Blob store

Payloads over 1 MB are automatically stored on disk by SHA-256 hash, keeping tokens lightweight. Use maybe_store() and maybe_resolve() for transparent handling.

When blob storage is used

Symbolic expressions in computational mathematics can exceed 1 MB in serialized form. The blob store handles large payloads by storing them on disk, referenced by SHA-256 hash.

On-disk layout and blob references

Payloads exceeding 1 MB are stored in a two-level directory structure:

.elegua/blobs/
  ab/
    cdef0123...  (SHA-256 hex, first 2 chars as directory)

The token stores a reference instead of the full payload:

{"blob": "abcdef0123456789..."}

Create a BlobStore

from pathlib import Path
from elegua.blobstore import BlobStore

store = BlobStore(root=Path(".elegua/blobs"))

Store and retrieve

ref = store.put({"large_expression": "..." * 1_000_000})
# ref = {"blob": "a1b2c3d4..."}

data = store.get(ref["blob"])
# data == original payload

Transparent handling

Use maybe_store() and maybe_resolve() for payloads that may or may not exceed the threshold:

# Stores only if > 1 MB, otherwise returns the payload unchanged
result = store.maybe_store(payload)

# Resolves blob refs back to full payloads, passes plain dicts through
data = store.maybe_resolve(result)

Check the threshold

store.should_store(small_payload)  # False
store.should_store(large_payload)  # True

Blob-store guarantees

• Deterministic — the same payload always produces the same hash
• Content-addressed — duplicate payloads share the same file
• Two-level directory — prevents filesystem performance issues with thousands of files
• Idempotent — storing the same payload twice is safe