Compressor

Optional value compression wired into CacheManager. Compresses bytes before writing to the backend and decompresses on read. Reduces memory and disk usage for large LLM responses.

Built-in Compressors

Class	Algorithm	Use when
NoopCompressor	None (passthrough)	Default — no overhead
GzipCompressor	gzip (Python stdlib)	Large text responses, embedding payloads

---

Usage

Default (no compression)

from chengeta_ai import CacheManager, InMemoryBackend, CacheKeyBuilder

manager = CacheManager(
    backend=InMemoryBackend(),
    key_builder=CacheKeyBuilder(),
)  # NoopCompressor by default

Gzip compression

from chengeta_ai import GzipCompressor, CacheManager, InMemoryBackend, CacheKeyBuilder

manager = CacheManager(
    backend=InMemoryBackend(),
    key_builder=CacheKeyBuilder(),
    compressor=GzipCompressor(level=6),  # 0=fastest, 9=best compression
)

Compression applies automatically on manager.set() for bytes values and decompresses on manager.get(). Non-bytes values (e.g. dicts stored directly) are passed through unchanged.

With Redis backend (reduces network transfer)

from chengeta_ai.backends.redis_backend import RedisBackend

manager = CacheManager(
    backend=RedisBackend(url="redis://localhost:6379/0"),
    key_builder=CacheKeyBuilder(),
    compressor=GzipCompressor(level=6),
)

---

Custom Compressor

Implement the Compressor protocol — two methods required:

from chengeta_ai import Compressor

class Lz4Compressor:
    def compress(self, data: bytes) -> bytes:
        import lz4.frame
        return lz4.frame.compress(data)

    def decompress(self, data: bytes) -> bytes:
        import lz4.frame
        return lz4.frame.decompress(data)

manager = CacheManager(
    backend=InMemoryBackend(),
    key_builder=CacheKeyBuilder(),
    compressor=Lz4Compressor(),
)

LZ4 is ~4× faster than gzip with similar compression ratios for LLM text — good for high-throughput production deployments.

---

When to Use Compression

Scenario	Recommendation
Short LLM responses (<1KB)	NoopCompressor — overhead exceeds savings
Long LLM responses (>4KB)	GzipCompressor(level=6)
Redis backend (reduces network bytes)	GzipCompressor
Disk backend (reduces storage)	GzipCompressor
Latency-critical hot path	NoopCompressor or LZ4

---

API Reference

Parameter	Type	Default	Description
level (GzipCompressor)	int	6	Compression level 0–9. 0=no compression, 9=max

Compressor is @runtime_checkable — verify with isinstance(obj, Compressor).