Info Processor

Intelligent information intake, classification, and routing system for knowledge management.

What This Does

Takes incoming information (documents, research, data, insights) and intelligently routes it to the appropriate storage location for future retrieval and use.

Information Flow

code

Input (any format)
    ↓
Classify (determine type & purpose)
    ↓
Route to appropriate storage
    ↓
Index for retrieval
    ↓
Confirm storage location

Classification System

1. Operational Data

Purpose: Active transaction/entity data for portfolio operations Storage: Mech Storage (PostgreSQL collections) Examples:

•Entity profiles (companies, subsidiaries)
•Transactions (bank, payment processor)
•Health scores, reports
•Vendor rules

Routing:

typescript

// Create via brain repository
await repo.createEntity({...})
await repo.createTransaction({...})

2. Strategic Knowledge

Purpose: Long-term reference, business wisdom, learned insights Storage: memory/long-term/ Examples:

•Capital allocation principles
•Portfolio management strategies
•M&A criteria
•Financial analysis frameworks
•Industry insights

Routing:

code

memory/long-term/
├── capital-allocation.md
├── portfolio-strategy.md
├── ma-criteria.md
└── financial-frameworks.md

3. Reference Library

Purpose: External resources, research, documentation Storage: library/ Examples:

•Business books (Berkshire letters already there)
•Industry reports
•Market research
•Competitor analysis
•Technical documentation

Routing:

code

library/
├── business-wisdom/
│   └── berkshire-hathaway/
├── industry-reports/
├── market-research/
└── technical-docs/

4. Session Context

Purpose: Current session activities, today's learnings Storage: memory/daily/YYYY-MM-DD.md Examples:

•Implementation progress
•Decisions made today
•Bugs fixed
•New patterns learned

Routing: Append to daily log with timestamp and context

5. Skills & Capabilities

Purpose: Permanent learned capabilities Storage: .gemini/skills/<skill-name>/ Examples:

•New API integrations
•Analysis frameworks
•Automation workflows
•Decision-making processes

Routing: Create new skill directory with SKILL.md

6. Configuration & Secrets

Purpose: Credentials, API keys, environment config Storage: Mech Vault (encrypted) Examples:

•API keys (Mercury, Stripe, etc.)
•SSH keys
•Environment variables
•Deployment secrets

Routing:

typescript

await mech.createSecret({
  namespace: 'decisive/production',
  name: 'MERCURY_API_KEY',
  value: 'xxx'
})

Usage

Process Document

code

/info-processor "process this PDF about capital allocation strategies"

I will:

•Read/extract content from document
•Classify: Strategic Knowledge (capital allocation)
•Route to: memory/long-term/capital-allocation.md
•Index key concepts
•Confirm storage location

Process Research

code

/info-processor "research SaaS metrics and store for future use"

I will:

•Research SaaS metrics (MRR, CAC, LTV, etc.)
•Classify: Strategic Knowledge (financial frameworks)
•Route to: memory/long-term/saas-metrics.md
•Create structured reference
•Confirm storage location

Process Data

code

/info-processor "here's a CSV of transactions from our new entity"

I will:

•Parse CSV
•Classify: Operational Data (transactions)
•Route to: Mech Storage via repo.bulkCreateTransactions()
•Deduplicate, validate
•Confirm import statistics

Process Insight

code

/info-processor "I learned that AMZN charges often include AWS - categorize as mixed"

I will:

•Extract pattern: AMZN → mixed (AWS + retail)
•Classify: Operational Data (vendor rule)
•Route to: Mech Storage as VendorRule
•Update categorization logic
•Confirm rule created

Decision Tree

code

Is this operational data (entities, transactions, etc)?
├─ Yes → Mech Storage (via Repository)
└─ No ↓

Is this strategic knowledge (principles, frameworks)?
├─ Yes → memory/long-term/
└─ No ↓

Is this external reference (books, reports, docs)?
├─ Yes → library/
└─ No ↓

Is this session-specific (today's work)?
├─ Yes → memory/daily/
└─ No ↓

Is this a new capability (API, framework, process)?
├─ Yes → .gemini/skills/
└─ No ↓

Is this sensitive (credentials, keys)?
├─ Yes → Mech Vault
└─ No → Ask for clarification

Retrieval System

By Type

Operational Data:

typescript

// Via Repository
const entities = await repo.listEntities()
const transactions = await repo.listTransactions({ entityId })

Strategic Knowledge:

bash

# Full-text search
grep -r "capital allocation" memory/long-term/

Reference Library:

bash

# Use ask-buffett for Berkshire letters
/ask-buffett "capital allocation principles"

# Search other library content
find library/ -type f -name "*.md" -exec grep -l "SaaS metrics" {} \;

Session Context:

bash

# Today's work
cat memory/daily/$(date +%Y-%m-%d).md

# Historical
ls memory/daily/ | head -10

Skills:

bash

# List all skills
ls .gemini/skills/

# Search skill content
grep -r "API integration" .gemini/skills/

By Semantic Search (Future)

Once vector search is implemented:

typescript

// Find related knowledge
const results = await mech.vectorSearch({
  query: "How should I think about capital allocation?",
  collection: "memories",
  limit: 10
})

Auto-Classification Hints

I look for these patterns to classify:

Operational Data:

•Transaction data, bank statements
•Entity profiles, company details
•Health metrics, financial data
•Vendor patterns, categorization rules

Strategic Knowledge:

•"Principles of...", "Framework for..."
•"How to think about...", "Approach to..."
•Investment criteria, decision frameworks
•Portfolio management strategies

Reference Library:

•Books, papers, reports
•External documentation
•Industry research, market analysis
•Competitor information

Session Context:

•"Today I learned...", "Fixed bug..."
•Implementation progress
•Temporary notes, TODOs

Skills:

•"How to integrate with..."
•New capability, automation
•Reusable process, framework

Secrets:

•API keys, passwords, tokens
•SSH keys, certificates
•Environment variables

Storage Locations Summary

code

decisive_redux/
├── brain/lib/db/              # Code for accessing operational data
├── memory/
│   ├── MEMORY.md              # Core project context
│   ├── daily/                 # Session logs
│   └── long-term/             # Strategic knowledge
├── library/                   # External references
├── .gemini/skills/            # Learned capabilities
└── [Mech Storage]             # Operational data (remote)
    └── [Mech Vault]           # Secrets (remote)

Integration with Brain

The brain automatically stores certain data:

Every Heartbeat:

•Logs to memories collection (heartbeat_log type)
•Transaction imports to transactions collection
•Vendor rules to vendor_rules collection
•Health scores to health_scores collection

On Demand (via this skill):

•Strategic insights to memory/long-term/
•Research findings to library/
•New capabilities to .gemini/skills/
•Configuration to Mech Vault

Examples

Example 1: Process Business Document

Input: "Process this Berkshire 2023 letter and extract capital allocation insights"

Actions:

•Read letter content
•Extract key insights about capital allocation
•Classify: Strategic Knowledge
•Route to: memory/long-term/capital-allocation-berkshire.md
•Index key quotes and principles
•Cross-reference with existing capital allocation knowledge

Output:

code

✓ Processed Berkshire 2023 letter
✓ Extracted 12 capital allocation insights
✓ Stored: memory/long-term/capital-allocation-berkshire.md
✓ Key themes: share buybacks, acquisitions, cash deployment

Example 2: Process Transaction Data

Input: "Here's a Mercury CSV with 500 transactions from Q4 2025"

Actions:

•Parse CSV format
•Map to Transaction schema
•Classify: Operational Data
•Route to: Mech Storage via bulkCreateTransactions()
•Deduplicate against existing transactions
•Run AI categorization on uncategorized

Output:

code

✓ Parsed 500 transactions from CSV
✓ Imported 487 new transactions (13 duplicates)
✓ Stored in Mech Storage: transactions collection
✓ Categorized 487 transactions (12 new rules learned)

Example 3: Process Research

Input: "Research best practices for SaaS metrics and store them"

Actions:

•Research SaaS metrics (MRR, ARR, CAC, LTV, churn, etc.)
•Compile frameworks and benchmarks
•Classify: Strategic Knowledge
•Route to: memory/long-term/saas-metrics.md
•Structure with formulas and thresholds

Output:

code

✓ Researched SaaS metrics best practices
✓ Compiled 8 key metrics with formulas
✓ Stored: memory/long-term/saas-metrics.md
✓ Includes: MRR, CAC, LTV, churn, NRR, expansion

Example 4: Process Insight

Input: "I noticed AWS charges from Amazon should be categorized as Infrastructure, not Shopping"

Actions:

•Extract pattern: "AWS" in description → Infrastructure
•Classify: Operational Data (vendor rule)
•Route to: Mech Storage via createVendorRule()
•Update existing Amazon rules or create new one
•Apply retroactively to past transactions

Output:

code

✓ Created vendor rule: AWS → Infrastructure
✓ Pattern: /AWS|Amazon Web Services/i
✓ Confidence: 1.0 (manual rule)
✓ Applied retroactively: 23 transactions updated

Keywords for Discovery

information management, knowledge management, data routing, document processing, information architecture, knowledge base, content organization, data classification, information intake, smart filing, knowledge storage

Related Skills

•/memory-manager - Manage persistent memory
•/knowledgebase - Store and query knowledge
•/library-search - Search reference library
•/ask-buffett - Query Berkshire letters

This skill makes me smarter over time. Every document processed, every insight captured, every data point stored makes me more effective at managing the portfolio.