📚 Knowledgebase Types

Five Types of Knowledge.
One Retrieval Platform.

Company knowledge does not come in one shape. It lives in PDFs, spreadsheets, CAD files, past proposals, and compliance archives. RenderDraw's five knowledgebase types are each tuned for a different data shape — so retrieval is precise regardless of where the knowledge lives.

Type 01

Document Knowledgebases

PDFs, Word, PowerPoint, HTML, Plain Text

Document KBs are the most common knowledgebase type. They ingest prose-heavy files — technical specifications, product manuals, standards documents, service bulletins, and engineering guides — and make them retrievable by semantic meaning, not just keyword match.

When a workflow AI block queries a Document KB, it sends the current task context (e.g. "what are the tensile strength requirements for Grade 5 bolts in this assembly?") as a natural language embedding query. The KB returns the most semantically similar paragraphs — not just pages containing the word "tensile" — enabling the AI to generate precise, citation-grounded answers.

Document KBs support OCR for scanned PDFs, structured heading extraction for section-level retrieval, and table detection for simple tabular data embedded in prose documents. For predominantly tabular files (Excel pricing sheets, CSV catalogs), use a Structured Data KB instead.

When to Use Document KBs

  • Technical specification libraries and engineering standards
  • Product manuals, service bulletins, and installation guides
  • RFP requirements documents and scope-of-work PDFs
  • Safety data sheets (SDS) and environmental compliance documents
  • Legal agreements, contract templates, and terms of service

Setup Tips

  • Enable OCR for any scanned PDFs — text-layer documents process significantly faster
  • Use section-level splitting for long manuals with clear heading structure
  • Set chunk size to 512–768 tokens for dense technical content
  • Enable heading metadata extraction so retrieved chunks include their section context
  • Tag documents with product line, revision date, and document type for faceted filtering
📄

Supported Formats

  • PDF (text-layer + OCR for scanned)
  • Word (.docx, .doc)
  • PowerPoint (.pptx, .ppt)
  • HTML and web pages
  • Markdown (.md)
  • Plain text (.txt)
  • XML (with schema mapping)

Ideal Chunk Settings

  • Technical specs: 512 tokens, paragraph split
  • Long manuals: 512 tokens, section split, 20% overlap
  • Proposals: 1,024 tokens, section split
  • Standards docs: 768 tokens, 15% overlap

Type 02

Structured Data Knowledgebases

Excel, CSV, Database Tables, ERP Exports

Structured Data KBs treat tabular data as first-class knowledge. Instead of flattening rows and columns into prose chunks (which destroys the relational context that makes tabular data useful), RenderDraw preserves row-column structure during ingestion and supports both semantic and SQL-style querying at runtime.

A pricing table KB, for example, lets an AI block query: "What is the unit price for a 2-inch Schedule 40 carbon steel elbow in quantity 500 with next-day delivery?" The KB returns the exact row matching those parameters — not a paraphrased estimate drawn from general training data. This is the difference between a workflow that prices quotes correctly and one that hallucinates plausible-looking numbers.

Structured Data KBs support schema definition (you specify column names, data types, and units), column mapping (rename ambiguous headers to semantically meaningful labels), and live sync with ERP and CRM systems via the datasource integration layer. See the Structured Data guide for the full technical reference.

When to Use Structured Data KBs

  • Pricing tables and discount tiers
  • Bills of materials (BOMs) and component catalogs
  • Supplier catalogs and lead time tables
  • Materials specifications with dimensional data
  • Product configuration matrices and option compatibility tables
📊

Supported Sources

  • Excel (.xlsx, .xls)
  • CSV and TSV files
  • PostgreSQL / SQL Server tables (live sync)
  • Salesforce objects (live sync)
  • ERP exports (SAP, Oracle, Infor)
  • Google Sheets (live sync)

Query Modes

  • Semantic: Natural language lookup — "find the lowest price for stainless fittings under 2 inches"
  • Exact match: Row lookup by primary key or unique identifier
  • Range filter: "All items with quantity breaks > 100 and unit price < $5"

Type 03

CAD Asset Knowledgebases

3D Models, Assembly Files, Component Libraries

CAD Asset KBs are unique to RenderDraw. The platform's native CAD pipeline — built on HOOPS Exchange and processing over 30 major CAD formats — extracts structured knowledge from 3D models: part metadata, assembly hierarchies, dimensional tolerances, material assignments, surface areas, volumes, and BOM relationships.

This extracted metadata becomes a queryable knowledgebase. When a takeoff workflow needs to determine the surface area of a component for a coating estimate, or an RFP workflow needs to confirm that an assembly meets a specific dimensional requirement, the CAD Asset KB provides the answer from the actual model — not from a PDF approximation that may be out of sync with the current design revision.

CAD Asset KBs update automatically when new model revisions are processed. The KB version history tracks every model version, so you can always trace a workflow output back to the exact CAD revision it used.

When to Use CAD Asset KBs

  • Takeoff workflows that need dimensional data from models
  • RFP responses referencing specific product configurations
  • Spare parts catalogs with accurate part geometry
  • Assembly compliance checks against dimensional specifications
  • Cost estimation workflows using surface area or volume data
🏰

Extracted Metadata

  • Part names, numbers, and descriptions
  • Assembly hierarchy and parent-child relationships
  • Dimensions, tolerances, and surface areas
  • Material assignments and mass properties
  • BOM structure and quantity rollups
  • View representations and configurations

Supported CAD Formats

STEP, IGES, SolidWorks, Inventor, CATIA, PTC Creo, Siemens NX, Parasolid, JT, OBJ, STL, and 25+ additional formats via HOOPS Exchange.


Type 04

Response Template Knowledgebases

Winning Bids, Proposal Libraries, Executive Summaries

Response Template KBs store your organization's best-performing written outputs: winning RFP responses, executive summaries, proposal introductions, qualification statements, project narratives, and cover letters. The AI learns not just what to say, but how your organization says it — your industry vocabulary, your proposal structure, your value proposition framing.

When a workflow AI block is generating a new response section, the Response Template KB retrieves the most relevant examples from your historical library. These examples ground the AI in your actual voice and proven patterns, rather than generating from scratch in a generic style that doesn't represent your brand or align with evaluator expectations from previous submissions.

Template KBs benefit enormously from rich metadata tagging. Tag each template with: industry vertical, project type, project size, client type, win/loss outcome, and year. This metadata drives faceted filtering — a workflow responding to a $15M industrial HVAC RFP can specifically retrieve winning proposals from similar projects, not generic construction examples.

When to Use Response Template KBs

  • RFP response workflows needing voice and structure consistency
  • Qualification statement generation for new verticals
  • Executive summary sections for large proposals
  • RFI response libraries for recurring question types
  • Cover letter and introduction generation
🏆

Recommended Tags

  • Industry vertical (industrial, commercial, government)
  • Project type (HVAC, structural, electrical)
  • Project size ($1M–$5M, $5M–$25M, $25M+)
  • Outcome (win, loss, no-bid)
  • Client type (owner, GC, subcontractor)
  • Year and region

Retrieval Strategy

For template KBs, increase topK to 5–8 to surface multiple strong examples per section. The AI uses retrieved examples as structural anchors and voice calibration, not as direct copy — each output is original and tailored to the current document context.


Type 05

Compliance Knowledgebases

Regulations, Standards, Certifications, Internal Policies

Compliance KBs store the regulatory and standards documents that workflow outputs must be checked against. Building codes, industry standards (ASHRAE, ASTM, ISO, OSHA), environmental regulations, internal quality procedures, and certification requirements all belong here.

Compliance KBs are typically attached to workflow blocks that perform checklist validation: "Does this response address all mandatory submission requirements?" or "Does this scope of work include the required safety disclosures?" The AI block retrieves the relevant standard, cross-references it against the current document, and flags gaps — before a human reviewer sees the output.

Compliance KBs require the strictest versioning discipline of all KB types. When a standard is updated — a new code cycle, a regulation amendment, a revised internal policy — the KB must be updated before workflows run. RenderDraw supports notification rules: when a compliance KB document is updated, all workflows using that KB receive an alert that a review of their context rules is recommended.

When to Use Compliance KBs

  • RFP response section compliance checking
  • Takeoff specifications validated against building codes
  • Subcontractor qualification verification against certification requirements
  • Environmental impact section review against applicable regulations
  • Internal quality gate checks before document submission

Common Compliance Sources

  • Building codes (IBC, IFC, NFPA, local amendments)
  • ASHRAE standards (90.1, 62.1, 170)
  • ASTM and ANSI standards
  • ISO 9001 / ISO 14001 procedures
  • OSHA safety requirements
  • Agency-specific submission requirements (DOT, EPA, FAA)
  • Internal quality management procedures

⚠ Critical: Version Currency

A compliance KB citing a superseded standard is worse than no compliance KB — it gives false confidence. Set quarterly re-sync schedules and configure version-change notifications. Audit KB document dates before every major submission cycle.

Which Type Do I Need?

Most organizations end up using all five types. Here's a quick decision guide to get started.

My content is primarily prose paragraphs

Technical manuals, spec sheets, standards documents, RFP requirements → Document KB

My content is rows and columns of data

Pricing tables, BOMs, supplier catalogs, compatibility matrices → Structured Data KB

My content is 3D model files

Product assemblies, component libraries, CAD archives → CAD Asset KB

My content is past proposals and responses

Winning bids, executive summaries, qualification statements → Response Template KB

My content is regulations and standards

Building codes, industry standards, internal policies → Compliance KB

Set Up Your First KB →