The Lucas Knowledge Base is a collection of documents you upload to ground AI-generated scripts in your organization's approved content. When Lucas generates a script, it searches the Knowledge Base first — pulling facts, terminology, and context from your materials instead of generating them from scratch.
What the Knowledge Base Does
Without a Knowledge Base, Lucas generates scripts using general AI knowledge. With one, it generates scripts using your knowledge — your product details, your terminology, your approved messaging.
This reduces hallucination risk and keeps scripts factually accurate and on-brand without requiring manual review of every word.
What to Upload
The Knowledge Base works by finding text that answers a question. Upload documents where the answer to a likely script question is written out in plain language.
| Content Type | Example | Why It Works |
|---|---|---|
| Product overviews | Feature descriptions, service explainers | Lucas asks "What does X do?" — the answer is in the text |
| FAQs | Product FAQ, onboarding FAQ | Questions and answers already in natural language |
| Policy documents | Terms of service, return policy, compliance summaries | User asks "What's the policy on X?" — answer is a paragraph |
| How-to guides | "How to file a claim," onboarding walkthroughs | Steps are written out and searchable |
What Not to Upload
Some content looks like it belongs in the Knowledge Base but won't work well — and in some cases can introduce inaccurate or misleading output. The Knowledge Base finds text that resembles a question. It cannot filter tables, evaluate conditions, or follow decision logic.
| Content Type | Example | Why It Doesn't Work |
|---|---|---|
| Pricing tables | Plan tiers with rates | Requires filtering a row, not finding meaning in text |
| Product spec sheets | Technical dimensions, field definitions, API specs | Lucas may summarize exact values incorrectly |
| Eligibility matrices | "If customer has X and Y, offer Z" | Requires logic evaluation, not semantic search |
| Data exports | CRM exports, analytics reports | Structured data — not searchable by meaning |
| Slide decks with only visuals | Decks where content is in images, not text | Lucas cannot read text embedded in images |
| Audience segmentation files | Persona definitions, segment criteria, customer profiles | Describes who to target, not what to say — Lucas can't apply this to script content |
| Compliance and regulatory documents | Legal disclaimers, regulatory requirements, audit trails | Contains rules that must be applied precisely — AI retrieval risks partial or incorrect application |
| Call center scripts | Agent response trees, objection handling guides, escalation flows | Conditional and conversational — correct answer depends on context Lucas doesn't have |
A note on compliance content
Regulatory and legal documents are often uploaded with the intention of keeping Lucas "within bounds." This is understandable, but the Knowledge Base is not the right mechanism. Lucas may retrieve a relevant-sounding clause and incorporate it incorrectly, out of context, or incompletely.
If compliance guardrails are required, work with your account manager to configure them at the platform level — not through Knowledge Base uploads.
Tip: If you'd normally use a spreadsheet filter or a decision tree to find the answer, keep it out of the Knowledge Base.
When the Knowledge Base Is Used
The Knowledge Base is searched during the brief flow — when you start a video from a prompt or a short brief and Lucas needs to find relevant content to build the script from.
Other flows do not use the Knowledge Base because they are already grounded in the material you provide:
| Flow | Source of Truth |
|---|---|
| Brief / prompt to video | Knowledge Base |
| Document to video | The uploaded document (PDF, DOCX) |
| Presentation to video | The uploaded slide deck |
| Image / screenshot to video | The uploaded visuals |
| Walkthrough to video | The recorded or uploaded screen content |
In document-based flows, Lucas generates the script entirely from the file you uploaded — the Knowledge Base is not searched. There is no need to re-upload those same files to the Knowledge Base.
How Lucas Uses the Knowledge Base
When you generate a video from a brief, Lucas retrieves the most relevant sections from your Knowledge Base and uses them to write the script. It does not use the full document — it finds the passages most relevant to your prompt and grounds the output in those.
This means:
- Specific product details come from your approved content, not AI inference
- Terminology matches what your documents use
- Scripts stay within the scope of what you've authorized
Note: The Knowledge Base improves script accuracy. It does not replace Brand DNA, which governs tone, visual style, and voice. The two work together.
Managing Your Knowledge Base
Location: Lucas > Settings > Knowledge Base
From here you can upload new documents, view existing ones, and remove outdated content.
Supported file types: PDF, DOCX, TXT, URL
Tip: Keep your Knowledge Base current. If a product changes and the old document stays, Lucas may generate scripts with outdated information.
Common Questions
Q: How many documents can I upload? A: Limits depend on your account plan. Contact your account manager for details.
Q: Can I upload a URL instead of a file? A: Yes. Lucas will fetch and index the page content. Note that dynamically loaded content (content that requires user interaction to appear) may not be captured.
Q: Does every video use the Knowledge Base? A: No. The Knowledge Base is used in the brief flow only — when you start from a prompt. Document-based flows (document to video, presentation to video, walkthrough to video) use the uploaded file as the source, not the Knowledge Base.
Q: Can different teams have separate Knowledge Bases? A: Knowledge Base configuration is account-level. Contact your account manager if you need workspace-level separation.