BrainstormMSP routes AI tasks across 5 providers, selecting the optimal model for each task type. This multi-LLM approach ensures the best reasoning, fastest response, and highest reliability for every decision.
5 AI Providers
Provider Overview
|----------|------------|-------|
Why Multiple Providers?
• **No single LLM is best at everything**: Each excels in different domains
• **Reliability**: If one provider is down, others handle the load
• **Cost optimization**: Route simple tasks to cheaper models
• **Verification**: Cross-check critical decisions across providers
Routing Logic
How Routing Works
The Signal Processor determines which provider to use based on:
1. **Task type**: Reasoning vs. search vs. embedding vs. analysis
2. **Context size**: Large documents route to Gemini's large context window
3. **Freshness**: Tasks requiring current data route to Perplexity
4. **Domain**: Social/trend analysis routes to Grok
5. **Criticality**: High-risk decisions use Claude for the strongest reasoning
Routing Rules
• **Default**: Claude handles all reasoning and decision-making
• **Embeddings**: OpenAI text-embedding-3-large (1024 dimensions)
• **Web research**: Perplexity for any task requiring live web data
• **Large context**: Gemini for documents exceeding 100K tokens
• **Social signals**: Grok for X/Twitter analysis and trend detection
Model Selection
Automatic Model Selection
The brain selects models based on task requirements:
Claude Tasks
• Brain decisions and OODA loop reasoning
• Agent action planning
• Evidence chain analysis
• Code generation for remediation scripts
OpenAI Tasks
• Text embeddings for the knowledge base (pgvector)
• Alternative reasoning for cross-verification
Gemini Tasks
• Processing large compliance documents
• Analyzing lengthy log files
• Bulk evidence analysis
Perplexity Tasks
• CVE and vulnerability research
• Vendor status checks
• Industry news monitoring
Grok Tasks
• Social media threat intelligence
• Brand reputation monitoring
• MSP industry trend analysis
Provider Strengths
Why Each Provider Was Chosen
Claude (Anthropic)
• Strongest reasoning and instruction following
• Best at multi-step planning and evidence analysis
• Primary brain reasoning engine
OpenAI
• Industry-standard embedding model
• Strong at structured data analysis
• Proven reliability for embeddings workload
Gemini (Google)
• Largest context window (1M+ tokens)
• Strong at document understanding
• Good for bulk analysis tasks
Perplexity
• Built-in web search — no separate search API needed
• Always returns sourced, current information
• Ideal for "what's happening right now" queries
Grok (xAI)
• Unique access to X/Twitter data
• Real-time social intelligence
• Unfiltered analysis of public sentiment
Fallback Behavior
Provider Failover
If a provider is unavailable, the routing system fails over:
1. **Primary provider fails**: Route to secondary provider
2. **Retry with backoff**: 3 retries with exponential backoff
3. **Fallback mapping**:
- Claude fails → OpenAI o3 for reasoning
- OpenAI fails → Claude for embeddings (dimension adjustment)
- Gemini fails → Claude with chunked context
- Perplexity fails → Grok for web research
- Grok fails → Perplexity for social analysis
AI Sentinels
The 5 AI sentinels continuously monitor provider health:
• Response time tracking
• Error rate monitoring
• Token quota tracking
• Automatic routing updates based on current health
Viewing AI Provider Status
Go to **Brain > Observatory** to see:
• Current status of all 5 providers (GREEN/YELLOW/RED)
• Response time trends
• Usage statistics per provider
• Routing decisions log
Completed!
You've completed the Multi-LLM Routing & AI Providers guide. Ready to continue learning?