Float types are not supported. Use Decimal types instead.

TL;DR — boto3 raised a Python TypeError because you passed a float to DynamoDB. DynamoDB stores numbers as arbitrary-precision decimals (up to 38 digits), and binary floats can't represent those exactly — so boto3 refuses them. Convert to decimal.Decimal before writing, ideally via str() so you don't inherit the float's rounding.

What it means

TypeError: Float types are not supported. Use Decimal types instead.

This is a client-side error raised by boto3 (the AWS SDK for Python), not a DynamoDB service response — the SDK's serializer rejects float before the request is ever sent. DynamoDB's N type holds a decimal number with up to 38 digits of precision; Python float is IEEE-754 binary, which can't round-trip those values losslessly. boto3 fails loud rather than silently store an approximation.

Why it happens

• Writing a raw float — a price 30.51, a computed average, a json.loads() result (JSON numbers with decimals become Python floats).
• Nested floats — a float buried inside a dict/list you're putting; boto3 walks the whole structure and rejects the first one.
• Arithmetic results — total / count, sum(...), or any division that yields a float.
• Third-party data (pandas, an API response) that hands you numpy/float64 values.

How to fix it

1. Convert to decimal.Decimal before writing:

   from decimal import Decimal
   table.put_item(Item={'pk': 'ORDER#1', 'total': Decimal('30.51')})

2. Build the Decimal from a str, not the float — Decimal(30.51) inherits the binary-float error (30.510000000000001...); Decimal(str(30.51)) gives you exactly 30.51.
3. Convert recursively for nested data — walk the dict/list and turn every float into Decimal(str(x)) before put_item. A common pattern is json.loads(json.dumps(obj), parse_float=Decimal).
4. Reading back, Decimal values come out as Decimal; convert to float/int at the edge of your app if you need native types, and configure the resource's deserializer if you want a specific numeric type.
5. For values beyond 38 digits of precision (IDs, huge integers), store them as strings instead of numbers — see number overflow.

FAQ

Why doesn't DynamoDB accept Python floats? DynamoDB numbers are arbitrary-precision decimals (up to 38 digits). Python float is IEEE-754 binary and can't represent most decimals exactly, so boto3 refuses to store a lossy approximation and raises "Float types are not supported. Use Decimal types instead."

How do I convert a float to Decimal correctly for DynamoDB? Build the Decimal from the string form: Decimal(str(value)), not Decimal(value). Decimal(30.51) carries the float's binary rounding error, while Decimal(str(30.51)) is exactly 30.51. For nested structures, use json.loads(json.dumps(obj), parse_float=Decimal).

Related errors

• Number overflow — a value beyond DynamoDB's 38-digit magnitude range.
• SerializationException — a number/string wire-type mismatch.
• Learn: DynamoDB data types