Skip to content
Salamandr Docs

Knowledge Base & Private AI-Assist

Salamandr provides a unified Knowledge Base & AI-Assist Engine designed to keep institutional knowledge organized, accessible to customers, and searchable for agents without sending sensitive data to external AI clouds.

┌────────────────────────────────────────────────────────────────────────┐
│                   Salamandr Knowledge Ecosystem                        │
├───────────────────────────────────┬────────────────────────────────────┤
│    1. Help Center & Runbooks      │    2. Local AI & Document RAG      │
│       [Read Guide →](help-center) │       [Read Guide →](ai-assist)    │
├───────────────────────────────────┼────────────────────────────────────┤
│ • Public /help self-service       │ • On-premises Ollama integration   │
│ • Internal SOP runbooks for staff │ • Semantic PDF, Word, Excel parser │
│ • Multilingual translation groups │ • Web sources crawler              │
│ • Revision history & snapshots    │ • PostgreSQL pgvector embeddings   │
│ • Hierarchical page trees         │ • Grounded response drafts with    │
│ • Staff article comments          │   exact file citations             │
└───────────────────────────────────┴────────────────────────────────────┘

Dedicated Guides

Section titled “Dedicated Guides”

Explore the detailed technical documentation for each component:

  1. Help Center & Articles:

    • Public customer portal configuration at /help.
    • Category hierarchies, visual branding, and hex color gradients.
    • Multilingual localization with canonical translation groups.
    • Non-destructive article versioning and rollback.
    • Confluence-style hierarchical page trees.
    • Syncing articles from a connected GitHub/GitLab repository.
    • Tree view grouped by visibility & category, with inline recategorize.
    • Internal staff collaboration comments on runbooks.
    • PostgreSQL full-text search (tsvector & GIN indexes).
    • Complete REST API reference.

  2. Local AI & Document RAG:

    • 100% air-gapped local architecture using self-hosted Ollama (bge-m3 & gemma2:9b).
    • Ingestion of technical manuals, PDF/Word/Excel specifications, and internal wikis.
    • Automated web crawler for internal and partner portals.
    • Semantic chunking (512 tokens with 50-token overlap) and HNSW vector storage in PostgreSQL with pgvector.
    • AI response drafting in the ticket detail context rail with direct runbook citations.
    • Strict “Propose, don’t act” human-in-the-loop safety posture.
    • Complete REST API reference.

  3. AI Log Triage:

    • Automated first-pass triage of uploaded log bundles, grounded against your KB and issue tracker.
    • Filtered excerpt extraction, screenshot OCR, per-tenant tuning.
    • Complete REST API reference.