Skip to main content

Hermes Python client

Async Python client for the Hermes gRPC search server.

Installation

pip install hermes-client-python

Python 3.10 or newer is required.

Quick start

import asyncio

from hermes_client_python import HermesClient


async def main():
    async with HermesClient("localhost:50051") as client:
        await client.create_index(
            "articles",
            """
            index articles {
                field id: text<raw> [primary, stored]
                field title: text<simple> [indexed, stored]
                field body: text<simple> [indexed, stored]
            }
            """,
        )

        indexed, error_count, errors = await client.index_documents(
            "articles",
            [
                {"id": "1", "title": "Hello World", "body": "First article"},
                {"id": "2", "title": "Hermes Search", "body": "Fast retrieval"},
            ],
        )
        if error_count:
            raise RuntimeError(errors)
        print(f"Indexed {indexed} documents")

        await client.commit("articles")

        results = await client.search(
            "articles",
            query={"match": {"field": "title", "text": "hello"}},
            fields_to_load=["title", "body"],
        )
        for hit in results.hits:
            print(hit.address, hit.score, hit.fields)

        if results.hits:
            document = await client.get_document("articles", results.hits[0].address)
            print(document.fields if document else "document not found")


asyncio.run(main())

The context manager connects and closes the channel. Manual callers use await client.connect() / await client.close().

Index management

await client.list_indexes()
info = await client.get_index_info("articles")
await client.reorder("articles")
await client.retrain_vector_index("articles")

index_documents returns (indexed_count, error_count, errors); index_documents_stream takes an async iterable and returns counts. Inspect errors, then commit() to publish accepted work. Repeated values are lists; flat numeric lists are dense vectors, (dimension, weight) pairs are sparse vectors. See client types.

Delete and upsert documents

With a text primary key, delete by exact key or supply a complete replacement:

await client.delete_document("articles", "2")
await client.upsert_document("articles", {"id": "1", "title": "Updated title"})
await client.commit("articles")

delete_documents / upsert_documents return DocumentMutationResult with accepted_count and errors: [{index, error}]. Single-item helpers raise on rejection. Missing deletes succeed; upserts insert missing keys. Staged rows can be replaced/deleted again before commit; the latest accepted version wins. Optional content hashes skip unchanged writes.

Limits: 100,000 deletion keys / 8 MiB key bytes; 1,000 replacements / 32 MiB encoded protobuf, or one replacement / 200 MiB including the request envelope. Broker commits are atomic per partition. See mutation semantics.

Compact deleted rows

await client.force_merge("articles")  # Retain tombstones.
await client.force_merge("articles", compact=True)  # Physically remove deleted rows.

Compaction handles singleton segments and may change addresses and BM25 scores. Index info exposes num_docs, physical_num_docs, num_deleted_docs, and deleted_ratio. Use primary keys for durable identity.

Searching

search(index_name, query=..., limit=10, fields_to_load=[...]) accepts one query variant: term, match, phrase, boolean, sparse_vector, dense_vector, binary_dense_vector, boost, range, prefix, all, or fusion. See query types and the wire contract for options.

results = await client.search(
    "articles",
    query={"boolean": {
        "must": [{"match": {"field": "title", "text": "search"}}],
        "must_not": [{"term": {"field": "title", "term": "draft"}}],
    }},
    fields_to_load=["title"],
)

get_document(index_name, hit.address) uses the full segment/document address and returns None on NOT_FOUND.

Ranking diagnostics and recall traces

Search options include_rrf_scores=True and tracing=True default to false. RRF diagnostics describe organic branch nominations; traces retain bounded candidates and query trees, including hits outside the final page. Neither changes retrieval depth or ranking. Oversized exports fail explicitly.

For named branches, use l1={"formula": "0.2 * title + 0.8 * body + 3 * rrf"}. The formula is the only L1 scoring interface; old coefficient fields are removed. See candidate scoring for backfill, passage selection, expression limits, capability versions, and distributed behavior.

Deadlines and errors

RPCs accept timeout in seconds, overriding the constructor's default_timeout. gRPC failures raise grpc.aio.AioRpcError, except document NOT_FOUND as above. An expired mutation may already be staged, and an accepted commit continues after disconnection. Resolve the outcome before retrying replacements.

Development

From this directory:

uv sync --group dev --group test
uv run ruff check .
uv run ruff format --check .
uv run pytest tests/test_client_unit.py
uv run --group dev python generate_proto.py

Integration tests require target/debug/hermes-server. Regenerate bindings after protocol changes.

Metadata

Release files for hermes-client-python 1.8.149

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for hermes-client-python 1.8.149
File Size Uploaded Provenance
hermes_client_python-1.8.149.tar.gz 24.2 kB Without provenance Details

Built distribution (wheel)

Table of built distributions (wheels) for hermes-client-python 1.8.149
File Interpreter ABI Platform
hermes_client_python-1.8.149-py3-none-any.whl Python 3 none any Details

Total release size: 49.9 kB

Release files / hermes_client_python-1.8.149.tar.gz

Download URL hermes_client_python-1.8.149.tar.gz
Size 24.2 kB
Tags Source
SHA-256 checksum
How to use checksums
4fba04aeff5752a1c7a711a7c07fea0809a8796a1bcc44a9a9508410c29e75f5
BLAKE2b-256 checksum
How to use checksums
b4646a093675890bac452b479fd958c1effe0da53dc820cc0eaeed246110292f
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.12.18 {"installer":{"name":"uv","version":"0.12.18","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

Release files / hermes_client_python-1.8.149-py3-none-any.whl

Download URL hermes_client_python-1.8.149-py3-none-any.whl
Size 25.8 kB
Tags Python 3
SHA-256 checksum
How to use checksums
b117b1f3a588c6d4792167c24742984f6ab9b7083788a3e3da00c0f9618127e6
BLAKE2b-256 checksum
How to use checksums
acce62d8fd3571d52c842a606e1e95c39e1c477d2bdcd2979d11bc47a3dabb91
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.12.18 {"installer":{"name":"uv","version":"0.12.18","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

Security for hermes-client-python 1.8.149

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.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

Release history Release notifications | RSS feed

This release

1.8.149 This release

2 release files No provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page