hindsightai/ hs-embed-1

21 1 week ago

Multilingual text embedding model built for Hindsight's agent memory. 384 dimensions, truncatable to 256, 128 or 64.

embedding 
ollama pull hindsightai/hs-embed-1

Models

View all →

1 model

hs-embed-1:latest

238MB · 8K context window · Text · 1 week ago

Readme

Hindsight

hs-embed-1

hs-embed-1 is a multilingual text embedding model built for Hindsight, an open-source memory system for agents.

Highlights

  • Multilingual: English, Spanish, French, German, Portuguese, Japanese, Korean and Chinese (Simplified and Traditional).
  • Variable dimensions: 384-dimensional vectors that can be truncated to 256, 128 or 64.
  • Long inputs: up to 8,192 tokens.
  • Small: 112M parameters. 98M of them are the token embedding table, so each forward pass runs a 13M-parameter transformer.
  • No prompts: queries and documents are encoded the same way, with no prefix.

Model overview
Model type Text embedding (bi-encoder)
Base model hotchpotch/bekko-embedding-v1-a25m
Parameters 112M
Languages en, es, fr, de, pt, ja, ko, zh
Context length 8,192 tokens
Embedding dimensions 384 (truncatable to 256, 128, 64)
Pooling Mean
Similarity Cosine
Weights GGUF, F16

Evaluation

hs-embed-1 benchmarks

Usage

ollama pull hindsightai/hs-embed-1

REST API

curl http://localhost:11434/api/embed -d '{
  "model": "hindsightai/hs-embed-1",
  "input": ["Where did Maya move last spring?", "Maya moved to Lisbon in April."]
}'

For shorter vectors, add "dimensions": 256 (or 128 or 64). Ollama keeps the first values and normalizes them again.

Python

import ollama

response = ollama.embed(
    model="hindsightai/hs-embed-1",
    input=["Where did Maya move last spring?", "Maya moved to Lisbon in April."],
)
print(len(response.embeddings[0]))  # 384

Hindsight

Hindsight’s OpenAI embeddings provider works with Ollama’s OpenAI-compatible API:

HINDSIGHT_API_EMBEDDINGS_PROVIDER=openai
HINDSIGHT_API_EMBEDDINGS_OPENAI_BASE_URL=http://localhost:11434/v1
HINDSIGHT_API_EMBEDDINGS_OPENAI_MODEL=hindsightai/hs-embed-1
HINDSIGHT_API_EMBEDDINGS_OPENAI_API_KEY=ollama

More

Safetensors, ONNX, and serving with Sentence Transformers or Text Embeddings Inference: hindsight-ai/hs-embed-1 on Hugging Face.

License

MIT