0.2.0

Adaptensor Python SDK

Private Document AI Platform — Upload, index, and search your documents with AI. 100% self-hosted, no external APIs.

Features

• 🔒 100% Private — Your data never leaves your infrastructure
• ⚡ Fast Search — Sub-150ms semantic queries across thousands of documents
• 📦 AdaptHex Compression — 4-8x smaller vectors with 99.6% accuracy
• 🔌 Simple API — Upload, index, search in 3 lines of code
• 🚀 TPU Accelerated — 25% cheaper than GPU clouds

Installation

pip install adaptensor

Quick Start

from adaptensor import Adaptensor

# Initialize client
client = Adaptensor(api_key="your-api-key")

# Upload documents
client.documents.upload("contract.pdf")
client.documents.upload("report.docx")

# Build search index
result = client.documents.index()
print(f"Indexed {result.total_chunks} chunks")

# Search
results = client.query("liability clauses", top_k=5)

for chunk in results:
    print(f"[{chunk.score:.2f}] {chunk.text[:100]}...")

Authentication

Set your API key via environment variable or pass directly:

# Environment variable
export ADAPTENSOR_API_KEY="your-api-key"

# Or pass directly
client = Adaptensor(api_key="your-api-key")

Upload Documents

Single File

doc = client.documents.upload("report.pdf")
print(f"Uploaded: {doc.doc_id} ({doc.size_bytes:,} bytes)")

Multiple Files

docs = client.documents.upload_many([
    "doc1.pdf",
    "doc2.pdf",
    "doc3.pdf"
], on_progress=lambda c, t, d: print(f"{c}/{t}: {d.filename}"))

Supported Formats

• PDF (.pdf)
• Word (.docx)
• Text (.txt)
• HTML (.html)
• Markdown (.md)
• CSV (.csv)

Build Index

After uploading documents, build the search index:

result = client.documents.index()

print(f"Status: {result.status}")
print(f"Documents: {result.documents_processed}")
print(f"Chunks: {result.total_chunks}")
print(f"Time: {result.elapsed_seconds:.1f}s")

Search

Basic Search

results = client.query("contract termination", top_k=5)

for chunk in results:
    print(f"Score: {chunk.score:.3f}")
    print(f"Text: {chunk.text}")
    print(f"Source: {chunk.metadata.get('filename')}")
    print("---")

Search Result Object

results = client.query("machine learning")

print(f"Query: {results.query}")
print(f"Results: {results.count}")
print(f"Time: {results.elapsed_ms:.1f}ms")

# Iterate through results
for chunk in results:
    print(chunk.text)

Embeddings

Generate embeddings with optional AdaptHex compression:

# With compression (default)
result = client.embed("machine learning algorithms")
print(f"Dimensions: {result.dimensions}")
print(f"Hex: {result.hex_embedding[:50]}...")

# Without compression
result = client.embed("machine learning", quantize=False)
print(f"Vector: {result.embedding[:5]}...")

Compression Modes

# Use different compression modes
result = client.embed("text", mode="hex4")  # 8x compression

RAG Chat (Coming Soon)

response = client.chat("What are the key terms in the contract?")
print(response)

Error Handling

from adaptensor import (
    Adaptensor,
    AdaptensorError,
    AuthenticationError,
    DocumentNotFoundError,
    APIError
)

try:
    client = Adaptensor(api_key="invalid-key")
    client.query("test")
except AuthenticationError:
    print("Invalid API key")
except APIError as e:
    print(f"API error: {e} (status: {e.status_code})")
except AdaptensorError as e:
    print(f"General error: {e}")

Configuration

client = Adaptensor(
    api_key="your-api-key",
    base_url="https://your-instance.adaptensor.com",  # Custom endpoint
    timeout=600  # Request timeout in seconds
)

Health Check

if client.health():
    print("API is healthy")
else:
    print("API is down")

Usage Statistics

stats = client.stats()
print(f"Documents: {stats['documents']}")
print(f"Chunks: {stats['chunks']}")
print(f"Queries: {stats['queries']}")

Full Example

from adaptensor import Adaptensor

# Initialize
client = Adaptensor()

# Upload a batch of legal documents
docs = client.documents.upload_many([
    "contracts/agreement_2024.pdf",
    "contracts/amendment_1.pdf",
    "contracts/amendment_2.pdf",
])
print(f"Uploaded {len(docs)} documents")

# Index everything
result = client.documents.index()
print(f"Indexed {result.total_chunks} chunks in {result.elapsed_seconds:.1f}s")

# Search for specific clauses
results = client.query("indemnification obligations", top_k=10)

print(f"\nFound {results.count} results in {results.elapsed_ms:.1f}ms:\n")
for i, chunk in enumerate(results, 1):
    print(f"{i}. [{chunk.score:.2f}] {chunk.metadata.get('filename')}")
    print(f"   {chunk.text[:150]}...")
    print()

Requirements

• Python 3.8+
• requests >= 2.25.0

Links

• Documentation: docs.adaptensor.com
• API Reference: docs.adaptensor.com/api
• GitHub: github.com/adaptensor/adaptensor-python
• PyPI: pypi.org/project/adaptensor

License

MIT License - see LICENSE for details.

---

Adaptensor — Your Data. Your Infrastructure. Your AI.