This project has been archived by its maintainers, and is no longer receiving any updates.
ARKADIA AI.DATA-FORMAT
:SB@MMWBZSr,
,irXX7ri, iaBWMMMMM@a: :,
i2WMMMMMMMMMMMM@a:7ZZWMMMMM@7 :i,02,
,7S7;i::i;72Z2S0@ZZWMW;r7iBWWM@M2 rM0BM;
ia@@08Z0B2: :,i0r rBZ,: XSBMZM: 2MMMMi
iBZ7i:, iXZZSi i; ia ;:@22: ;M@MMS
,: ,, ,7: ,Xi ;r, : ,SS, rBZ8MiX7,
r8WZ: XW@0Z8WMZ: rX: :M2;BZ ;r: ;, ::Si;ii :7
:8MM@a,:@0: XM2 , i@WZMZ 28MMMBr , :ZW@a iB@;
,Z0MMWZi BB ,aZ8i S7,, ;SXZMMMMMMMa::r , 7M8,
:;MMZ: 8Z :07Z7 @X XaSa,:8MMMM2ZMMMMMMM@MS i2 ,BX :WW:
iM0: ,X,i8r XM:,S7a; XZWMMMM2,8MMMMMWr8M2, iS, ,B@:
87 iB0:,,,XZX, XW S@ ,;aZ00ZZ2SX7,:S8082,iM00, :i ar ,WW,
: :0MB@Zi,, iM2 Sr :;Xriirr: ZM8 ,XBM@X : ,: i@a
rMMMW0X: 8Z : ;a00Z7,,2BMMWBZXi :, iaWMMMMB7 :; 2Wi
rMMBr ZW ,i rW7a00Z702 78@MMMMMMMM0@MMM@@@MMMMMM0r iWS
:MZ, ,7: ;M2;Zii,:MrB, 2S07 ZZ8BZ,XB0WMMMMMMMMW280BWWWW8i 8Z
S: ,ZZ XMZW:rr, Z;B: ,ZXSi S2222XXX7ZBZ;:BMMM@X: irr: ZZ
;MZ; rMMX : ,27XZZS;S; ii ,i,ZMMMMMMMMM@7 ,8Z aa
;MM0: , BMi iS ,irr;:;, :7,,Wa S8Z,,aXMZWBBMMMMMMS:MM8 ,ZX
,WMM0,2 :@, 0rr ;@Br2,7S ::r7,ir;rr8ZZMMMMMMSiMMa ;ai
rMMM8a7 ,: @0: ,: , , ,Z222X77@MW02i iZW: ,SX
X@SM0@: BB ;r :S ,;7XS77; ,XZZZZ82r, :;i ,, ;X:
;riMMM: , r , ;0 ,:7Z ;SS ; , ,: iZ@MMMMM2
8MM; ; Z ;i ;, XrMi B8 X7 , ,7 ,XZWMMMMMM@B
,WM2 7;87:8 XXMW, ,Bi ;7@i i Z ,;2aa8MWWMMMM02
,B@, :ZaMiBX :ZBMWr, ; XB@iiXX2:: Xa7;80Z@0a :
X8 r8WMaMr iZ0MM0, ,X@;;ZWX8 :S; 7:
,i rBWMMMr :X;WM@r ir:0@W0
,0@MMM2 : ;BMWi 7@MW:
:0MMMWi ;aBS: i8MZ:
XBMMW; ,;r,
iS0@8;
The High-Density, Token-Efficient Data Protocol for Large Language Models. Stop wasting context window on JSON braces.
AI.DATAis a unified, schema-first data format designed specifically for AI understanding. It offers up to 25% token savings, faster parsing, and human-readable structure that LLMs love.
📦 Installation
Get started immediately with pip:
pip install arkadia-ai-data
🚀 Fast Example
Encoding to AI.DATA:
echo '{ "data": 2}' | aid enc - -c
# Output: <data:number>(2)
Decoding back to JSON:
echo '<data:number>(2)' | aid dec - -f json
# Output: { "data": 2 }
⚡ Performance & Token Savings
Why switch? Because every token counts. AICD (Arkadia Compressed Data) consistently outperforms standard formats in both token efficiency.
BENCHMARK SUMMARY:
JSON █████████████████████░░░░ 6921 tok ░░░░░░░░░░░░░░░░░░░░░░░░░ 0.15 ms
AICD ████████████████░░░░░░░░░ 5416 tok █████████████████████████ 4.40 ms
AID ███████████████████░░░░░░ 6488 tok ████████████████████████░ 4.29 ms
TOON █████████████████████████ 8198 tok █████████████░░░░░░░░░░░░ 2.36 ms
FORMAT TOKENS TIME (Total) AVG TIME/FILE VS JSON
----------------------------------------------------------------------
AICD 5416 4.40 ms 0.37 ms -21.7%
AID 6488 4.29 ms 0.36 ms -6.3%
JSON 6921 0.15 ms 0.01 ms +0.0%
TOON 8198 2.36 ms 0.20 ms +18.5%
CONCLUSION: Switching to AICD saves 1505 tokens (21.7%) compared to JSON.
🛠 CLI Usage
The package comes with a powerful CLI tool aid for encoding, decoding, and benchmarking.
Arkadia AI DATA TOOL
--------------------------------------------------
Unified interface for AI Data Format operations.
USAGE:
aid <command> [flags]
COMMANDS:
enc [ENCODE] Convert JSON/YAML/TOON to AI.Data format
dec [DECODE] Parse AI.Data format back to JSON
benchmark [BENCHMARK] Run performance and token usage tests
ai-benchmark [AI] Run AI understanding tests (not implemented yet)
GLOBAL OPTIONS:
-h, --help Show this help message
-v, --version Show version info
📖 Syntax Specification (Current Version)
This section describes the actual, currently implemented syntax of AI.DATA-FORMAT.
1. Type Definition
A type defines a name and an ordered list of fields. Comments are allowed within the definition to assist the LLM.
User</comment/ ={(23,"A",3) #tag1 #tag2} %[{ id: 4, b: "a", c: 43}]: id:number,
b: string , c:number, >
@Users
<
@list
a: number,
b: string
>
[
@size=5
/example list of values/
(1,`text`,5)
(2,`Text can be
multiline
`,5)
{
id:3,
b: "text"
}
]
Key Rules:
- The type name (
@Name) is optional but recommended. - The header
<...>defines field names and their order. - Comments (
/ ... /) are allowed in the header.
2. Data Structures
The format supports compact positional records and explicit named records.
| Structure | Syntax | Description |
|---|---|---|
| Positional Record | (a,b,c) |
Must follow the exact order of fields in the type header. |
| Named Record | {key:value} |
Keys must match field names. No spaces allowed in keys/values. |
| List | [ ... ] |
Contains positional or named records. |
| Multiline Text | text |
Ends with a line containing only a backtick. |
3. Comments
/ this is a comment /
- Allowed only inside type definitions.
- Forbidden in raw data blocks to save space.
4. General Rules
- Data must contain NO spaces. (Compactness is priority).
- Schema/Type definitions may contain spaces and comments.
- Named fields always use
key:valuewithout spaces. - Positional order must exactly match the declared order.
5. Inline Type Usage
You can declare a type and immediately use it:
@User<id:number name:string desc:string>
value:@User(2,"Alice","Hello")
value2:@User(3,"Bob","World")
6. Nested Types
Currently, nested types are allowed as structural definitions:
@User<
id:string
name:string
profile: < level:number, score:number >
>
[
("u1","Aga",{level:5,score:82})
("u2","Marek",{level:7,score:91})
]
🔮 Futures / Roadmap
The following features are planned for future releases and are not yet implemented.
-
Modifiers:
-
!required- field must be included. -
?empty- field must not be empty. -
=value- default value. -
N..M- numeric range validation. -
Binary Data Types:
-
Hex:
~[hex]1A0F4F~ -
Base64:
~[b64]ADFKDXKZK...~ -
Pointers/References:
-
Reference existing objects by ID:
(1, "Alex", *User[2])
📄 License
This project is licensed under the [MIT License].
Metadata
Release files for arkadia-ai-data 0.1.3
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| arkadia_ai_data-0.1.3.tar.gz | 112.4 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| arkadia_ai_data-0.1.3-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 163.5 kB
Release files / arkadia_ai_data-0.1.3.tar.gz
| Download URL | arkadia_ai_data-0.1.3.tar.gz |
|---|---|
| Size | 112.4 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
24316e205742d4556f872dacdde4aa1abed31c4d81f04a9801cea6774a754558
|
|
BLAKE2b-256 checksum How to use checksums |
61d1f3ad1ad5fe03fe84b0454954418650151d9b08d9a91c3b9ddc0737e261e4
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.10.0
|
Release files / arkadia_ai_data-0.1.3-py3-none-any.whl
| Download URL | arkadia_ai_data-0.1.3-py3-none-any.whl |
|---|---|
| Size | 51.1 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
b1ea07e6caf03e7bd0d6d09cda6c5cfc198f369ca92cb357094dd6f23c231d14
|
|
BLAKE2b-256 checksum How to use checksums |
9cbc9514d9bfdc671bd10d21ee33f6c281b8f8d74f28d8342bef15a7fda3baab
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.10.0
|