Knowledge specific to your business
Add policies, manuals, frequently asked questions and product documentation so they become part of the context available to AI.
Knowledge base for context-aware answers
WhatsApp RAG chatbot: answer using your documents
Turn PDF files, text documents and web pages into a knowledge base for your assistant. When a question arrives, ChatCol retrieves related passages and adds them as context before generating an answer.
- PDF and TXT files
- Web pages and sitemaps
- Similarity retrieval
- Language-adapted answers
Answers connected to your business information
RAG adds a retrieval step before generation so the assistant can use the content you indexed instead of relying only on general knowledge.
Sources from files or web pages
Upload PDF or TXT documents, index a public HTTPS URL or use a sitemap to add several pages.
Relevant passage retrieval
Content is divided and vectorised so the system can find passages closest to the question before building its context.
Answers in the contact's language
The assistant adapts retrieved information to the language detected in the latest message, even when the relevant source uses another language.
Step-by-step preparation
How to prepare a RAG knowledge base for your chatbot
Answer quality depends on clear, current and consistent sources. Start with a small set and test real questions before expanding it.
Select reliable information
Collect policies, product sheets, instructions or frequently asked questions that are current and do not contradict one another.
Add files or URLs
Upload a PDF or TXT, enter a public HTTPS page or add a compatible sitemap from the Intelligence section.
Wait for indexing
ChatCol extracts the text, divides it into passages and generates the representations required to retrieve related context.
Test real questions
Try common cases, review whether retrieved information is sufficient and adjust sources or instructions when necessary.
Clear boundaries for preparing sources
The current implementation prioritises controlled text formats and bounded ingestion so you can review what information enters the knowledge base.
- Accepts one PDF or TXT file up to 10 MB per upload and rejects empty or incompatible files.
- Can index one public HTTPS page or attempt the first 15 entries in an XML sitemap; each URL must pass HTTPS validation.
- Retrieves up to three related passages to add to the context of each AI answer.
- Combines document context with configured instructions and intents before generating an answer.
Business information that fits a RAG knowledge base
Choose content with concrete answers that can be kept current. Sensitive or ambiguous questions should retain a route to a person.
Policies and customer service
Answer questions about exchanges, conditions, coverage or processes using the policies you have indexed.
Product catalogue and documentation
Help people find features, compatibility details or instructions from technical sheets and help pages.
Procedures and frequently asked questions
Reuse approved guides and answers to help contacts and hand over to the team when the content is insufficient.
WhatsApp RAG chatbot questions
Answers about sources, retrieval and the limits of a knowledge base connected to AI.
What does RAG mean in a chatbot?
RAG combines retrieval and generation. Before drafting an answer, the system searches indexed content for related passages and adds them to the context given to the AI model.
Which files can I add to the knowledge base?
ChatCol's current implementation accepts PDF and TXT files up to 10 MB. You can also index public HTTPS pages and compatible XML sitemaps.
Can I index my entire website?
You can add an individual page or a sitemap. ChatCol attempts no more than the first 15 entries in each sitemap and only indexes valid public HTTPS URLs.
Do documents update automatically when a source changes?
Not currently. The knowledge base reflects content captured during indexing. If a policy, file or page changes, delete the old source and index its current version again.
Does RAG guarantee that AI will never be wrong?
No. RAG supplies relevant context and reduces reliance on general knowledge, but it does not guarantee perfect answers. Test real questions, keep sources current and route sensitive or insufficient cases to a person.
What do I need to use these answers on WhatsApp?
You need to connect a compatible WhatsApp Business Platform channel and meet Meta's applicable requirements. Conversation windows and approved-template requirements still apply when relevant.