jammi-ai
The Python client (import jammi) for a Jammi
engine — one connect(target) for local (embedded) and remote targets.
jammi-ai is the base client: a universal py3-none-any wheel that
links no candle/ML stack — just grpcio + protobuf + pyarrow — for
production where the engine runs behind a server. It also discovers the compiled
in-process engine as an OPTIONAL backend: pip install jammi-ai[embedded]
pulls jammi-ai-native, and then connect("file://…") resolves the local target
to an in-process engine (direct FFI) through the SAME front door. Absent the
extra, the base stays native-free and a file:// target raises a truthful
NoEmbeddedEngineError.
One front door
import jammi
db = jammi.connect("https://engine.example.com")
db.add_source("patents", url="s3://corpus/patents.parquet", format="parquet")
db.generate_embeddings(source="patents", model="local:tiny_bert",
columns=["abstract"], key="id", modality="text")
q = db.encode_query(model="local:tiny_bert", query="quantum computing")
hits = db.search("patents", query=q, k=5) # -> pyarrow.Table
connect(target) is the Python mirror of the Rust Jammi::open(Target):
transport is config, not a code path.
Authenticated channels
A bearer-protected endpoint is reached by attaching credentials to the channel — the bearer rides the channel (attached once at connect), not threaded through every call:
from jammi import connect, BearerCredentials
db = connect("https://engine.example.com", credentials=BearerCredentials(token))
The same works on a plaintext grpc:// target for local development. The
bearer covers the typed gRPC verbs (on the channel) and db.sql() (the Flight
SQL lane, per call) alike.
| target | transport |
|---|---|
https://host / grpcs://host:8081 |
secure remote |
http://host / grpc://host:8081 |
plaintext remote |
file:///data |
in-process engine (direct FFI) with the [embedded] extra installed; otherwise raises NoEmbeddedEngineError pointing at pip install jammi-ai[embedded] |
Scaling local→remote is an env flip (connect(os.environ["JAMMI_TARGET"]))
with no code change. The lean deploy build and the embedded build are ONE
package — the same import jammi, connect unchanged — differing only by
whether the [embedded] extra is present.
Generated from the canonical proto
The wire stubs under jammi/_generated/ are generated from the same
jammi.v1 proto the engine, npm client, and embed wheel speak — one source, no
parallel schema. Regenerate after a proto change:
pip install -e '.[dev]'
make generate
1 maintainer
Metadata
1 maintainer
Release files for jammi-ai 0.54.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| jammi_ai-0.54.0-py3-none-any.whl | Python 3 | none | any | Details |
Release files / jammi_ai-0.54.0-py3-none-any.whl
| Download URL | jammi_ai-0.54.0-py3-none-any.whl |
|---|---|
| Size | 199.0 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
7ce0315960436d78765846fed7f7a7d3549139500bb3087b3827c843bc7e8532
|
|
BLAKE2b-256 checksum How to use checksums |
6bede1de4184cabfcfd4edab98e7b98519cbc81a3e69abdfd4c583476faafc44
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 3, 2026.
Transparency log
