Glossary
Plain-language definitions for buyers and practitioners.
RAG (Retrieval-Augmented Generation)
Combines search over your documents with LLM generation for cited answers.
Embedding
Numeric vector representation of text used for semantic similarity search.
Vector database
Stores embeddings for fast nearest-neighbor retrieval (e.g. Qdrant).
Large language model (LLM)
Neural model that generates text from prompts; KobiGPT uses Gemini by default.
Fine-tuning
Training a base model on custom data; often costlier than RAG for SMEs.
Prompt engineering
Crafting instructions and context so the model behaves reliably.
Chunking
Splitting documents into segments before embedding (e.g. 512 tokens).
Cosine similarity
Metric comparing vector angles; common in semantic search.
HNSW index
Approximate nearest-neighbor graph index used in Qdrant.
Knowledge base
Curated company documents indexed for assistants.
Multi-tenant
Isolated data per company on shared infrastructure.
KVKK
Turkey’s personal data protection law; relevant for SME AI deployments.
GDPR
EU data protection regulation; consider for cross-border SaaS.
Token
Subword unit billed by LLM APIs; drives Kobi Kredi estimates.
Context window
Maximum tokens the model can read in one request.
Hallucination
Confident but incorrect model output; mitigated by RAG citations.
Semantic search
Search by meaning via embeddings rather than keywords only.
BM25
Classic lexical ranking; sometimes hybridized with vectors.
Hybrid search
Combines keyword and vector retrieval for better recall.
Grounding
Anchoring answers to retrieved sources.
Citation
Reference to document snippets shown with answers.
Department assistant
Scoped chatbot for HR, legal, sales, etc.
Google Drive sync
Connecting folders as a document source in KobiGPT.
Gemini
Google’s LLM family used as KobiGPT’s default chat model.
OpenRouter
API gateway for embeddings and alternate models.
Qdrant
Vector database used for per-company collections.
BullMQ
Redis queue for async document embedding jobs.
Self-hosted
Running KobiGPT on your infrastructure for data residency.
Kobi Kredi
Usage credits metering external AI and platform costs.
SME AI
AI products priced and designed for small and mid-size firms.
Chatbot
Conversational UI over your knowledge base.
Enterprise AI
Org-wide AI with governance, SSO, and policies.
On-premise
Software deployed inside customer network.
Data residency
Geographic location where data is stored and processed.
PII
Personally identifiable information subject to KVKK/GDPR.
Access control
Rules for who can read or chat with which documents.
Role-based access
Admin, manager, viewer roles in KobiGPT.
API key
Secret used to authenticate programmatic API calls.
Webhook
HTTP callback for integrations and automation.
SSL/TLS
Encryption in transit for web and API traffic.