Context Profiles

Context profiles tune ClawVault's retrieval behavior for specific scenarios. Instead of one-size-fits-all context, profiles optimize what memories surface based on your current task.

Available Profiles

default

Balanced retrieval for general work.

• Equal weight to semantic relevance and graph connections
• Moderate temporal decay (recent items favored but not dominant)
• Mixed memory types based on query content

Best for: Regular development work, research, general questions

clawvault context "API integration" --profile default

planning

Strategic context for roadmapping and architecture.

• Emphasizes decisions, commitments, and projects
• Longer temporal window (includes older strategic context)
• Prioritizes high-level connections over implementation details
• Includes related lessons for informed planning

Best for: Sprint planning, architecture reviews, project kickoffs

clawvault context "Q1 roadmap" --profile planning

incident

Crisis mode - focus on recent problems and blockers.

• Heavy emphasis on recent memories (last 48 hours)
• Prioritizes errors, blockers, and critical decisions
• Includes related people (who to contact) and facts (system details)
• Minimal historical context to reduce noise

Best for: Production issues, debugging, urgent problems

clawvault context "API timeouts production" --profile incident

handoff

Session transition context for continuity.

• Focus on recent handoffs, commitments, and active projects
• Includes feelings and subjective context for smooth transitions
• Emphasizes next steps and blockers from recent sessions
• Graph connections to related ongoing work

Best for: Session starts, context recovery, team handoffs

clawvault context "current work status" --profile handoff

Auto Profile Detection

Use --profile auto to let ClawVault infer the best profile from your query:

clawvault context "urgent: database crashed" --profile auto
# → Detected: incident (keywords: urgent, crashed)

clawvault context "planning next quarter goals" --profile auto 
# → Detected: planning (keywords: planning, quarter, goals)

clawvault context "what was I working on yesterday" --profile auto
# → Detected: handoff (keywords: working on, yesterday)

clawvault context "how to implement OAuth" --profile auto
# → Detected: default (general technical query)

Detection rules:

• incident: urgent, error, failed, broken, down, crash, bug
• planning: roadmap, quarter, plan, strategy, architecture, goals
• handoff: working on, yesterday, last session, status, progress
• default: everything else

Profile Behavior Details

Temporal Weighting

How each profile weighs recency:

Memory Type Preferences

What types each profile emphasizes:

incident

• High: errors, blockers, recent decisions
• Medium: people (contacts), facts (system info)
• Low: historical lessons, old projects

planning

• High: decisions, commitments, projects
• Medium: lessons, preferences, facts
• Low: feelings, daily observations

handoff

• High: recent handoffs, commitments, projects
• Medium: feelings, blockers, next steps
• Low: old decisions, historical lessons

default

• Balanced across all types based on semantic relevance

Graph Traversal Depth

How far each profile explores connections:

• incident: 1 hop (direct connections only)
• handoff: 2 hops (close relationships)
• default: 2 hops (balanced exploration)
• planning: 3 hops (broader context)

OpenClaw Integration

The OpenClaw hook automatically uses context profiles:

// In session:start event
const profile = inferContextProfile(sessionQuery);
const context = await vault.context(query, { profile: 'auto' });

This ensures new sessions get appropriately-scoped context without manual configuration.

Usage Examples

Morning Session Start

clawvault context "what needs attention today" --profile handoff

Returns: recent handoffs, pending commitments, blockers

Code Review Preparation

clawvault context "authentication refactor decisions" --profile default

Returns: related technical decisions, implementation notes, connected discussions

Production Incident

clawvault context "payment processing down users complaining" --profile incident

Returns: recent errors, system facts, escalation contacts, similar past incidents

Sprint Planning Meeting

clawvault context "user dashboard roadmap Q1 priorities" --profile planning 

Returns: strategic decisions, project commitments, lessons from past sprints, stakeholder preferences

JSON Output with Explain

Get detailed context signals with JSON format:

clawvault context "database migration" --profile planning --format json

{
 "query": "database migration",
 "profile": "planning",
 "entries": [
 {
 "id": "decisions/database-choice",
 "title": "Use PostgreSQL over SQLite",
 "excerpt": "Need concurrent writes...",
 "explain": {
 "semanticScore": 0.87,
 "graphScore": 0.65,
 "temporalScore": 0.45,
 "signals": ["direct_match", "connected_project", "recent_decision"],
 "rationale": "High semantic relevance + connected to current migration project"
 }
 }
 ]
}

This helps debug why certain memories surfaced and tune queries.

Customization

Profiles are currently built-in, but you can approximate custom behavior:

# Incident-like behavior manually
clawvault context "query" --recent-days 2 --types "errors,blockers"

# Planning-like behavior
clawvault context "query" --graph-depth 3 --types "decisions,projects"

Context profiles make ClawVault's retrieval contextually intelligent, surfacing the right memories for your current situation.