Metadata-Version: 2.4
Name: rivera-edge
Version: 0.1.0
Summary: Rivera Edge — on-device semantic search: local embeddings + exact cosine retrieval in a single SQLite file. No server, no cloud calls.
Project-URL: Homepage, https://api.wirtel.ca
Project-URL: Documentation, https://api.wirtel.ca/documentation
Project-URL: Repository, https://github.com/Mohit1298/rivera-edge
Author-email: Mohit Bendale <mohitab2005@gmail.com>
License-Expression: MIT
Keywords: edge,embeddings,on-device,rivera,semantic-search,vector-search
Requires-Python: >=3.10
Requires-Dist: fastembed>=0.4
Requires-Dist: numpy>=1.24
Description-Content-Type: text/markdown

# Rivera Edge (beta)

On-device semantic search from the [Rivera](https://api.wirtel.ca) family:
local embeddings (fastembed, CPU-only ONNX) + exact cosine retrieval, stored in
a single SQLite file. **No server. No cloud calls. No API key.**

```bash
pip install rivera-edge
```

```python
from rivera_edge import RiveraEdge

edge = RiveraEdge()                     # data at ~/.rivera-edge/edge.db
edge.add("notes", [
    {"id": "1", "text": "User prefers pytest over unittest"},
    {"id": "2", "text": "Deploys go through Railway", "metadata": {"type": "fact"}},
])
hits = edge.search("notes", "which test framework?")   # searchable instantly
print(hits[0]["text"], hits[0]["score"])
```

Same retrieval semantics as Rivera Cloud/On-Prem (exact, deterministic, instant
ingestion) — scoped to a single process and disk. Namespaces, upsert by id,
metadata, multi-namespace search, thresholds.

First call downloads the embedding model (~130 MB, cached). Beta: API may
evolve; sync with Rivera Cloud is on the roadmap.

MIT © Mohit Bendale
