Summary

jsonable_encoder(some_bytes) currently calls some_bytes.decode() with no error handling. For any bytes value that isn't valid UTF-8 (raw binary data — hashes, encrypted blobs, protobuf, DB blob columns), this raises an unhandled UnicodeDecodeError straight from deep inside FastAPI's internals, with no indication of what went wrong or how to fix it — even though a working escape hatch (custom_encoder={bytes: ...}) already exists and just isn't discoverable from the error.

Fix

Wrap the decode in a named bytes_encoder() function that re-raises as a ValueError (consistent with the ValueError jsonable_encoder already raises for other unencodable types) with a message that explains the problem and points at custom_encoder as the fix. The existing pattern of naming non-trivial encoders as top-level functions (e.g. decimal_encoder, isoformat) is followed rather than keeping an inline lambda.

Testing

Added three tests to tests/test_jsonable_encoder.py:

• test_encode_bytes_valid_utf8: valid UTF-8 bytes still encode correctly (unchanged behavior)
• test_encode_bytes_invalid_utf8_raises_clear_error: invalid UTF-8 now raises ValueError with an actionable message
• test_encode_bytes_invalid_utf8_with_custom_encoder: the existing custom_encoder escape hatch still works and now has a regression test

Not a breaking change for the success path; the only change is the exception type/message for previously-crashing input.