svo-client 4.0.1.2

Async client for SVO semantic chunker microservice.

svo-client

Async Python client for the SVO semantic chunker. Canonical API: SvoChunkerClient (adapter / JsonRpcClient only).

Install

pip install -e .
# or: pip install svo-client

Quick start

import asyncio
from svo_client import SvoChunkerClient

async def main():
    async with SvoChunkerClient(
        protocol="mtls",
        host="127.0.0.1",
        port=8009,
        cert="mtls_certificates/client/svo-chunker.crt",
        key="mtls_certificates/client/svo-chunker.key",
        ca="mtls_certificates/ca/ca.crt",
        check_hostname=False,
    ) as client:
        chunks = await client.chunk(
            text="Your text.",
            chunk_type="Comment",
            aggregation_mode="paragraph",
            respect_structural_blocks=True,
            language="en",
        )
        print(len(chunks))

asyncio.run(main())

Current Chunker Parameters

• chunk_type="Comment" is passed as RPC field type="Comment" for comments, docstrings, or other pre-extracted blocks. The server owns the Comment keep-block policy.
• aggregation_mode is optional. Supported values are "sentence", "none", and "paragraph". When omitted, the client does not send the field.
• respect_structural_blocks is optional. Use True or False to send respect_structural_blocks; omit it to preserve the server default.

from svo_client import build_file_chunk_params
from svo_client.file_processor import file_chunk_from_local_path

chunks = await client.chunk(
    text="# module comment",
    chunk_type="Comment",
    aggregation_mode="paragraph",
    respect_structural_blocks=True,
)

batch = await client.chunk_batch(
    ["first paragraph", "second paragraph"],
    aggregation_mode="paragraph",
    respect_structural_blocks=False,
)

params = build_file_chunk_params(
    filename="comments.py",
    file_content_base64=encoded,
    chunk_type="Comment",
    aggregation_mode="paragraph",
    respect_structural_blocks=True,
)
result = await client.file_chunk(params)

result = await file_chunk_from_local_path(
    client,
    "comments.py",
    chunk_type="Comment",
    aggregation_mode="paragraph",
    respect_structural_blocks=True,
)

CLI: svo-client or python -m svo_client — thin wrapper; local file chunking orchestration is file_chunk_from_local_path in svo_client.file_processor. Text chunking: chunk (default) and chunk text; file chunking: chunk file or file-chunk (alias); ws prints inbound WebSocket JSON lines to stdout. --help lists all flags. RPC discovery: svo-chunker (list / help / run).

svo-client chunk text -i sample.txt -o out.json --format json --chunk-type Comment
svo-client file-chunk -i sample.md -o out.jsonl --aggregation-mode paragraph --respect-structural-blocks
svo-client file-chunk -i comments.py --no-download --chunk-type Comment --no-respect-structural-blocks

Configuration precedence (CLI → SVO_CHUNKER_CLIENT_* → JSON), --config, and CLI details (ws, --quiet-status, file-chunk status on stderr): svo_client/README.md.

Tests

.venv/bin/python -m pytest tests/ -q

Live integration (chunker on 127.0.0.1:8009): port must be listening or free for managed spawn; see tests/conftest.py.