Library, not the tab

Your strategy is Python. The tab will not eval it.

If the cut you want is not in the dropdown, do not paste a generated worker into the playground. Register a BaseChunker. Then create_chunker, the CLI, compare, and benchmark all see it. If it belongs in the lab later, port it with a golden fixture. That order is the point.

Minimum chunker

from chunking_strategy.core.base import (
    BaseChunker, Chunk, ChunkingResult, ChunkMetadata, ModalityType,
)
from chunking_strategy.core.registry import register_chunker

@register_chunker(
    name="my_word_count_chunker",
    category="text",
    description="Chunks on a word budget",
    default_parameters={"words_per_chunk": 50},
)
class MyWordCountChunker(BaseChunker):
    def __init__(self, words_per_chunk=50, **kwargs):
        super().__init__(name="my_word_count_chunker", category="text", **kwargs)
        self.words_per_chunk = words_per_chunk

    def chunk(self, content, source_info=None, **kwargs):
        if isinstance(content, bytes):
            content = content.decode("utf-8")
        words = str(content).split()
        chunks = []
        for i in range(0, len(words), self.words_per_chunk):
            text = " ".join(words[i:i + self.words_per_chunk])
            chunks.append(Chunk(
                id=f"word_{i // self.words_per_chunk}",
                content=text,
                modality=ModalityType.TEXT,
                metadata=ChunkMetadata(
                    source=(source_info or {}).get("source", "unknown"),
                    chunker_used="my_word_count_chunker",
                    extra={"word_count": len(text.split())},
                ),
            ))
        return ChunkingResult(chunks=chunks)

from chunking_strategy import create_chunker
print(create_chunker("my_word_count_chunker", words_per_chunk=40).chunk(text).chunks)

CLI

chunking-strategy custom load ./my_chunker.py
chunking-strategy custom load-dir ./examples/custom_algorithms
chunking-strategy custom list --detailed
chunking-strategy custom create-template ./my_chunker.py --algorithm-name my_word_count
chunking-strategy custom validate ./my_chunker.py
chunking-strategy custom validate-batch ./examples/custom_algorithms
chunking-strategy custom validate-config ./my-config.yaml
chunking-strategy custom benchmark ./my_chunker.py --compare-with fixed_size --compare-with sentence_based
chunking-strategy benchmark doc.txt --custom-algorithms ./my_chunker.py

from chunking_strategy import load_custom_algorithm, create_chunker
load_custom_algorithm("./my_chunker.py")
print(create_chunker("my_word_count_chunker").chunk(text).chunks)

Worked examples in the repo

CUSTOM_ALGORITHMS_GUIDE.md is the long form (validation, YAML, parallel custom runs). Start with the three files above.

Getting it into the lab later

1. Python chunker + tests.
2. A golden under fixtures/ so JS and Python agree on offsets (Unicode scalar offsets, not UTF-16).
3. A JS port in poc/chunkers/, then move the name from lab_later / python_only to lab in schemas/strategy-tiers.yaml.