feat: Add Text Embeddings Inference (TEI) provider support

[jayscambler]

Summary

Adds support for Hugging Face's Text Embeddings Inference (TEI) server as an embedding provider in ContextFrame, providing high-performance, self-hosted embeddings for 100+ open-source models.

Changes

• ✨ New TEIProvider class implementing the EmbeddingProvider interface
• 🔧 Updated factory function to support provider_type="tei"
• 📚 Added comprehensive documentation in embedding providers guide
• 🧪 Unit tests with mocks for TEI functionality
• 💡 Complete example demonstrating TEI usage patterns
• 📦 Added httpx as optional dependency for lightweight HTTP client

Features

• Local and Remote Support: Works with TEI servers running locally or remotely
• Authentication: Bearer token support for secured instances
• Health Checks: Built-in server health monitoring
• Error Handling: Automatic retries with exponential backoff
• Configuration: Flexible timeout, truncation, and normalization options
• Minimal Dependencies: Only requires httpx (25KB) for HTTP communication

Example Usage

from contextframe.embed import create_embedder

# Local TEI server
embedder = create_embedder(
    model="BAAI/bge-large-en-v1.5",
    provider_type="tei",
    api_base="http://localhost:8080"
)

# Embed documents
results = embedder.embed_batch(["Document 1", "Document 2"])

Benefits

• 🚀 Performance: Flash Attention, ONNX optimization, dynamic batching
• 🔐 Privacy: Self-hosted solution for sensitive data
• 🎯 Flexibility: Supports any Sentence Transformer or BERT-based model
• 📊 Production Ready: Built-in metrics, monitoring, health checks
• 💻 Hardware Support: GPU acceleration (CUDA) and CPU optimizations

Testing

The implementation includes comprehensive unit tests using mocks. Note: Current test suite has NumPy compatibility issues unrelated to this PR that will be addressed separately.

Related

• Linear Issue: CFOS-46
• TEI Documentation: https://huggingface.co/docs/text-embeddings-inference

Docker Setup

# GPU deployment
docker run --gpus all -p 8080:80 -v $PWD/data:/data \
  ghcr.io/huggingface/text-embeddings-inference:1.7 \
  --model-id BAAI/bge-large-en-v1.5

# CPU deployment
docker run -p 8080:80 -v $PWD/data:/data \
  ghcr.io/huggingface/text-embeddings-inference:cpu-1.7 \
  --model-id BAAI/bge-large-en-v1.5

[linear]

CFOS-46 Add Text Embeddings Inference (TEI) support to embeddings module