Data Connectors
Ingest from anywhere
Connect your existing data sources in minutes. New connectors added monthly.
How It Works
Three steps to grounded AI
Connect & Ingest
Point Tensoras at your data sources -- S3 buckets, databases, Confluence wikis, Notion workspaces, or upload files directly. We parse, clean, and prepare your content automatically.
Chunk, Embed & Index
Choose from smart chunking strategies (recursive, semantic, sentence-window). Embed with BGE, E5, or Cohere models. Build a hybrid index that combines vector similarity with BM25 keyword search.
Retrieve & Generate
Query your knowledge base through our inference API. We retrieve the most relevant chunks, rerank them, and generate grounded responses with inline citations and confidence scores.
Smart Chunking
The right chunk for every document
Recursive Character
Split by paragraph, sentence, then character with overlap
Best for: General-purpose documents
Semantic Chunking
Use embeddings to find natural topic boundaries
Best for: Long-form content with distinct topics
Sentence Window
Small chunks for retrieval, expanded context for generation
Best for: Precision-critical use cases
Markdown / Code
Respect heading hierarchy and code block boundaries
Best for: Technical documentation & source code
Hybrid Search
Combine the precision of BM25 keyword search with the semantic understanding of dense vector embeddings. Reciprocal Rank Fusion merges both result sets into a single, high-quality ranking.
- Dense vector similarity (cosine, dot product)
- Sparse BM25 keyword matching
- Reciprocal Rank Fusion (RRF)
- Cross-encoder reranking
- Metadata filtering & facets
Vector Search
Semantic relevance: 0.92
BM25 Search
Keyword match: 0.78
RRF Fusion
Combined score: 0.96
A few lines of code
Create a knowledge base, ingest documents, and query with citations.
from tensoras import Tensoras
client = Tensoras()
# Create a knowledge base
kb = client.knowledge_bases.create(
name="product-docs",
embedding_model="bge-large-en-v1.5"
)
# Ingest documents
kb.ingest(source="s3://my-bucket/docs/")
# Query with citations
response = client.chat.completions.create(
model="llama-3.3-70b",
knowledge_base=kb.id,
messages=[{
"role": "user",
"content": "How do I configure webhooks?"
}]
)