2. **Connection pooling with PgBouncer**
   - Rationale: Handle 1000+ concurrent connections
   - Date: 2025-11-20
   - Relevance: 0.87

3. **Schema versioning with Alembic**
   - Rationale: Reproducible migrations
   - Date: 2025-11-18
   - Relevance: 0.82

Finding Code Patterns

/recall --patterns --language python

# Output:
## Python Patterns

1. **async_handler** - Async request handling with error recovery
   ```python
   async def handler(request):
       try:
           result = await process(request)
           return Response(result)
       except Exception as e:
           logger.error(f"Handler error: {e}")
           return ErrorResponse(500)

2. repository_base - Data access abstraction

   class Repository(Generic[T]):
       async def get(self, id: str) -> T: ...
       async def save(self, entity: T) -> None: ...

### Finding Error Solutions
```bash
/recall --errors "connection pool"

# Output:
## Error Solutions: Connection Pool

1. **ConnectionPoolExhausted**
   - Error: `pool exhausted, no connections available`
   - Solution: Increase max_connections to 20, add connection timeout
   - Code:
     ```python
     pool = await asyncpg.create_pool(
         max_size=20,
         command_timeout=30
     )
     ```
   - Success rate: 100% (3 occurrences)

Deep Context for Complex Task

/recall --deep "implement payment system"

# Output:
## Comprehensive Context Report (4823 tokens)

### Work Status
- [BLOCKED] Payment integration - awaiting Stripe keys
- [COMPLETED] Pricing model design
- [PLANNED] Subscription management

### Decision History
1. **Stripe over PayPal** (ADR-007)
   - Rationale: Better API, lower fees, webhook support
   - Alternatives considered: PayPal, Square, Braintree
   - Decision date: 2025-11-25

### Architecture Notes
- Payment service isolated in `services/payment/`
- Event-driven with webhooks
- Idempotency keys for all mutations

### Code Patterns
[Full implementations...]

### Related Errors
[Previous payment-related issues...]

### Expand Available
- /recall --expand decisions
- /recall --expand patterns
- /recall --expand errors

JSON Output for Automation

/recall "authentication" --json

# Output:
{
  "retrieval_id": "mem_20251212_143022",
  "topic": "authentication",
  "token_count": 1847,
  "work_status": [...],
  "decisions": [...],
  "patterns": [...],
  "errors": [...],
  "relevance_summary": {
    "high": 3,
    "medium": 5,
    "low": 2
  }
}

Options

Topic & Scope

Option	Description
"topic"	Main search topic (positional)
--deep	Comprehensive retrieval (5000 token budget)
--status	Work status summary only
--wip	Work-in-progress items only
--blockers	Blocked items only

Knowledge Types

Option	Description
--decisions	Decision records
--patterns	Code patterns
--errors	Error-solution pairs
--all	All knowledge types (default)

Filters

Option	Description
--language LANG	Filter patterns by language
--type TYPE	Filter by decision/error type
--since DATE	Only items since date
--project NAME	Filter by project

Output

Option	Description
--json	JSON output
--detailed	Full content, no summarization
--budget N	Token budget (default: 2000)

Information

Option	Description
--stats	Memory system statistics
--help	Show help

Token Budgets

Budget	Tokens	Use Case
minimal	500	Quick status check
standard	2000	Normal session start (default)
comprehensive	5000	Complex task planning
full	10000	Complete project history

Integration

With Orchestrators

Orchestrators automatically call /recall during planning phase to ensure plans account for prior work.

With Session Hooks

The pre-session-memory hook can auto-run /recall at session start based on detected context.

With /cxq

/recall builds on /cxq but adds:

• Intelligent relevance scoring
• Token-efficient summarization
• Work status awareness
• Hierarchical output

Workflow

Session Start

# Check what's relevant to today's work
/recall "today's task topic"

Before Planning

# Ensure you know prior decisions
/recall --decisions "feature area"

When Stuck

# Find if similar issues were solved
/recall --errors "error message"

End of Day

# Update memory for next session
/export && /cx

Success Output

When memory retrieval completes:

✅ COMMAND COMPLETE: /recall
Topic: <search-topic>
Results: N items
Token Count: X
Categories: decisions, patterns, errors, status
Relevance: High (N), Medium (M), Low (L)

Completion Checklist

Before marking complete:

• Topic identified
• Memory database queried
• Results ranked by relevance
• Token budget respected
• Summary generated

Failure Indicators

This command has FAILED if:

• ❌ Memory database not found
• ❌ No results returned
• ❌ Token budget exceeded
• ❌ Invalid query

When NOT to Use

Do NOT use when:

• Memory database not initialized (run /cx first)
• Searching for new information (use web search)
• Simple file lookup (use /where)

Anti-Patterns (Avoid)

Anti-Pattern	Problem	Solution
Too broad topic	Noise in results	Be specific
Skip --deep for complex tasks	Missing context	Use --deep for planning
Ignore token budget	Context bloat	Use appropriate --budget

Principles

This command embodies:

• #9 Based on Facts - Historical context retrieval
• #4 Keep It Simple - Token-efficient summaries

Full Standard: CODITECT-STANDARD-AUTOMATION.md

---

Agent: memory-context-agent Skill: skills/memory-retrieval/SKILL.md Script: scripts/memory-retrieval.py Version: 1.0.0 Last Updated: 2025-12-12