Skip to content

Feature: Knowledge Base (Twitter-Style)

The Knowledge Base is a lightweight, self-ranking documentation system that replaces traditional enterprise wikis with atomic, friction-free knowledge snippets.

The Innovation: MasalaRank

Traditional knowledge management fails because good content is buried under stale documentation. Ticket Masala solves this through Usage-Based Ranking:

  • Atomic Units: Documentation is broken into "Tweets"—snippets of 50-300 words.
  • #Hashtag Organization: No folders. Content is categorized via hashtags within the text.
  • MasalaRank Algorithm: Content quality is determined by actual usage in resolved tickets: Score = (UsageCount × 10) + (IsVerified × 50) + (FTS_Rank × -1)

Business Value

The Problem: "The Wiki Trap"

Most enterprise wikis are stagnant because contribution is too difficult and search results are prioritized by date rather than relevance or quality.

The Solution: "Living Knowledge"

We treat knowledge as a stream. High-quality, verified answers rise to the top of search results based on how often they actually help close cases.

Technical Architecture

graph TD
    subgraph "Contribution"
        User[Agent] -->|Write Snippet| Editor[Markdown Editor]
        Editor -->|Extract #Tags| Parser[Tag Parser]
        Parser -->|Save| DB[(SQLite FTS5)]
    end

    subgraph "Ranking Loop"
        Resolved[Ticket Closed] -->|Link Snippet| Score[Usage++]
        Score -->|Calculate| Rank[MasalaRank]
    end

Detailed Capabilities

1. The Atomic Editor

A simplified Markdown interface that encourages agents to share "micro-learnings" immediately after resolving a complex case.

2. Verified Signals

Domain experts can "Verify" a snippet (✓), giving it a permanent +50 boost in the MasalaRank algorithm and marking it as reliable for AI context injection.

Powered by SQLite's FTS5 extension with trigram tokenization, allowing for blazing-fast, typo-tolerant search across thousands of snippets. 

-- Example MasalaRank Search Query
SELECT Content, (UsageCount * 10) + (CASE WHEN IsVerified THEN 50 ELSE 0 END) AS Score
FROM KnowledgeIndex
WHERE KnowledgeIndex MATCH @query
ORDER BY Score DESC;

Operational Scenarios

The "Anita" Problem (Quality Control)

To prevent the knowledge base from becoming "noisy," the system runs a similarity check during submission. If an agent tries to post a duplicate answer, the system warns them: "We found 3 similar snippets. Consider clarifying theirs instead of creating a new one."

AI Foundation

The Enrichment Pipeline uses these verified snippets as the primary source of truth for RAG (Retrieval-Augmented Generation). The AI doesn't just guess; it cites your best internal documentation.

Success Criteria

  1. Adoption: 80% of agents contributing at least one snippet per month.
  2. Signal: Top 3 search results are relevant in 90%+ of queries.
  3. Speed: Search latency <50ms even with 100,000+ snippets.

References