buildlog 0.8.0__py3-none-any.whl → 0.10.0__py3-none-any.whl
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.
- buildlog/cli.py +491 -30
- buildlog/constants.py +121 -0
- buildlog/core/__init__.py +44 -0
- buildlog/core/operations.py +1189 -13
- buildlog/data/seeds/bragi.yaml +61 -0
- buildlog/llm.py +51 -4
- buildlog/mcp/__init__.py +51 -3
- buildlog/mcp/server.py +40 -0
- buildlog/mcp/tools.py +526 -12
- buildlog/seed_engine/__init__.py +2 -0
- buildlog/seed_engine/llm_extractor.py +121 -0
- buildlog/seed_engine/pipeline.py +45 -1
- {buildlog-0.8.0.data → buildlog-0.10.0.data}/data/share/buildlog/post_gen.py +10 -5
- buildlog-0.10.0.data/data/share/buildlog/template/buildlog/.gitkeep +0 -0
- buildlog-0.10.0.data/data/share/buildlog/template/buildlog/assets/.gitkeep +0 -0
- buildlog-0.10.0.dist-info/METADATA +248 -0
- {buildlog-0.8.0.dist-info → buildlog-0.10.0.dist-info}/RECORD +27 -22
- buildlog-0.8.0.dist-info/METADATA +0 -151
- {buildlog-0.8.0.data → buildlog-0.10.0.data}/data/share/buildlog/copier.yml +0 -0
- {buildlog-0.8.0.data/data/share/buildlog/template/buildlog → buildlog-0.10.0.data/data/share/buildlog/template/buildlog/.buildlog}/.gitkeep +0 -0
- {buildlog-0.8.0.data/data/share/buildlog/template/buildlog/assets → buildlog-0.10.0.data/data/share/buildlog/template/buildlog/.buildlog/seeds}/.gitkeep +0 -0
- {buildlog-0.8.0.data → buildlog-0.10.0.data}/data/share/buildlog/template/buildlog/2026-01-01-example.md +0 -0
- {buildlog-0.8.0.data → buildlog-0.10.0.data}/data/share/buildlog/template/buildlog/BUILDLOG_SYSTEM.md +0 -0
- {buildlog-0.8.0.data → buildlog-0.10.0.data}/data/share/buildlog/template/buildlog/_TEMPLATE.md +0 -0
- {buildlog-0.8.0.data → buildlog-0.10.0.data}/data/share/buildlog/template/buildlog/_TEMPLATE_QUICK.md +0 -0
- {buildlog-0.8.0.dist-info → buildlog-0.10.0.dist-info}/WHEEL +0 -0
- {buildlog-0.8.0.dist-info → buildlog-0.10.0.dist-info}/entry_points.txt +0 -0
- {buildlog-0.8.0.dist-info → buildlog-0.10.0.dist-info}/licenses/LICENSE +0 -0
|
@@ -0,0 +1,121 @@
|
|
|
1
|
+
"""LLM-backed rule extraction for the seed engine pipeline.
|
|
2
|
+
|
|
3
|
+
Adapts LLMBackend.extract_rules() into the RuleExtractor interface,
|
|
4
|
+
bridging the LLM module with the seed engine's 4-step pipeline.
|
|
5
|
+
"""
|
|
6
|
+
|
|
7
|
+
from __future__ import annotations
|
|
8
|
+
|
|
9
|
+
import logging
|
|
10
|
+
from typing import TYPE_CHECKING, Any
|
|
11
|
+
|
|
12
|
+
from buildlog.seed_engine.extractors import RuleExtractor
|
|
13
|
+
from buildlog.seed_engine.models import CandidateRule, Source
|
|
14
|
+
|
|
15
|
+
if TYPE_CHECKING:
|
|
16
|
+
from buildlog.llm import LLMBackend
|
|
17
|
+
|
|
18
|
+
logger = logging.getLogger(__name__)
|
|
19
|
+
|
|
20
|
+
_PLACEHOLDER = "Not specified by LLM"
|
|
21
|
+
|
|
22
|
+
|
|
23
|
+
class LLMExtractor(RuleExtractor):
|
|
24
|
+
"""LLM-backed rule extraction from source content.
|
|
25
|
+
|
|
26
|
+
Wraps any LLMBackend to produce CandidateRules with full
|
|
27
|
+
defensibility fields. Fields the LLM doesn't populate get
|
|
28
|
+
placeholder values so downstream validation passes.
|
|
29
|
+
|
|
30
|
+
Usage:
|
|
31
|
+
from buildlog.llm import OllamaBackend
|
|
32
|
+
from buildlog.seed_engine.llm_extractor import LLMExtractor
|
|
33
|
+
|
|
34
|
+
backend = OllamaBackend(model="llama3.2")
|
|
35
|
+
extractor = LLMExtractor(backend, source_content={"https://...": "..."})
|
|
36
|
+
|
|
37
|
+
rules = extractor.extract(source)
|
|
38
|
+
"""
|
|
39
|
+
|
|
40
|
+
def __init__(
|
|
41
|
+
self,
|
|
42
|
+
backend: LLMBackend,
|
|
43
|
+
source_content: dict[str, str] | None = None,
|
|
44
|
+
) -> None:
|
|
45
|
+
"""Initialize with an LLM backend.
|
|
46
|
+
|
|
47
|
+
Args:
|
|
48
|
+
backend: Any LLMBackend (Ollama, Anthropic, etc.).
|
|
49
|
+
source_content: Optional map of source.url → text content.
|
|
50
|
+
For sources that need pre-fetched content.
|
|
51
|
+
"""
|
|
52
|
+
self._backend = backend
|
|
53
|
+
self._source_content = source_content or {}
|
|
54
|
+
|
|
55
|
+
def extract(self, source: Source) -> list[CandidateRule]:
|
|
56
|
+
"""Extract candidate rules from a source via LLM.
|
|
57
|
+
|
|
58
|
+
Resolution for content:
|
|
59
|
+
1. source_content dict (keyed by source.url)
|
|
60
|
+
2. source.description as fallback
|
|
61
|
+
|
|
62
|
+
Returns empty list on LLM failure (logged, not raised).
|
|
63
|
+
"""
|
|
64
|
+
content = self._source_content.get(source.url, "").strip()
|
|
65
|
+
if not content:
|
|
66
|
+
content = source.description.strip()
|
|
67
|
+
if not content:
|
|
68
|
+
logger.warning("No content for source %s, skipping", source.name)
|
|
69
|
+
return []
|
|
70
|
+
|
|
71
|
+
try:
|
|
72
|
+
extracted = self._backend.extract_rules(content)
|
|
73
|
+
except Exception:
|
|
74
|
+
logger.exception("LLM extraction failed for %s", source.name)
|
|
75
|
+
return []
|
|
76
|
+
|
|
77
|
+
candidates: list[CandidateRule] = []
|
|
78
|
+
for er in extracted:
|
|
79
|
+
if not er.rule.strip():
|
|
80
|
+
continue
|
|
81
|
+
|
|
82
|
+
metadata: dict[str, Any] = {
|
|
83
|
+
"extractor": "llm",
|
|
84
|
+
"severity": er.severity,
|
|
85
|
+
"scope": er.scope,
|
|
86
|
+
}
|
|
87
|
+
# Include backend class name (public info only)
|
|
88
|
+
metadata["backend_type"] = type(self._backend).__name__
|
|
89
|
+
|
|
90
|
+
candidates.append(
|
|
91
|
+
CandidateRule(
|
|
92
|
+
rule=er.rule,
|
|
93
|
+
context=er.context or _PLACEHOLDER,
|
|
94
|
+
antipattern=er.antipattern or _PLACEHOLDER,
|
|
95
|
+
rationale=er.rationale or _PLACEHOLDER,
|
|
96
|
+
source=source,
|
|
97
|
+
raw_tags=[er.category] + er.applicability,
|
|
98
|
+
confidence=0.7,
|
|
99
|
+
metadata=metadata,
|
|
100
|
+
)
|
|
101
|
+
)
|
|
102
|
+
|
|
103
|
+
logger.info("LLM extracted %d rules from %s", len(candidates), source.name)
|
|
104
|
+
return candidates
|
|
105
|
+
|
|
106
|
+
def validate(self, rule: CandidateRule) -> list[str]:
|
|
107
|
+
"""Validate a candidate rule.
|
|
108
|
+
|
|
109
|
+
Warns on placeholder defensibility fields.
|
|
110
|
+
Requires non-empty rule text.
|
|
111
|
+
"""
|
|
112
|
+
issues: list[str] = []
|
|
113
|
+
if not rule.rule.strip():
|
|
114
|
+
issues.append("Rule text is empty")
|
|
115
|
+
if rule.context == _PLACEHOLDER:
|
|
116
|
+
issues.append("Context is LLM placeholder — consider enriching")
|
|
117
|
+
if rule.antipattern == _PLACEHOLDER:
|
|
118
|
+
issues.append("Antipattern is LLM placeholder — consider enriching")
|
|
119
|
+
if rule.rationale == _PLACEHOLDER:
|
|
120
|
+
issues.append("Rationale is LLM placeholder — consider enriching")
|
|
121
|
+
return issues
|
buildlog/seed_engine/pipeline.py
CHANGED
|
@@ -12,13 +12,16 @@ from __future__ import annotations
|
|
|
12
12
|
import logging
|
|
13
13
|
from dataclasses import dataclass
|
|
14
14
|
from pathlib import Path
|
|
15
|
-
from typing import Any
|
|
15
|
+
from typing import TYPE_CHECKING, Any
|
|
16
16
|
|
|
17
17
|
from buildlog.seed_engine.categorizers import Categorizer, TagBasedCategorizer
|
|
18
18
|
from buildlog.seed_engine.extractors import ManualExtractor, RuleExtractor
|
|
19
19
|
from buildlog.seed_engine.generators import SeedGenerator
|
|
20
20
|
from buildlog.seed_engine.models import CandidateRule, CategorizedRule, Source
|
|
21
21
|
|
|
22
|
+
if TYPE_CHECKING:
|
|
23
|
+
from buildlog.llm import LLMBackend
|
|
24
|
+
|
|
22
25
|
logger = logging.getLogger(__name__)
|
|
23
26
|
|
|
24
27
|
|
|
@@ -174,6 +177,7 @@ class Pipeline:
|
|
|
174
177
|
Returns:
|
|
175
178
|
List of validation issues (empty if valid).
|
|
176
179
|
"""
|
|
180
|
+
allowed_schemes = {"https", "http", "file"}
|
|
177
181
|
issues = []
|
|
178
182
|
for i, source in enumerate(sources):
|
|
179
183
|
prefix = f"Source {i + 1} ({source.name})"
|
|
@@ -181,10 +185,50 @@ class Pipeline:
|
|
|
181
185
|
issues.append(f"{prefix}: Missing name")
|
|
182
186
|
if not source.url.strip():
|
|
183
187
|
issues.append(f"{prefix}: Missing URL")
|
|
188
|
+
else:
|
|
189
|
+
# Validate URL scheme
|
|
190
|
+
scheme = (
|
|
191
|
+
source.url.split("://")[0].lower() if "://" in source.url else ""
|
|
192
|
+
)
|
|
193
|
+
if scheme not in allowed_schemes:
|
|
194
|
+
issues.append(
|
|
195
|
+
f"{prefix}: URL scheme '{scheme}' not in allowlist {allowed_schemes}"
|
|
196
|
+
)
|
|
184
197
|
if not source.domain.strip():
|
|
185
198
|
issues.append(f"{prefix}: Missing domain")
|
|
186
199
|
return issues
|
|
187
200
|
|
|
201
|
+
@classmethod
|
|
202
|
+
def with_llm(
|
|
203
|
+
cls,
|
|
204
|
+
persona: str,
|
|
205
|
+
backend: LLMBackend,
|
|
206
|
+
source_content: dict[str, str] | None = None,
|
|
207
|
+
default_category: str = "general",
|
|
208
|
+
version: int = 1,
|
|
209
|
+
) -> Pipeline:
|
|
210
|
+
"""Convenience constructor wiring LLMExtractor + TagBasedCategorizer.
|
|
211
|
+
|
|
212
|
+
Args:
|
|
213
|
+
persona: Persona name for the seed file.
|
|
214
|
+
backend: Any LLMBackend implementation.
|
|
215
|
+
source_content: Optional pre-fetched content map.
|
|
216
|
+
default_category: Fallback category for uncategorized rules.
|
|
217
|
+
version: Seed file version.
|
|
218
|
+
|
|
219
|
+
Returns:
|
|
220
|
+
Pipeline configured with LLMExtractor.
|
|
221
|
+
"""
|
|
222
|
+
from buildlog.seed_engine.llm_extractor import LLMExtractor
|
|
223
|
+
|
|
224
|
+
return cls(
|
|
225
|
+
persona=persona,
|
|
226
|
+
default_category=default_category,
|
|
227
|
+
version=version,
|
|
228
|
+
extractor=LLMExtractor(backend, source_content),
|
|
229
|
+
categorizer=TagBasedCategorizer(default_category=default_category),
|
|
230
|
+
)
|
|
231
|
+
|
|
188
232
|
def dry_run(self, sources: list[Source]) -> dict[str, Any]:
|
|
189
233
|
"""Run pipeline without writing, returning preview.
|
|
190
234
|
|
|
@@ -3,7 +3,11 @@
|
|
|
3
3
|
|
|
4
4
|
from pathlib import Path
|
|
5
5
|
|
|
6
|
-
|
|
6
|
+
try:
|
|
7
|
+
from buildlog.constants import CLAUDE_MD_BUILDLOG_SECTION
|
|
8
|
+
except ImportError:
|
|
9
|
+
# Fallback for when buildlog isn't installed (e.g., copier from GitHub)
|
|
10
|
+
CLAUDE_MD_BUILDLOG_SECTION = """
|
|
7
11
|
## Build Journal
|
|
8
12
|
|
|
9
13
|
After completing significant work (features, debugging sessions, deployments,
|
|
@@ -40,15 +44,16 @@ def main():
|
|
|
40
44
|
|
|
41
45
|
content = claude_md.read_text()
|
|
42
46
|
|
|
43
|
-
|
|
44
|
-
|
|
47
|
+
# Check for either old or new section marker
|
|
48
|
+
if "## buildlog Integration" in content or "## Build Journal" in content:
|
|
49
|
+
print("buildlog section already exists in CLAUDE.md")
|
|
45
50
|
return
|
|
46
51
|
|
|
47
52
|
# Append to end of file
|
|
48
53
|
with open(claude_md, "a") as f:
|
|
49
|
-
f.write("\n" +
|
|
54
|
+
f.write("\n" + CLAUDE_MD_BUILDLOG_SECTION)
|
|
50
55
|
|
|
51
|
-
print("Added
|
|
56
|
+
print("Added buildlog Integration section to CLAUDE.md")
|
|
52
57
|
|
|
53
58
|
|
|
54
59
|
if __name__ == "__main__":
|
|
File without changes
|
|
File without changes
|
|
@@ -0,0 +1,248 @@
|
|
|
1
|
+
Metadata-Version: 2.4
|
|
2
|
+
Name: buildlog
|
|
3
|
+
Version: 0.10.0
|
|
4
|
+
Summary: Engineering notebook for AI-assisted development
|
|
5
|
+
Project-URL: Homepage, https://github.com/Peleke/buildlog-template
|
|
6
|
+
Project-URL: Repository, https://github.com/Peleke/buildlog-template
|
|
7
|
+
Author: Peleke Sengstacke
|
|
8
|
+
License-Expression: MIT
|
|
9
|
+
License-File: LICENSE
|
|
10
|
+
Keywords: ai,buildlog,development,documentation,journal
|
|
11
|
+
Classifier: Development Status :: 4 - Beta
|
|
12
|
+
Classifier: Environment :: Console
|
|
13
|
+
Classifier: Intended Audience :: Developers
|
|
14
|
+
Classifier: License :: OSI Approved :: MIT License
|
|
15
|
+
Classifier: Programming Language :: Python :: 3
|
|
16
|
+
Classifier: Programming Language :: Python :: 3.10
|
|
17
|
+
Classifier: Programming Language :: Python :: 3.11
|
|
18
|
+
Classifier: Programming Language :: Python :: 3.12
|
|
19
|
+
Classifier: Programming Language :: Python :: 3.13
|
|
20
|
+
Classifier: Topic :: Documentation
|
|
21
|
+
Classifier: Topic :: Software Development :: Documentation
|
|
22
|
+
Requires-Python: >=3.10
|
|
23
|
+
Requires-Dist: click>=8.0.0
|
|
24
|
+
Requires-Dist: copier>=9.0.0
|
|
25
|
+
Requires-Dist: mcp>=1.0.0
|
|
26
|
+
Requires-Dist: numpy>=1.21.0
|
|
27
|
+
Requires-Dist: pymupdf>=1.26.7
|
|
28
|
+
Requires-Dist: pyyaml>=6.0.0
|
|
29
|
+
Provides-Extra: all
|
|
30
|
+
Requires-Dist: anthropic>=0.40.0; extra == 'all'
|
|
31
|
+
Requires-Dist: ollama>=0.4.0; extra == 'all'
|
|
32
|
+
Requires-Dist: openai>=1.0.0; extra == 'all'
|
|
33
|
+
Requires-Dist: sentence-transformers>=2.2.0; extra == 'all'
|
|
34
|
+
Provides-Extra: anthropic
|
|
35
|
+
Requires-Dist: anthropic>=0.40.0; extra == 'anthropic'
|
|
36
|
+
Provides-Extra: dev
|
|
37
|
+
Requires-Dist: black>=24.0.0; extra == 'dev'
|
|
38
|
+
Requires-Dist: flake8>=7.0.0; extra == 'dev'
|
|
39
|
+
Requires-Dist: isort>=5.13.0; extra == 'dev'
|
|
40
|
+
Requires-Dist: mkdocs-material>=9.5.0; extra == 'dev'
|
|
41
|
+
Requires-Dist: mypy>=1.8.0; extra == 'dev'
|
|
42
|
+
Requires-Dist: pre-commit>=3.6.0; extra == 'dev'
|
|
43
|
+
Requires-Dist: pytest-asyncio>=0.21.0; extra == 'dev'
|
|
44
|
+
Requires-Dist: pytest-cov>=4.0.0; extra == 'dev'
|
|
45
|
+
Requires-Dist: pytest>=7.0.0; extra == 'dev'
|
|
46
|
+
Requires-Dist: types-pyyaml>=6.0.0; extra == 'dev'
|
|
47
|
+
Provides-Extra: embeddings
|
|
48
|
+
Requires-Dist: sentence-transformers>=2.2.0; extra == 'embeddings'
|
|
49
|
+
Provides-Extra: engine
|
|
50
|
+
Provides-Extra: llm
|
|
51
|
+
Requires-Dist: anthropic>=0.40.0; extra == 'llm'
|
|
52
|
+
Requires-Dist: ollama>=0.4.0; extra == 'llm'
|
|
53
|
+
Provides-Extra: mcp
|
|
54
|
+
Provides-Extra: ollama
|
|
55
|
+
Requires-Dist: ollama>=0.4.0; extra == 'ollama'
|
|
56
|
+
Provides-Extra: openai
|
|
57
|
+
Requires-Dist: openai>=1.0.0; extra == 'openai'
|
|
58
|
+
Description-Content-Type: text/markdown
|
|
59
|
+
|
|
60
|
+
<div align="center">
|
|
61
|
+
|
|
62
|
+
# buildlog
|
|
63
|
+
|
|
64
|
+
### A measurable learning loop for AI-assisted work
|
|
65
|
+
|
|
66
|
+
[](https://pypi.org/project/buildlog/)
|
|
67
|
+
[](https://python.org/)
|
|
68
|
+
[](https://github.com/Peleke/buildlog-template/actions/workflows/ci.yml)
|
|
69
|
+
[](https://opensource.org/licenses/MIT)
|
|
70
|
+
[](https://peleke.github.io/buildlog-template/)
|
|
71
|
+
|
|
72
|
+
**Track what works. Prove it. Drop what doesn't.**
|
|
73
|
+
|
|
74
|
+
<img src="assets/hero-banner-perfectdeliberate.png" alt="buildlog - A measurable learning loop for AI-assisted work" width="800"/>
|
|
75
|
+
|
|
76
|
+
> **RE: The art.** Yes, it's AI-generated. Yes, that's hypocritical for a project about rigor over vibes. Looking for an actual artist to pay for a real logo. If you know someone good, [open an issue](https://github.com/Peleke/buildlog-template/issues) or DM me. Budget exists.
|
|
77
|
+
|
|
78
|
+
**[Read the full documentation](https://peleke.github.io/buildlog-template/)**
|
|
79
|
+
|
|
80
|
+
</div>
|
|
81
|
+
|
|
82
|
+
---
|
|
83
|
+
|
|
84
|
+
## The Problem
|
|
85
|
+
|
|
86
|
+
Most AI agents do not learn. They execute without retaining context. You can bolt on memory stores and tool routers, but if the system cannot demonstrably improve its decision-making over time, you have a persistent memory store, not a learning system.
|
|
87
|
+
|
|
88
|
+
Every AI-assisted work session produces a trajectory: goals, decisions, tool uses, corrections, outcomes. Almost all of this is discarded. The next session starts from scratch with the same blind spots.
|
|
89
|
+
|
|
90
|
+
buildlog exists to close that gap. It captures structured trajectories from real work, extracts decision patterns, and uses statistical methods to select which patterns to surface in future sessions, then measures whether that selection actually reduced mistakes.
|
|
91
|
+
|
|
92
|
+
buildlog measures whether the system actually got better, and proves it.
|
|
93
|
+
|
|
94
|
+
## How It Works
|
|
95
|
+
|
|
96
|
+
### 1. Capture structured work trajectories
|
|
97
|
+
|
|
98
|
+
Each session is a dated entry documenting what you did, what went wrong, and what you learned. Each session is a structured record of decisions and outcomes, not a chat transcript.
|
|
99
|
+
|
|
100
|
+
```bash
|
|
101
|
+
buildlog init # scaffold a project
|
|
102
|
+
buildlog new my-feature # start a session
|
|
103
|
+
# ... work ...
|
|
104
|
+
buildlog commit -m "feat: add auth"
|
|
105
|
+
```
|
|
106
|
+
|
|
107
|
+
### 2. Extract decision patterns as seeds
|
|
108
|
+
|
|
109
|
+
The seed engine watches your development patterns and extracts **seeds**: atomic observations about what works. A seed might be "always define interfaces before implementations" or "mock at the boundary, not the implementation." Each seed carries a category, a confidence score, and source provenance.
|
|
110
|
+
|
|
111
|
+
Extraction runs through a pipeline: `sources -> extractors -> categorizers -> generators`. Extractors range from regex-based (fast, cheap, brittle) to LLM-backed (accurate, expensive). The pipeline deduplicates semantically using embeddings.
|
|
112
|
+
|
|
113
|
+
### 3. Select which patterns to surface using Thompson Sampling
|
|
114
|
+
|
|
115
|
+
Seeds compete for inclusion in your agent's instruction set. The system treats each seed as an arm in a contextual bandit and uses **Thompson Sampling** to balance exploration (trying under-tested rules) against exploitation (surfacing rules with strong track records).
|
|
116
|
+
|
|
117
|
+
Each seed maintains a Beta posterior updated by observed outcomes. Over time, the system converges on the rules that actually reduce mistakes in your specific codebase and workflow, not rules that sound good in the abstract.
|
|
118
|
+
|
|
119
|
+
### 4. Render to every agent format
|
|
120
|
+
|
|
121
|
+
Selected rules are written into the instruction files your agents actually read:
|
|
122
|
+
|
|
123
|
+
- `CLAUDE.md` (Claude Code)
|
|
124
|
+
- `.cursorrules` (Cursor)
|
|
125
|
+
- `.github/copilot-instructions.md` (GitHub Copilot)
|
|
126
|
+
- Windsurf, Continue.dev, generic `settings.json`
|
|
127
|
+
|
|
128
|
+
The same knowledge base renders to every agent format.
|
|
129
|
+
|
|
130
|
+
```bash
|
|
131
|
+
buildlog skills # render current policy to agent files
|
|
132
|
+
```
|
|
133
|
+
|
|
134
|
+
### 5. Close the loop with experiments
|
|
135
|
+
|
|
136
|
+
Track whether the selected rules are working. Run experiments, measure Repeated Mistake Rate (RMR) across sessions, and get statistical evidence, not feelings, about what improved.
|
|
137
|
+
|
|
138
|
+
```bash
|
|
139
|
+
buildlog experiment start
|
|
140
|
+
# ... work across sessions ...
|
|
141
|
+
buildlog experiment end
|
|
142
|
+
buildlog experiment report
|
|
143
|
+
```
|
|
144
|
+
|
|
145
|
+
## What Else Is In the Box
|
|
146
|
+
|
|
147
|
+
- **Review gauntlet:** automated quality gate with curated reviewer personas. Runs on commits (via Claude Code hooks or CI) and files GitHub issues for findings, categorized by severity.
|
|
148
|
+
- **LLM-backed extraction:** when regex isn't enough, the seed engine can use OpenAI, Anthropic, or Ollama to extract patterns from code and logs. Metered backend tracks token usage and cost.
|
|
149
|
+
- **MCP server:** buildlog exposes itself as an MCP server so agents can query seeds, skills, and build history programmatically during sessions.
|
|
150
|
+
- **npm wrapper:** `npx @peleke.s/buildlog` for JS/TS projects. Thin shim that finds and invokes the Python CLI.
|
|
151
|
+
|
|
152
|
+
## Current Limits
|
|
153
|
+
|
|
154
|
+
This is v0.8, not the end state.
|
|
155
|
+
|
|
156
|
+
- **Extraction quality is uneven.** Regex extractors miss nuance; LLM extractors are accurate but expensive. The middle ground is still being found.
|
|
157
|
+
- **Feedback signals are coarse.** Repeated Mistake Rate works but requires manual tagging. Richer automatic signals (test outcomes, review results, revision distance) are on the roadmap.
|
|
158
|
+
- **Credit assignment is limited.** When multiple rules are active, the system doesn't yet isolate which one was responsible for an outcome.
|
|
159
|
+
- **Single-agent only.** Multi-agent coordination (shared learning across agents) is designed but not implemented.
|
|
160
|
+
- **Long-horizon learning is not modeled.** The bandit operates per-session. Longer arcs of competence building need richer policy models.
|
|
161
|
+
|
|
162
|
+
The roadmap: contextual bandits (now) -> richer policy models -> longer-horizon RL -> multi-agent coordination. Each step builds on the same foundation: measuring whether rule changes actually reduce mistakes.
|
|
163
|
+
|
|
164
|
+
## Installation
|
|
165
|
+
|
|
166
|
+
### Quick start
|
|
167
|
+
|
|
168
|
+
```bash
|
|
169
|
+
pip install buildlog # MCP server included by default
|
|
170
|
+
buildlog init --defaults # scaffold project, register MCP, update CLAUDE.md
|
|
171
|
+
```
|
|
172
|
+
|
|
173
|
+
That's it. Claude Code will now have access to all 29 buildlog tools.
|
|
174
|
+
|
|
175
|
+
### Global install (recommended)
|
|
176
|
+
|
|
177
|
+
```bash
|
|
178
|
+
uv tool install buildlog # or: pipx install buildlog
|
|
179
|
+
```
|
|
180
|
+
|
|
181
|
+
This puts `buildlog` and `buildlog-mcp` on your PATH. Works from any directory.
|
|
182
|
+
|
|
183
|
+
### Per-project (virtual environment)
|
|
184
|
+
|
|
185
|
+
```bash
|
|
186
|
+
uv pip install buildlog # or: pip install buildlog
|
|
187
|
+
```
|
|
188
|
+
|
|
189
|
+
### For JS/TS projects
|
|
190
|
+
|
|
191
|
+
```bash
|
|
192
|
+
npx @peleke.s/buildlog init
|
|
193
|
+
```
|
|
194
|
+
|
|
195
|
+
### MCP server for Claude Code
|
|
196
|
+
|
|
197
|
+
`buildlog init` auto-registers the MCP server. For existing projects:
|
|
198
|
+
|
|
199
|
+
```bash
|
|
200
|
+
buildlog init-mcp # register MCP in .claude/settings.json
|
|
201
|
+
buildlog mcp-test # verify all 29 tools are registered
|
|
202
|
+
```
|
|
203
|
+
|
|
204
|
+
This exposes buildlog tools (seeds, skills, experiments, gauntlet, bandit status) to any Claude Code session.
|
|
205
|
+
|
|
206
|
+
## Quick Start
|
|
207
|
+
|
|
208
|
+
```bash
|
|
209
|
+
buildlog init --defaults # scaffold + MCP + CLAUDE.md
|
|
210
|
+
buildlog new my-feature # start a session
|
|
211
|
+
# ... work ...
|
|
212
|
+
buildlog commit -m "feat: add auth"
|
|
213
|
+
buildlog experiment start
|
|
214
|
+
# ... work across sessions ...
|
|
215
|
+
buildlog experiment end
|
|
216
|
+
buildlog experiment report
|
|
217
|
+
```
|
|
218
|
+
|
|
219
|
+
## Documentation
|
|
220
|
+
|
|
221
|
+
| Section | Description |
|
|
222
|
+
|---------|------------|
|
|
223
|
+
| [Installation](https://peleke.github.io/buildlog-template/getting-started/installation/) | Setup, extras, and initialization |
|
|
224
|
+
| [Quick Start](https://peleke.github.io/buildlog-template/getting-started/quick-start/) | Full pipeline walkthrough |
|
|
225
|
+
| [Core Concepts](https://peleke.github.io/buildlog-template/getting-started/concepts/) | The problem, the claim, and the metric |
|
|
226
|
+
| [CLI Reference](https://peleke.github.io/buildlog-template/guides/cli-reference/) | Every command documented |
|
|
227
|
+
| [MCP Integration](https://peleke.github.io/buildlog-template/guides/mcp-integration/) | Claude Code setup and available tools |
|
|
228
|
+
| [Experiments](https://peleke.github.io/buildlog-template/guides/experiments/) | Running and measuring experiments |
|
|
229
|
+
| [Review Gauntlet](https://peleke.github.io/buildlog-template/guides/review-gauntlet/) | Reviewer personas and the gauntlet loop |
|
|
230
|
+
| [Multi-Agent Setup](https://peleke.github.io/buildlog-template/guides/multi-agent/) | Render rules to any AI coding agent |
|
|
231
|
+
| [Theory](https://peleke.github.io/buildlog-template/theory/00-background/) | The math behind Thompson Sampling |
|
|
232
|
+
| [Philosophy](https://peleke.github.io/buildlog-template/philosophy/) | Principles and honest limitations |
|
|
233
|
+
|
|
234
|
+
## Contributing
|
|
235
|
+
|
|
236
|
+
```bash
|
|
237
|
+
git clone https://github.com/Peleke/buildlog-template
|
|
238
|
+
cd buildlog-template
|
|
239
|
+
uv venv && source .venv/bin/activate
|
|
240
|
+
uv pip install -e ".[dev]"
|
|
241
|
+
pytest
|
|
242
|
+
```
|
|
243
|
+
|
|
244
|
+
We're especially interested in better context representations, credit assignment approaches, statistical methodology improvements, and real-world experiment results (positive or negative).
|
|
245
|
+
|
|
246
|
+
## License
|
|
247
|
+
|
|
248
|
+
MIT License. See [LICENSE](./LICENSE)
|
|
@@ -1,16 +1,18 @@
|
|
|
1
1
|
buildlog/__init__.py,sha256=wxluyg3fDOiaKUAOJa0cav39hwtNJS5Y_0X20uL_yi4,90
|
|
2
|
-
buildlog/cli.py,sha256=
|
|
2
|
+
buildlog/cli.py,sha256=YSLegAPnIqPR_AGEkSTe3DaZilhRH4kdDUhbk1dLCD4,74956
|
|
3
3
|
buildlog/confidence.py,sha256=jPvN0_3drGHQG1C7iUxUtYjKC62nNKsHHxu6WXMfJFg,10137
|
|
4
|
+
buildlog/constants.py,sha256=nwRAiI4NcXczdUK1C332i36BfnPqjMp4gHg2E4y4n5M,5783
|
|
4
5
|
buildlog/distill.py,sha256=PL7UBToBb27BrCOTWGBTDIXGggtrUumHHBs0_MfG6vY,14166
|
|
5
6
|
buildlog/embeddings.py,sha256=vPydWjJVkYp172zFou-lJ737qsu6vRMQAMs143RGIpA,12364
|
|
6
|
-
buildlog/llm.py,sha256=
|
|
7
|
+
buildlog/llm.py,sha256=t_3KnJ_eBmrtHaay4owG3EgGbJrVp_T379P0tTuyJf8,16986
|
|
7
8
|
buildlog/seeds.py,sha256=L-lzO7rrDjMdVHcYKvEBm6scVDfCmFD0koiyafMZDLo,8596
|
|
8
9
|
buildlog/skills.py,sha256=3CihyyAFfJh0w0VNXQauNhGm_RM8m_GIO1iiNLNPmsQ,33789
|
|
9
10
|
buildlog/stats.py,sha256=2WdHdmzUNGobtWngmm9nA_UmqM7DQeAnZL8_rLQN8aw,13256
|
|
10
|
-
buildlog/core/__init__.py,sha256=
|
|
11
|
+
buildlog/core/__init__.py,sha256=yjmo_pWE7ijCvMSgBXDPVl5lgsUlTXjc1Nv7eIRjlVI,2642
|
|
11
12
|
buildlog/core/bandit.py,sha256=0z58s42jpQiBgXdjb465apenWdBVTvoc7bDOcnAcPLo,24565
|
|
12
|
-
buildlog/core/operations.py,sha256=
|
|
13
|
+
buildlog/core/operations.py,sha256=fVpUcxt2AzktGy1iwSpazj3sLob6l4Rio9QtxDWHOec,104634
|
|
13
14
|
buildlog/data/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
|
|
15
|
+
buildlog/data/seeds/bragi.yaml,sha256=2bheCsPHp6gHvgi8jCFTebs_q5X8WLVUduJe-Q-kQ0Y,3702
|
|
14
16
|
buildlog/data/seeds/security_karen.yaml,sha256=Hjwp4j6FB-e0vCXp6AkyhAEkwASR4H1owdhPBID740E,8775
|
|
15
17
|
buildlog/data/seeds/test_terrorist.yaml,sha256=-QY2x6d0LrlM6msMYaYwR43Wny9ZybQmU7pH_E7gSB0,11113
|
|
16
18
|
buildlog/engine/__init__.py,sha256=KNTSUZTUHK6r2abAA-2PIb0wUcRgbE4juvBmoywaXHI,1750
|
|
@@ -19,9 +21,9 @@ buildlog/engine/confidence.py,sha256=7Zhw8bwT_S-qeVaGXT6dHOyjDTvwyhetLKCWhJH9jMc
|
|
|
19
21
|
buildlog/engine/embeddings.py,sha256=78InITcBbaPZ29tUlb4hJNNVZ1fhxstvVa2fYvOh6QA,597
|
|
20
22
|
buildlog/engine/experiments.py,sha256=38f_UsbY4TP3LTGM9yUsZxoXqxUGOxmLCoYrfZqjVks,20041
|
|
21
23
|
buildlog/engine/types.py,sha256=W7U2B4TE79xACV0cuaidKevmTEskxqXfltH5hjIxV5k,734
|
|
22
|
-
buildlog/mcp/__init__.py,sha256=
|
|
23
|
-
buildlog/mcp/server.py,sha256=
|
|
24
|
-
buildlog/mcp/tools.py,sha256=
|
|
24
|
+
buildlog/mcp/__init__.py,sha256=dSjiS__oyYuGipAQDfNlM37NYZ60TMCCA9WXtQk81x0,1668
|
|
25
|
+
buildlog/mcp/server.py,sha256=Ua0Jk5jRuHGDSSLS0QEJhsXW2fKtrQlUdow-aeumI9M,2233
|
|
26
|
+
buildlog/mcp/tools.py,sha256=MRM9UMwoDmX0WaRq3JD1DlNnOOJdjNOzq7CQs2f0IC4,32764
|
|
25
27
|
buildlog/render/__init__.py,sha256=jj-o-vPdqpivdFKK85xskCOT8Zn_2tb3Q5Jb0gWLjAg,2567
|
|
26
28
|
buildlog/render/base.py,sha256=gQfvOsH1zttAo10xtEyNsAbqZ4NRSPiDihO-aiGgTsw,533
|
|
27
29
|
buildlog/render/claude_md.py,sha256=R5z4dHh-3YwclbW3MAzHxgYQHtd5lXS3McJJ1isLGqE,4901
|
|
@@ -32,23 +34,26 @@ buildlog/render/settings_json.py,sha256=4DS5OWksPrFCa7MIgWIu0t4rxYmItpMdGfTqMX3a
|
|
|
32
34
|
buildlog/render/skill.py,sha256=_7umIS1Ms1oQ2_PopYueFjX41nMq1p28yJp6DhXFdgU,5981
|
|
33
35
|
buildlog/render/tracking.py,sha256=1y_Mf1s3KblV8TC0W9x2a_ynSJHLwkVZJcW1gs1a6q4,1786
|
|
34
36
|
buildlog/render/windsurf.py,sha256=jfJLb-luQLjdBM38_KKp0mZ08bUSDTBPxSLOIFWxawQ,3047
|
|
35
|
-
buildlog/seed_engine/__init__.py,sha256=
|
|
37
|
+
buildlog/seed_engine/__init__.py,sha256=QHo59cq-Ar_M6UthSOu_zpfYR1OF2ePu1qWrbgQqLak,1949
|
|
36
38
|
buildlog/seed_engine/categorizers.py,sha256=uzatJQPG4Vf7z7ZAeyQbL-bVJ8cUKBHG-3U93-BFIRQ,4649
|
|
37
39
|
buildlog/seed_engine/extractors.py,sha256=rkztg62LzFH018ob5Tl1eFp_yfPo8syn3Ib4h6bAyK4,4824
|
|
38
40
|
buildlog/seed_engine/generators.py,sha256=T3N3uHg359xSBypIGmvVWYI0Qh8WLlGIFQWX0_gs0I4,4358
|
|
41
|
+
buildlog/seed_engine/llm_extractor.py,sha256=nfTmN4vuEiPQuhiL_2DX1QZ7TZlDR2vxYy0tVrkf2Vc,4124
|
|
39
42
|
buildlog/seed_engine/models.py,sha256=qESSuoF3CJy8pp96E3Vmb38V-snzeaW0Xg0RfbH8v1U,3418
|
|
40
|
-
buildlog/seed_engine/pipeline.py,sha256=
|
|
43
|
+
buildlog/seed_engine/pipeline.py,sha256=cyR2Vf32Hf0PK5DxnaFcq56ygjJIiJWiIkHY4Uz7AS8,8082
|
|
41
44
|
buildlog/seed_engine/sources.py,sha256=8j9oUFZCSKMr5VpIuAxTPY3wTzfTEmw6M_41_aismiE,11184
|
|
42
|
-
buildlog-0.
|
|
43
|
-
buildlog-0.
|
|
44
|
-
buildlog-0.
|
|
45
|
-
buildlog-0.
|
|
46
|
-
buildlog-0.
|
|
47
|
-
buildlog-0.
|
|
48
|
-
buildlog-0.
|
|
49
|
-
buildlog-0.
|
|
50
|
-
buildlog-0.
|
|
51
|
-
buildlog-0.
|
|
52
|
-
buildlog-0.
|
|
53
|
-
buildlog-0.
|
|
54
|
-
buildlog-0.
|
|
45
|
+
buildlog-0.10.0.data/data/share/buildlog/copier.yml,sha256=hD_UcLGB9eDxTGkWjHAvUoiqQKhLC0w3G-kPKX_p43I,802
|
|
46
|
+
buildlog-0.10.0.data/data/share/buildlog/post_gen.py,sha256=YQJYQjbRqfmyKrRRkdqk8qTkXwqompg-AgroKUR8W9M,2020
|
|
47
|
+
buildlog-0.10.0.data/data/share/buildlog/template/buildlog/.gitkeep,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
|
|
48
|
+
buildlog-0.10.0.data/data/share/buildlog/template/buildlog/2026-01-01-example.md,sha256=7x9sKmydfmfKyNz9hV7MtYnQJuBwbxNanbPOcpQDDZQ,7040
|
|
49
|
+
buildlog-0.10.0.data/data/share/buildlog/template/buildlog/BUILDLOG_SYSTEM.md,sha256=osclytWwl5jUiTgSpuT4cT3h3oPvCkZ5GPCnFuJZNcY,3802
|
|
50
|
+
buildlog-0.10.0.data/data/share/buildlog/template/buildlog/_TEMPLATE.md,sha256=CUvxgcx1-9XT_EdQ8e_vnuPq_h-u1uhXJgForJU2Pso,2932
|
|
51
|
+
buildlog-0.10.0.data/data/share/buildlog/template/buildlog/_TEMPLATE_QUICK.md,sha256=eUr5MiqLsM6drV7rAq53R1SLkK8G7LkMAUjWKXx81IA,409
|
|
52
|
+
buildlog-0.10.0.data/data/share/buildlog/template/buildlog/.buildlog/.gitkeep,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
|
|
53
|
+
buildlog-0.10.0.data/data/share/buildlog/template/buildlog/.buildlog/seeds/.gitkeep,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
|
|
54
|
+
buildlog-0.10.0.data/data/share/buildlog/template/buildlog/assets/.gitkeep,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
|
|
55
|
+
buildlog-0.10.0.dist-info/METADATA,sha256=3rhwCyjU4M6hABubghAUvxA2uEz1YvK8hjMaoQY_PYk,11465
|
|
56
|
+
buildlog-0.10.0.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
|
|
57
|
+
buildlog-0.10.0.dist-info/entry_points.txt,sha256=BMFclPOomp_sgaa0OqBg6LfqCMlqzjZV88ww5TrPPoo,87
|
|
58
|
+
buildlog-0.10.0.dist-info/licenses/LICENSE,sha256=fAgt-akug9nAwIj6M-SIf8u3ck-T7pJTwfmy9vWYASk,1074
|
|
59
|
+
buildlog-0.10.0.dist-info/RECORD,,
|