NLU Brain — Intelligent Routing
Natural language understanding that routes your requests to the right Bootspring capability
Bootspring's NLU Brain is a unified natural language router that understands what you're trying to do and dispatches to the right skill, workflow, agent, or pipeline — automatically.
How It Works
The Brain uses a 3-stage cascade to find the best match:
- Exact Match — Direct command name lookup (fastest, 100% confidence)
- Semantic Match — TF-IDF vectorization + cosine similarity over 200+ trained phrases
- Keyword Fallback — Weighted keyword scoring when semantic match is inconclusive
Example Usage
CLI
Loading code block...
MCP Tool
The bootspring_brain MCP tool integrates directly with AI coding assistants:
Input: "I need to check if this is ready to deploy"
→ Target: deploy-readiness (skill)
→ Confidence: 0.88
→ Command: bootspring deploy check
Supported Targets
| Type | Targets | Examples |
|---|---|---|
| Skills | 9 | build-loop, seed, quality-gates, geo-seo, ai-marketing, deploy-readiness, cost-optimizer, observer, memory |
| Workflows | 5 | code-review, test-generation, refactoring, documentation, security-scan |
| Agents | 5 | file-analyzer, dependency-mapper, code-searcher, task-summarizer, git-insight |
| Pipeline | 2 | full-pipeline, fast-analysis |
Configuration
| Setting | Default | Description |
|---|---|---|
| minConfidence | 0.3 | Below this, ask user to clarify |
| autoDispatchThreshold | 0.85 | Above this, auto-invoke the action |
| maxAlternatives | 3 | Number of alternative matches shown |
| enableLearning | true | Log corrections for self-improvement |
Self-Improvement
The Brain tracks routing accuracy and learns from corrections:
- Every routing decision is logged with input, target, and confidence
- User corrections are recorded and used to adjust future rankings
- Analytics dashboard shows accuracy rate, method breakdown, and top targets