Managing Document Visibility
For certain workflows, you might want to attach documents that are visible to the virtual assistant but not to the users in the chat. This is particularly useful for:
Providing background information to the assistant
Including reference materials that don't need to clutter the user interface
Setting up knowledge bases that work behind the scenes
This functionality is available in task configurations, where you can set the "hidden" flag for documents while still specifying their accessibility status.
Understanding RAG (Retrieval Augmented Generation)
RAG (Retrieval Augmented Generation) is a powerful approach that enhances AI language models by connecting them to external knowledge sources. In PrimeThink, RAG is implemented through the document system with the "Search" accessibility status.
How RAG Works in PrimeThink
Indexing Phase
When you upload a document and set its accessibility status to "Search," PrimeThink processes and indexes the document
The system breaks down the document into meaningful chunks
Each chunk is converted into a vector representation (embedding) that captures its semantic meaning
These embeddings are stored in a vector database for efficient retrieval
Retrieval Phase
When a user asks a question or makes a request in the chat
The system converts the query into the same vector space as the document chunks
It performs a similarity search to find the most relevant document sections
The most relevant chunks are retrieved based on semantic similarity, not just keyword matching
Generation Phase
The retrieved document chunks are added to the context window for the AI assistant
The assistant uses both its training knowledge and the retrieved information to generate a response
This allows for more accurate, up-to-date, and contextually relevant answers
Benefits of RAG in PrimeThink
Access to Specialized Knowledge: RAG enables virtual assistants to leverage specific information from your documents that wouldn't be in their general training data.
Reduced Hallucinations: By grounding responses in retrieved document content, RAG significantly reduces the likelihood of AI assistants generating incorrect information.
Customized Responses: The system provides answers that are specific to your organization's knowledge, policies, or domain expertise.
Transparency: Responses can reference specific sources from your documents, making information more traceable and verifiable.
Efficiency: Only the most relevant parts of documents are used, rather than overwhelming the AI with entire document contents.
Understanding CAG (Context Augmented Generation)
CAG (Context Augmented Generation) represents another approach to working with documents in PrimeThink, implemented through the "Context" access status. Unlike RAG, which retrieves only relevant portions of documents, CAG provides the entire document content to the AI assistant.
How CAG Works
When a document's access status is set to "Context," the entire document (if it fits within the context window) is directly provided to the AI assistant.
This approach effectively "augments" the model's context with the complete document, making all information readily available without a retrieval step.
The AI assistant can then reference, analyze, and utilize any part of the document without needing to explicitly request specific sections.
When to Use CAG
CAG is particularly effective for:
Smaller documents that need to be analyzed in their entirety
Situations where the whole document provides important context
Cases where you want to ensure nothing is missed through the retrieval process
Documents where the relationships between different sections are important
While CAG provides comprehensive access to document content, it's limited by the AI's context window size. For larger documents, RAG often provides a more efficient approach by retrieving only the most relevant sections.
RAG vs. Other Document Access Methods
Feature | Archived (Agentic) | Search (RAG) | Context (CAG) |
|---|---|---|---|
AI access method | Explicit tool use | Automatic retrieval | Always available |
Document size limitation | None | None | Limited by context window |
Precision | High (targeted retrieval) | Medium-High (semantic search) | Low (entire document) |
Use case | Specific document lookup | General knowledge queries | Full document analysis |
Context window usage | Efficient | Efficient | Can be inefficient |
Practical Applications
RAG is particularly effective for:
Knowledge Bases: Making company policies, procedures, and FAQ documents accessible to virtual assistants
Legal Document Analysis: Retrieving relevant precedents or clauses from large legal corpora
Research Support: Pulling relevant information from academic papers or reports
Customer Support: Finding accurate product information from technical documentation
Best Practices
Use Archived status when you want assistants to specifically reference documents only when needed
Use Search status for large reference documents where only portions may be relevant to any given query
Use Context status for smaller, highly relevant documents that should be fully considered
Regularly review and update document collections to ensure assistants have access to the most current information
Consider using a mix of status types to create the optimal knowledge environment for your specific use case
For optimal RAG performance, ensure documents are well-structured with clear headings and concise sections
Test different chunk sizes and overlap settings if you have access to advanced RAG configuration options