bocconcini 4.3.0__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- bocconcini-4.3.0/LICENSE +21 -0
- bocconcini-4.3.0/PKG-INFO +125 -0
- bocconcini-4.3.0/README.md +102 -0
- bocconcini-4.3.0/bocconcini.egg-info/PKG-INFO +125 -0
- bocconcini-4.3.0/bocconcini.egg-info/SOURCES.txt +9 -0
- bocconcini-4.3.0/bocconcini.egg-info/dependency_links.txt +1 -0
- bocconcini-4.3.0/bocconcini.egg-info/top_level.txt +2 -0
- bocconcini-4.3.0/bocconcini.py +2892 -0
- bocconcini-4.3.0/bocconcini_tree.py +636 -0
- bocconcini-4.3.0/pyproject.toml +33 -0
- bocconcini-4.3.0/setup.cfg +4 -0
bocconcini-4.3.0/LICENSE
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
1
|
+
MIT License
|
|
2
|
+
|
|
3
|
+
Copyright (c) 2026 Grasso Maurizio
|
|
4
|
+
|
|
5
|
+
Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
6
|
+
of this software and associated documentation files (the "Software"), to deal
|
|
7
|
+
in the Software without restriction, including without limitation the rights
|
|
8
|
+
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
9
|
+
copies of the Software, and to permit persons to whom the Software is
|
|
10
|
+
furnished to do so, subject to the following conditions:
|
|
11
|
+
|
|
12
|
+
The above copyright notice and this permission notice shall be included in all
|
|
13
|
+
copies or substantial portions of the Software.
|
|
14
|
+
|
|
15
|
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
16
|
+
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
17
|
+
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
|
18
|
+
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
19
|
+
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
20
|
+
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
|
21
|
+
SOFTWARE.
|
|
@@ -0,0 +1,125 @@
|
|
|
1
|
+
Metadata-Version: 2.4
|
|
2
|
+
Name: bocconcini
|
|
3
|
+
Version: 4.3.0
|
|
4
|
+
Summary: Local LLM client for small usable windows. Zero runtime dependencies.
|
|
5
|
+
Author: Grasso Maurizio
|
|
6
|
+
License-Expression: MIT
|
|
7
|
+
Project-URL: Homepage, https://github.com/MalgTruevero/bocconcini_crunch
|
|
8
|
+
Project-URL: Repository, https://github.com/MalgTruevero/bocconcini_crunch
|
|
9
|
+
Project-URL: Issues, https://github.com/MalgTruevero/bocconcini_crunch/issues
|
|
10
|
+
Project-URL: Benchmark, https://github.com/MalgTruevero/bocconcini_crunch/blob/main/benchmark_tank.py
|
|
11
|
+
Keywords: llm,local-llm,context-window,chunking,rag,koboldcpp,llama-cpp,termux,android,tokenizer
|
|
12
|
+
Classifier: Development Status :: 4 - Beta
|
|
13
|
+
Classifier: Intended Audience :: Developers
|
|
14
|
+
Classifier: Operating System :: OS Independent
|
|
15
|
+
Classifier: Programming Language :: Python :: 3 :: Only
|
|
16
|
+
Classifier: Topic :: Software Development :: Libraries :: Python Modules
|
|
17
|
+
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
|
|
18
|
+
Classifier: Topic :: Text Processing
|
|
19
|
+
Requires-Python: >=3.7
|
|
20
|
+
Description-Content-Type: text/markdown
|
|
21
|
+
License-File: LICENSE
|
|
22
|
+
Dynamic: license-file
|
|
23
|
+
|
|
24
|
+
# bocconcini
|
|
25
|
+
|
|
26
|
+
[](https://github.com/MalgTruevero/bocconcini_crunch/actions/workflows/tests.yml)
|
|
27
|
+
|
|
28
|
+
**Use it for:** documents and prompts on local LLMs with a genuinely
|
|
29
|
+
small context - Android/Termux, Raspberry Pi, old laptops, air-gapped machines.
|
|
30
|
+
|
|
31
|
+
**Zero runtime dependencies.** A Python client for local LLMs whose *usable*
|
|
32
|
+
window is smaller than the advertised one. Born on a Samsung S20 Ultra in
|
|
33
|
+
Termux, built to never lie about what it did.
|
|
34
|
+
|
|
35
|
+
Why "bocconcini"? A child, a huge cake, a small mouth: [STORIA.md](STORIA.md).
|
|
36
|
+
Italian edition: [README.it.md](README.it.md).
|
|
37
|
+
|
|
38
|
+
## Verification status
|
|
39
|
+
|
|
40
|
+
- **Verified live:** KoboldCpp, Samsung S20 Ultra, smol-tools-4b.
|
|
41
|
+
- **Implemented but NOT verified live:** llama.cpp native, Ollama, cloud.
|
|
42
|
+
- 43 checks: `python test_bocconcini.py` and `python test_regressioni.py`.
|
|
43
|
+
|
|
44
|
+
## What it does
|
|
45
|
+
|
|
46
|
+
- Chunks text / Python / JSON, never inside a string or an escape sequence.
|
|
47
|
+
- Output budget from the context window, not from generation speed.
|
|
48
|
+
- **Cognitive window:** `client.calibra()` measures how much the model actually
|
|
49
|
+
understands, not how much the server can hold. Only a passed probe is stored.
|
|
50
|
+
- Detects silent prompt truncation, and only near the window: far from it a
|
|
51
|
+
shortfall is our own estimate, not a cut.
|
|
52
|
+
- Continues answers cut by the token limit, stitching without duplicates.
|
|
53
|
+
- Disk-backed tree: `DocTree`, `naviga`, `albero_codice` (lossless AST partition).
|
|
54
|
+
- `trova_fatti`: deterministic scan for codes / IDs / measurements / keys,
|
|
55
|
+
each with its surrounding sentence.
|
|
56
|
+
|
|
57
|
+
Some public names are Italian (`trova_fatti`, `contesto`, `invia_adattivo`).
|
|
58
|
+
They are kept as-is for compatibility.
|
|
59
|
+
|
|
60
|
+
## Install
|
|
61
|
+
|
|
62
|
+
Python 3.7+. **No runtime dependencies.** Not on PyPI yet.
|
|
63
|
+
|
|
64
|
+
pip install git+https://github.com/MalgTruevero/bocconcini_crunch.git
|
|
65
|
+
|
|
66
|
+
Offline / Termux — no pip, no build, two files:
|
|
67
|
+
|
|
68
|
+
curl -O https://raw.githubusercontent.com/MalgTruevero/bocconcini_crunch/main/bocconcini.py
|
|
69
|
+
curl -O https://raw.githubusercontent.com/MalgTruevero/bocconcini_crunch/main/bocconcini_tree.py
|
|
70
|
+
|
|
71
|
+
## First result — no model, no server
|
|
72
|
+
|
|
73
|
+
from bocconcini_tree import trova_fatti
|
|
74
|
+
doc = "filler " * 4000 + "The unlock code is ALFA-1." + "filler " * 4000
|
|
75
|
+
print(trova_fatti(doc))
|
|
76
|
+
|
|
77
|
+
Python reads every character and returns the code with its sentence. This is the
|
|
78
|
+
mechanism that found 4/4 hidden codes in 52 s where blind compression found 0/4
|
|
79
|
+
in 42 min, on a Samsung S20 Ultra with a 4B model.
|
|
80
|
+
|
|
81
|
+
Reproduce it:
|
|
82
|
+
|
|
83
|
+
python benchmark_tank.py --no-model # scan only, no server, milliseconds
|
|
84
|
+
python benchmark_tank.py # scan + one model call
|
|
85
|
+
|
|
86
|
+
Exits 1 if a needle is missed.
|
|
87
|
+
|
|
88
|
+
Reproduce it:
|
|
89
|
+
|
|
90
|
+
python benchmark_tank.py --no-model # scan only, no server, milliseconds
|
|
91
|
+
python benchmark_tank.py # scan + one model call
|
|
92
|
+
|
|
93
|
+
Exits 1 if a needle is missed.
|
|
94
|
+
|
|
95
|
+
## With a local server
|
|
96
|
+
|
|
97
|
+
Needs a running llama.cpp / KoboldCpp server. **Verified live: KoboldCpp only.**
|
|
98
|
+
|
|
99
|
+
from bocconcini import LlamaCppClient
|
|
100
|
+
cl = LlamaCppClient("http://127.0.0.1:5001")
|
|
101
|
+
print(cl.chat("Answer with one word: ok"))
|
|
102
|
+
|
|
103
|
+
Ollama and OpenAI-compatible cloud paths are implemented but **not verified live**.
|
|
104
|
+
## Finding facts without blind compression
|
|
105
|
+
|
|
106
|
+
from bocconcini_tree import trova_fatti
|
|
107
|
+
facts = trova_fatti(document) # reads every character, no model call
|
|
108
|
+
context = "\n".join(f["contesto"] for f in facts)
|
|
109
|
+
|
|
110
|
+
Measured on the reference phone, 4 codes hidden in 32,000 characters:
|
|
111
|
+
blind compression 0/4 in 2562 s; trova_fatti 4/4 in 52 s,
|
|
112
|
+
~265 prompt tokens instead of ~10923.
|
|
113
|
+
|
|
114
|
+
`auto_reduce=True` is **lossy**: it can discard the middle of a document.
|
|
115
|
+
To find facts, use `trova_fatti` or explicit chunking.
|
|
116
|
+
|
|
117
|
+
## License
|
|
118
|
+
|
|
119
|
+
MIT. Keeping the name and the story is a courtesy, not a requirement.
|
|
120
|
+
|
|
121
|
+
## The name
|
|
122
|
+
|
|
123
|
+
"BOCCONCINI" reads like mozzarella to an English speaker. It is not: it means
|
|
124
|
+
small bites. If you fork and republish, please carry STORIA.md along — a
|
|
125
|
+
courtesy, not a licence requirement.
|
|
@@ -0,0 +1,102 @@
|
|
|
1
|
+
# bocconcini
|
|
2
|
+
|
|
3
|
+
[](https://github.com/MalgTruevero/bocconcini_crunch/actions/workflows/tests.yml)
|
|
4
|
+
|
|
5
|
+
**Use it for:** documents and prompts on local LLMs with a genuinely
|
|
6
|
+
small context - Android/Termux, Raspberry Pi, old laptops, air-gapped machines.
|
|
7
|
+
|
|
8
|
+
**Zero runtime dependencies.** A Python client for local LLMs whose *usable*
|
|
9
|
+
window is smaller than the advertised one. Born on a Samsung S20 Ultra in
|
|
10
|
+
Termux, built to never lie about what it did.
|
|
11
|
+
|
|
12
|
+
Why "bocconcini"? A child, a huge cake, a small mouth: [STORIA.md](STORIA.md).
|
|
13
|
+
Italian edition: [README.it.md](README.it.md).
|
|
14
|
+
|
|
15
|
+
## Verification status
|
|
16
|
+
|
|
17
|
+
- **Verified live:** KoboldCpp, Samsung S20 Ultra, smol-tools-4b.
|
|
18
|
+
- **Implemented but NOT verified live:** llama.cpp native, Ollama, cloud.
|
|
19
|
+
- 43 checks: `python test_bocconcini.py` and `python test_regressioni.py`.
|
|
20
|
+
|
|
21
|
+
## What it does
|
|
22
|
+
|
|
23
|
+
- Chunks text / Python / JSON, never inside a string or an escape sequence.
|
|
24
|
+
- Output budget from the context window, not from generation speed.
|
|
25
|
+
- **Cognitive window:** `client.calibra()` measures how much the model actually
|
|
26
|
+
understands, not how much the server can hold. Only a passed probe is stored.
|
|
27
|
+
- Detects silent prompt truncation, and only near the window: far from it a
|
|
28
|
+
shortfall is our own estimate, not a cut.
|
|
29
|
+
- Continues answers cut by the token limit, stitching without duplicates.
|
|
30
|
+
- Disk-backed tree: `DocTree`, `naviga`, `albero_codice` (lossless AST partition).
|
|
31
|
+
- `trova_fatti`: deterministic scan for codes / IDs / measurements / keys,
|
|
32
|
+
each with its surrounding sentence.
|
|
33
|
+
|
|
34
|
+
Some public names are Italian (`trova_fatti`, `contesto`, `invia_adattivo`).
|
|
35
|
+
They are kept as-is for compatibility.
|
|
36
|
+
|
|
37
|
+
## Install
|
|
38
|
+
|
|
39
|
+
Python 3.7+. **No runtime dependencies.** Not on PyPI yet.
|
|
40
|
+
|
|
41
|
+
pip install git+https://github.com/MalgTruevero/bocconcini_crunch.git
|
|
42
|
+
|
|
43
|
+
Offline / Termux — no pip, no build, two files:
|
|
44
|
+
|
|
45
|
+
curl -O https://raw.githubusercontent.com/MalgTruevero/bocconcini_crunch/main/bocconcini.py
|
|
46
|
+
curl -O https://raw.githubusercontent.com/MalgTruevero/bocconcini_crunch/main/bocconcini_tree.py
|
|
47
|
+
|
|
48
|
+
## First result — no model, no server
|
|
49
|
+
|
|
50
|
+
from bocconcini_tree import trova_fatti
|
|
51
|
+
doc = "filler " * 4000 + "The unlock code is ALFA-1." + "filler " * 4000
|
|
52
|
+
print(trova_fatti(doc))
|
|
53
|
+
|
|
54
|
+
Python reads every character and returns the code with its sentence. This is the
|
|
55
|
+
mechanism that found 4/4 hidden codes in 52 s where blind compression found 0/4
|
|
56
|
+
in 42 min, on a Samsung S20 Ultra with a 4B model.
|
|
57
|
+
|
|
58
|
+
Reproduce it:
|
|
59
|
+
|
|
60
|
+
python benchmark_tank.py --no-model # scan only, no server, milliseconds
|
|
61
|
+
python benchmark_tank.py # scan + one model call
|
|
62
|
+
|
|
63
|
+
Exits 1 if a needle is missed.
|
|
64
|
+
|
|
65
|
+
Reproduce it:
|
|
66
|
+
|
|
67
|
+
python benchmark_tank.py --no-model # scan only, no server, milliseconds
|
|
68
|
+
python benchmark_tank.py # scan + one model call
|
|
69
|
+
|
|
70
|
+
Exits 1 if a needle is missed.
|
|
71
|
+
|
|
72
|
+
## With a local server
|
|
73
|
+
|
|
74
|
+
Needs a running llama.cpp / KoboldCpp server. **Verified live: KoboldCpp only.**
|
|
75
|
+
|
|
76
|
+
from bocconcini import LlamaCppClient
|
|
77
|
+
cl = LlamaCppClient("http://127.0.0.1:5001")
|
|
78
|
+
print(cl.chat("Answer with one word: ok"))
|
|
79
|
+
|
|
80
|
+
Ollama and OpenAI-compatible cloud paths are implemented but **not verified live**.
|
|
81
|
+
## Finding facts without blind compression
|
|
82
|
+
|
|
83
|
+
from bocconcini_tree import trova_fatti
|
|
84
|
+
facts = trova_fatti(document) # reads every character, no model call
|
|
85
|
+
context = "\n".join(f["contesto"] for f in facts)
|
|
86
|
+
|
|
87
|
+
Measured on the reference phone, 4 codes hidden in 32,000 characters:
|
|
88
|
+
blind compression 0/4 in 2562 s; trova_fatti 4/4 in 52 s,
|
|
89
|
+
~265 prompt tokens instead of ~10923.
|
|
90
|
+
|
|
91
|
+
`auto_reduce=True` is **lossy**: it can discard the middle of a document.
|
|
92
|
+
To find facts, use `trova_fatti` or explicit chunking.
|
|
93
|
+
|
|
94
|
+
## License
|
|
95
|
+
|
|
96
|
+
MIT. Keeping the name and the story is a courtesy, not a requirement.
|
|
97
|
+
|
|
98
|
+
## The name
|
|
99
|
+
|
|
100
|
+
"BOCCONCINI" reads like mozzarella to an English speaker. It is not: it means
|
|
101
|
+
small bites. If you fork and republish, please carry STORIA.md along — a
|
|
102
|
+
courtesy, not a licence requirement.
|
|
@@ -0,0 +1,125 @@
|
|
|
1
|
+
Metadata-Version: 2.4
|
|
2
|
+
Name: bocconcini
|
|
3
|
+
Version: 4.3.0
|
|
4
|
+
Summary: Local LLM client for small usable windows. Zero runtime dependencies.
|
|
5
|
+
Author: Grasso Maurizio
|
|
6
|
+
License-Expression: MIT
|
|
7
|
+
Project-URL: Homepage, https://github.com/MalgTruevero/bocconcini_crunch
|
|
8
|
+
Project-URL: Repository, https://github.com/MalgTruevero/bocconcini_crunch
|
|
9
|
+
Project-URL: Issues, https://github.com/MalgTruevero/bocconcini_crunch/issues
|
|
10
|
+
Project-URL: Benchmark, https://github.com/MalgTruevero/bocconcini_crunch/blob/main/benchmark_tank.py
|
|
11
|
+
Keywords: llm,local-llm,context-window,chunking,rag,koboldcpp,llama-cpp,termux,android,tokenizer
|
|
12
|
+
Classifier: Development Status :: 4 - Beta
|
|
13
|
+
Classifier: Intended Audience :: Developers
|
|
14
|
+
Classifier: Operating System :: OS Independent
|
|
15
|
+
Classifier: Programming Language :: Python :: 3 :: Only
|
|
16
|
+
Classifier: Topic :: Software Development :: Libraries :: Python Modules
|
|
17
|
+
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
|
|
18
|
+
Classifier: Topic :: Text Processing
|
|
19
|
+
Requires-Python: >=3.7
|
|
20
|
+
Description-Content-Type: text/markdown
|
|
21
|
+
License-File: LICENSE
|
|
22
|
+
Dynamic: license-file
|
|
23
|
+
|
|
24
|
+
# bocconcini
|
|
25
|
+
|
|
26
|
+
[](https://github.com/MalgTruevero/bocconcini_crunch/actions/workflows/tests.yml)
|
|
27
|
+
|
|
28
|
+
**Use it for:** documents and prompts on local LLMs with a genuinely
|
|
29
|
+
small context - Android/Termux, Raspberry Pi, old laptops, air-gapped machines.
|
|
30
|
+
|
|
31
|
+
**Zero runtime dependencies.** A Python client for local LLMs whose *usable*
|
|
32
|
+
window is smaller than the advertised one. Born on a Samsung S20 Ultra in
|
|
33
|
+
Termux, built to never lie about what it did.
|
|
34
|
+
|
|
35
|
+
Why "bocconcini"? A child, a huge cake, a small mouth: [STORIA.md](STORIA.md).
|
|
36
|
+
Italian edition: [README.it.md](README.it.md).
|
|
37
|
+
|
|
38
|
+
## Verification status
|
|
39
|
+
|
|
40
|
+
- **Verified live:** KoboldCpp, Samsung S20 Ultra, smol-tools-4b.
|
|
41
|
+
- **Implemented but NOT verified live:** llama.cpp native, Ollama, cloud.
|
|
42
|
+
- 43 checks: `python test_bocconcini.py` and `python test_regressioni.py`.
|
|
43
|
+
|
|
44
|
+
## What it does
|
|
45
|
+
|
|
46
|
+
- Chunks text / Python / JSON, never inside a string or an escape sequence.
|
|
47
|
+
- Output budget from the context window, not from generation speed.
|
|
48
|
+
- **Cognitive window:** `client.calibra()` measures how much the model actually
|
|
49
|
+
understands, not how much the server can hold. Only a passed probe is stored.
|
|
50
|
+
- Detects silent prompt truncation, and only near the window: far from it a
|
|
51
|
+
shortfall is our own estimate, not a cut.
|
|
52
|
+
- Continues answers cut by the token limit, stitching without duplicates.
|
|
53
|
+
- Disk-backed tree: `DocTree`, `naviga`, `albero_codice` (lossless AST partition).
|
|
54
|
+
- `trova_fatti`: deterministic scan for codes / IDs / measurements / keys,
|
|
55
|
+
each with its surrounding sentence.
|
|
56
|
+
|
|
57
|
+
Some public names are Italian (`trova_fatti`, `contesto`, `invia_adattivo`).
|
|
58
|
+
They are kept as-is for compatibility.
|
|
59
|
+
|
|
60
|
+
## Install
|
|
61
|
+
|
|
62
|
+
Python 3.7+. **No runtime dependencies.** Not on PyPI yet.
|
|
63
|
+
|
|
64
|
+
pip install git+https://github.com/MalgTruevero/bocconcini_crunch.git
|
|
65
|
+
|
|
66
|
+
Offline / Termux — no pip, no build, two files:
|
|
67
|
+
|
|
68
|
+
curl -O https://raw.githubusercontent.com/MalgTruevero/bocconcini_crunch/main/bocconcini.py
|
|
69
|
+
curl -O https://raw.githubusercontent.com/MalgTruevero/bocconcini_crunch/main/bocconcini_tree.py
|
|
70
|
+
|
|
71
|
+
## First result — no model, no server
|
|
72
|
+
|
|
73
|
+
from bocconcini_tree import trova_fatti
|
|
74
|
+
doc = "filler " * 4000 + "The unlock code is ALFA-1." + "filler " * 4000
|
|
75
|
+
print(trova_fatti(doc))
|
|
76
|
+
|
|
77
|
+
Python reads every character and returns the code with its sentence. This is the
|
|
78
|
+
mechanism that found 4/4 hidden codes in 52 s where blind compression found 0/4
|
|
79
|
+
in 42 min, on a Samsung S20 Ultra with a 4B model.
|
|
80
|
+
|
|
81
|
+
Reproduce it:
|
|
82
|
+
|
|
83
|
+
python benchmark_tank.py --no-model # scan only, no server, milliseconds
|
|
84
|
+
python benchmark_tank.py # scan + one model call
|
|
85
|
+
|
|
86
|
+
Exits 1 if a needle is missed.
|
|
87
|
+
|
|
88
|
+
Reproduce it:
|
|
89
|
+
|
|
90
|
+
python benchmark_tank.py --no-model # scan only, no server, milliseconds
|
|
91
|
+
python benchmark_tank.py # scan + one model call
|
|
92
|
+
|
|
93
|
+
Exits 1 if a needle is missed.
|
|
94
|
+
|
|
95
|
+
## With a local server
|
|
96
|
+
|
|
97
|
+
Needs a running llama.cpp / KoboldCpp server. **Verified live: KoboldCpp only.**
|
|
98
|
+
|
|
99
|
+
from bocconcini import LlamaCppClient
|
|
100
|
+
cl = LlamaCppClient("http://127.0.0.1:5001")
|
|
101
|
+
print(cl.chat("Answer with one word: ok"))
|
|
102
|
+
|
|
103
|
+
Ollama and OpenAI-compatible cloud paths are implemented but **not verified live**.
|
|
104
|
+
## Finding facts without blind compression
|
|
105
|
+
|
|
106
|
+
from bocconcini_tree import trova_fatti
|
|
107
|
+
facts = trova_fatti(document) # reads every character, no model call
|
|
108
|
+
context = "\n".join(f["contesto"] for f in facts)
|
|
109
|
+
|
|
110
|
+
Measured on the reference phone, 4 codes hidden in 32,000 characters:
|
|
111
|
+
blind compression 0/4 in 2562 s; trova_fatti 4/4 in 52 s,
|
|
112
|
+
~265 prompt tokens instead of ~10923.
|
|
113
|
+
|
|
114
|
+
`auto_reduce=True` is **lossy**: it can discard the middle of a document.
|
|
115
|
+
To find facts, use `trova_fatti` or explicit chunking.
|
|
116
|
+
|
|
117
|
+
## License
|
|
118
|
+
|
|
119
|
+
MIT. Keeping the name and the story is a courtesy, not a requirement.
|
|
120
|
+
|
|
121
|
+
## The name
|
|
122
|
+
|
|
123
|
+
"BOCCONCINI" reads like mozzarella to an English speaker. It is not: it means
|
|
124
|
+
small bites. If you fork and republish, please carry STORIA.md along — a
|
|
125
|
+
courtesy, not a licence requirement.
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
|