simplicio-cli 0.2.0__tar.gz → 0.2.2__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.
- {simplicio_cli-0.2.0/simplicio_cli.egg-info → simplicio_cli-0.2.2}/PKG-INFO +25 -8
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/README.md +23 -6
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/pyproject.toml +2 -2
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2/simplicio_cli.egg-info}/PKG-INFO +25 -8
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/LICENSE +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/setup.cfg +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/simplicio/__init__.py +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/simplicio/bench.py +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/simplicio/cache.py +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/simplicio/cli.py +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/simplicio/pipeline.py +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/simplicio/precedent.py +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/simplicio/prompt.py +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/simplicio/providers.py +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/simplicio/skill_router.py +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/simplicio/templates/simplicio_prompt.md +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/simplicio_cli.egg-info/SOURCES.txt +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/simplicio_cli.egg-info/dependency_links.txt +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/simplicio_cli.egg-info/entry_points.txt +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/simplicio_cli.egg-info/requires.txt +0 -0
- {simplicio_cli-0.2.0 → simplicio_cli-0.2.2}/simplicio_cli.egg-info/top_level.txt +0 -0
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: simplicio-cli
|
|
3
|
-
Version: 0.2.
|
|
4
|
-
Summary: Portable task-to-code pipeline that works with any LLM. Turn a one-line task into a verified code change — diff + test + verify loop. +
|
|
3
|
+
Version: 0.2.2
|
|
4
|
+
Summary: Portable task-to-code pipeline that works with any LLM. Turn a one-line task into a verified code change — diff + test + verify loop. +55 pts on a 156-check benchmark, 21% faster, ~same tokens.
|
|
5
5
|
Author-email: Wesley Simplicio <wesleybob4@gmail.com>
|
|
6
6
|
License: MIT
|
|
7
7
|
Project-URL: Homepage, https://github.com/wesleysimplicio/simplicio-cli
|
|
@@ -41,6 +41,8 @@ Dynamic: license-file
|
|
|
41
41
|
[](https://pypi.org/project/simplicio-cli/)
|
|
42
42
|
[](LICENSE)
|
|
43
43
|
|
|
44
|
+
[](output/imagegen/simplicio-cli-readme-hero.png)
|
|
45
|
+
|
|
44
46
|
> *"hide the Delete button for non-admins"* → diff + test + applied + verified.
|
|
45
47
|
> Works with **OpenRouter, OpenAI, Anthropic, GLM, DeepSeek, Ollama** — one env var.
|
|
46
48
|
|
|
@@ -56,18 +58,33 @@ Same model. Same task. Only the prompt changes. **Measured, reproducible, determ
|
|
|
56
58
|
|
|
57
59
|
| Model | Without simplicio | With simplicio | Gain |
|
|
58
60
|
|---|---|---|---|
|
|
59
|
-
| **
|
|
60
|
-
| **
|
|
61
|
-
| **Qwen 2.5 7B Instruct** |
|
|
62
|
-
| **Average across 3 models · 10 cases · 156 checks** | **
|
|
61
|
+
| **Gemma 3 12B IT** | 34% | **92%** | **+58 pts** |
|
|
62
|
+
| **Llama 3.1 8B Instruct** | 36% | **90%** | **+54 pts** |
|
|
63
|
+
| **Qwen 2.5 7B Instruct** | 34% | **88%** | **+54 pts** |
|
|
64
|
+
| **Average across 3 models · 10 cases · 156 checks** | **35%** | **90%** | **+55 pts (+156%)** |
|
|
63
65
|
|
|
64
66
|
### Output-quality signals (rate across all 30 runs)
|
|
65
67
|
|
|
66
68
|
| Signal | Raw prompt | With simplicio |
|
|
67
69
|
|---|---|---|
|
|
68
70
|
| **DIFF block present** | 0% | **100%** |
|
|
69
|
-
| Target file mentioned |
|
|
70
|
-
| TEST block present |
|
|
71
|
+
| Target file mentioned | 0% | **96%** |
|
|
72
|
+
| TEST block present | 80% | **96%** |
|
|
73
|
+
|
|
74
|
+
### Cost — tokens & wall-clock (measured, not estimated)
|
|
75
|
+
|
|
76
|
+
Same provider, same models, same cases. Token counts pulled from the API
|
|
77
|
+
`usage` field; latency from `time.perf_counter()` around each call.
|
|
78
|
+
|
|
79
|
+
| Side | Tokens / run | Wall-clock / run | Total tokens (30 runs) | Total time |
|
|
80
|
+
|---|---|---|---|---|
|
|
81
|
+
| Raw prompt | 759 | 12.4s | 22,774 | 6m 13s |
|
|
82
|
+
| With simplicio | **770** | **9.9s** | **23,127** | **4m 58s** |
|
|
83
|
+
| Δ | **+1%** | **−21%** | +353 | **−75s** |
|
|
84
|
+
|
|
85
|
+
simplicio wraps the objective in a 6-layer contract — more input tokens up
|
|
86
|
+
front, fewer completion tokens because the model stops guessing. Net effect
|
|
87
|
+
across 30 runs: roughly the same token bill, **21% faster, 90% pass-rate**.
|
|
71
88
|
|
|
72
89
|
> A 7B-parameter open model wrapped in simplicio's 6-layer contract outperforms
|
|
73
90
|
> the same model with raw prompting **by 42 to 64 points**. Without changing the
|
|
@@ -6,6 +6,8 @@
|
|
|
6
6
|
[](https://pypi.org/project/simplicio-cli/)
|
|
7
7
|
[](LICENSE)
|
|
8
8
|
|
|
9
|
+
[](output/imagegen/simplicio-cli-readme-hero.png)
|
|
10
|
+
|
|
9
11
|
> *"hide the Delete button for non-admins"* → diff + test + applied + verified.
|
|
10
12
|
> Works with **OpenRouter, OpenAI, Anthropic, GLM, DeepSeek, Ollama** — one env var.
|
|
11
13
|
|
|
@@ -21,18 +23,33 @@ Same model. Same task. Only the prompt changes. **Measured, reproducible, determ
|
|
|
21
23
|
|
|
22
24
|
| Model | Without simplicio | With simplicio | Gain |
|
|
23
25
|
|---|---|---|---|
|
|
24
|
-
| **
|
|
25
|
-
| **
|
|
26
|
-
| **Qwen 2.5 7B Instruct** |
|
|
27
|
-
| **Average across 3 models · 10 cases · 156 checks** | **
|
|
26
|
+
| **Gemma 3 12B IT** | 34% | **92%** | **+58 pts** |
|
|
27
|
+
| **Llama 3.1 8B Instruct** | 36% | **90%** | **+54 pts** |
|
|
28
|
+
| **Qwen 2.5 7B Instruct** | 34% | **88%** | **+54 pts** |
|
|
29
|
+
| **Average across 3 models · 10 cases · 156 checks** | **35%** | **90%** | **+55 pts (+156%)** |
|
|
28
30
|
|
|
29
31
|
### Output-quality signals (rate across all 30 runs)
|
|
30
32
|
|
|
31
33
|
| Signal | Raw prompt | With simplicio |
|
|
32
34
|
|---|---|---|
|
|
33
35
|
| **DIFF block present** | 0% | **100%** |
|
|
34
|
-
| Target file mentioned |
|
|
35
|
-
| TEST block present |
|
|
36
|
+
| Target file mentioned | 0% | **96%** |
|
|
37
|
+
| TEST block present | 80% | **96%** |
|
|
38
|
+
|
|
39
|
+
### Cost — tokens & wall-clock (measured, not estimated)
|
|
40
|
+
|
|
41
|
+
Same provider, same models, same cases. Token counts pulled from the API
|
|
42
|
+
`usage` field; latency from `time.perf_counter()` around each call.
|
|
43
|
+
|
|
44
|
+
| Side | Tokens / run | Wall-clock / run | Total tokens (30 runs) | Total time |
|
|
45
|
+
|---|---|---|---|---|
|
|
46
|
+
| Raw prompt | 759 | 12.4s | 22,774 | 6m 13s |
|
|
47
|
+
| With simplicio | **770** | **9.9s** | **23,127** | **4m 58s** |
|
|
48
|
+
| Δ | **+1%** | **−21%** | +353 | **−75s** |
|
|
49
|
+
|
|
50
|
+
simplicio wraps the objective in a 6-layer contract — more input tokens up
|
|
51
|
+
front, fewer completion tokens because the model stops guessing. Net effect
|
|
52
|
+
across 30 runs: roughly the same token bill, **21% faster, 90% pass-rate**.
|
|
36
53
|
|
|
37
54
|
> A 7B-parameter open model wrapped in simplicio's 6-layer contract outperforms
|
|
38
55
|
> the same model with raw prompting **by 42 to 64 points**. Without changing the
|
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
[project]
|
|
2
2
|
name = "simplicio-cli"
|
|
3
|
-
version = "0.2.
|
|
4
|
-
description = "Portable task-to-code pipeline that works with any LLM. Turn a one-line task into a verified code change — diff + test + verify loop. +
|
|
3
|
+
version = "0.2.2"
|
|
4
|
+
description = "Portable task-to-code pipeline that works with any LLM. Turn a one-line task into a verified code change — diff + test + verify loop. +55 pts on a 156-check benchmark, 21% faster, ~same tokens."
|
|
5
5
|
readme = "README.md"
|
|
6
6
|
license = { text = "MIT" }
|
|
7
7
|
requires-python = ">=3.9"
|
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: simplicio-cli
|
|
3
|
-
Version: 0.2.
|
|
4
|
-
Summary: Portable task-to-code pipeline that works with any LLM. Turn a one-line task into a verified code change — diff + test + verify loop. +
|
|
3
|
+
Version: 0.2.2
|
|
4
|
+
Summary: Portable task-to-code pipeline that works with any LLM. Turn a one-line task into a verified code change — diff + test + verify loop. +55 pts on a 156-check benchmark, 21% faster, ~same tokens.
|
|
5
5
|
Author-email: Wesley Simplicio <wesleybob4@gmail.com>
|
|
6
6
|
License: MIT
|
|
7
7
|
Project-URL: Homepage, https://github.com/wesleysimplicio/simplicio-cli
|
|
@@ -41,6 +41,8 @@ Dynamic: license-file
|
|
|
41
41
|
[](https://pypi.org/project/simplicio-cli/)
|
|
42
42
|
[](LICENSE)
|
|
43
43
|
|
|
44
|
+
[](output/imagegen/simplicio-cli-readme-hero.png)
|
|
45
|
+
|
|
44
46
|
> *"hide the Delete button for non-admins"* → diff + test + applied + verified.
|
|
45
47
|
> Works with **OpenRouter, OpenAI, Anthropic, GLM, DeepSeek, Ollama** — one env var.
|
|
46
48
|
|
|
@@ -56,18 +58,33 @@ Same model. Same task. Only the prompt changes. **Measured, reproducible, determ
|
|
|
56
58
|
|
|
57
59
|
| Model | Without simplicio | With simplicio | Gain |
|
|
58
60
|
|---|---|---|---|
|
|
59
|
-
| **
|
|
60
|
-
| **
|
|
61
|
-
| **Qwen 2.5 7B Instruct** |
|
|
62
|
-
| **Average across 3 models · 10 cases · 156 checks** | **
|
|
61
|
+
| **Gemma 3 12B IT** | 34% | **92%** | **+58 pts** |
|
|
62
|
+
| **Llama 3.1 8B Instruct** | 36% | **90%** | **+54 pts** |
|
|
63
|
+
| **Qwen 2.5 7B Instruct** | 34% | **88%** | **+54 pts** |
|
|
64
|
+
| **Average across 3 models · 10 cases · 156 checks** | **35%** | **90%** | **+55 pts (+156%)** |
|
|
63
65
|
|
|
64
66
|
### Output-quality signals (rate across all 30 runs)
|
|
65
67
|
|
|
66
68
|
| Signal | Raw prompt | With simplicio |
|
|
67
69
|
|---|---|---|
|
|
68
70
|
| **DIFF block present** | 0% | **100%** |
|
|
69
|
-
| Target file mentioned |
|
|
70
|
-
| TEST block present |
|
|
71
|
+
| Target file mentioned | 0% | **96%** |
|
|
72
|
+
| TEST block present | 80% | **96%** |
|
|
73
|
+
|
|
74
|
+
### Cost — tokens & wall-clock (measured, not estimated)
|
|
75
|
+
|
|
76
|
+
Same provider, same models, same cases. Token counts pulled from the API
|
|
77
|
+
`usage` field; latency from `time.perf_counter()` around each call.
|
|
78
|
+
|
|
79
|
+
| Side | Tokens / run | Wall-clock / run | Total tokens (30 runs) | Total time |
|
|
80
|
+
|---|---|---|---|---|
|
|
81
|
+
| Raw prompt | 759 | 12.4s | 22,774 | 6m 13s |
|
|
82
|
+
| With simplicio | **770** | **9.9s** | **23,127** | **4m 58s** |
|
|
83
|
+
| Δ | **+1%** | **−21%** | +353 | **−75s** |
|
|
84
|
+
|
|
85
|
+
simplicio wraps the objective in a 6-layer contract — more input tokens up
|
|
86
|
+
front, fewer completion tokens because the model stops guessing. Net effect
|
|
87
|
+
across 30 runs: roughly the same token bill, **21% faster, 90% pass-rate**.
|
|
71
88
|
|
|
72
89
|
> A 7B-parameter open model wrapped in simplicio's 6-layer contract outperforms
|
|
73
90
|
> the same model with raw prompting **by 42 to 64 points**. Without changing the
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|