buildlog 0.4.0__py3-none-any.whl → 0.5.0__py3-none-any.whl

buildlog-0.5.0.dist-info/METADATA

@@ -0,0 +1,419 @@
+Metadata-Version: 2.4
+Name: buildlog
+Version: 0.5.0
+Summary: Engineering notebook for AI-assisted development
+Project-URL: Homepage, https://github.com/Peleke/buildlog-template
+Project-URL: Repository, https://github.com/Peleke/buildlog-template
+Author: Peleke Sengstacke
+License-Expression: MIT
+License-File: LICENSE
+Keywords: ai,buildlog,development,documentation,journal
+Classifier: Development Status :: 4 - Beta
+Classifier: Environment :: Console
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Documentation
+Classifier: Topic :: Software Development :: Documentation
+Requires-Python: >=3.10
+Requires-Dist: click>=8.0.0
+Requires-Dist: copier>=9.0.0
+Requires-Dist: numpy>=1.21.0
+Requires-Dist: pyyaml>=6.0.0
+Provides-Extra: all
+Requires-Dist: mcp>=1.0.0; extra == 'all'
+Requires-Dist: openai>=1.0.0; extra == 'all'
+Requires-Dist: sentence-transformers>=2.2.0; extra == 'all'
+Provides-Extra: dev
+Requires-Dist: black>=24.0.0; extra == 'dev'
+Requires-Dist: flake8>=7.0.0; extra == 'dev'
+Requires-Dist: isort>=5.13.0; extra == 'dev'
+Requires-Dist: mypy>=1.8.0; extra == 'dev'
+Requires-Dist: pre-commit>=3.6.0; extra == 'dev'
+Requires-Dist: pytest-asyncio>=0.21.0; extra == 'dev'
+Requires-Dist: pytest-cov>=4.0.0; extra == 'dev'
+Requires-Dist: pytest>=7.0.0; extra == 'dev'
+Requires-Dist: types-pyyaml>=6.0.0; extra == 'dev'
+Provides-Extra: embeddings
+Requires-Dist: sentence-transformers>=2.2.0; extra == 'embeddings'
+Provides-Extra: mcp
+Requires-Dist: mcp>=1.0.0; extra == 'mcp'
+Provides-Extra: openai
+Requires-Dist: openai>=1.0.0; extra == 'openai'
+Description-Content-Type: text/markdown
+
+<div align="center">
+
+> **NOTE:** Forgive the agent dump; edit incoming.
+>
+> **RE: The art** — Yes, it's AI-generated. Yes, that's hypocritical for a project about rigor over vibes. Now that this is no longer internal, 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.
+
+# buildlog
+
+### The Only Agent Learning System You Can Prove Works
+
+[![PyPI](https://img.shields.io/pypi/v/buildlog?style=for-the-badge&logo=pypi&logoColor=white)](https://pypi.org/project/buildlog/)
+[![Python](https://img.shields.io/pypi/pyversions/buildlog?style=for-the-badge&logo=python&logoColor=white)](https://python.org/)
+[![CI](https://img.shields.io/github/actions/workflow/status/Peleke/buildlog-template/ci.yml?branch=main&style=for-the-badge&logo=github&label=CI)](https://github.com/Peleke/buildlog-template/actions/workflows/ci.yml)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg?style=for-the-badge)](https://opensource.org/licenses/MIT)
+
+**Falsifiable claims. Measurable outcomes. No vibes.**
+
+<img src="assets/hero-banner-perfectdeliberate.png" alt="buildlog - The Only Agent Learning System You Can Prove Works" width="800"/>
+
+[The Problem](#the-problem) · [The Claim](#the-claim) · [The Mechanism](#the-mechanism) · [Quick Start](#quick-start)
+
+---
+
+</div>
+
+## The Problem
+
+Everyone's building "agent memory." Blog posts announce breakthroughs. Tweets show impressive demos. Products ship with "learning" in the tagline.
+
+Ask them one question: **How do you know it works?**
+
+You'll get:
+- "It feels smarter"
+- "Users report better results"
+- "The agent remembers things now"
+
+That's not evidence. That's vibes.
+
+Here's what a real answer looks like:
+
+> "We track Repeated Mistake Rate (RMR) across sessions. Our null hypothesis is that the system makes no difference. After 50 sessions, RMR decreased from 34% to 12% (p < 0.01). The effect size is 0.65. Here's the data."
+
+If you can't say something like that, you don't have agent learning. You have a demo.
+
+---
+
+## The Claim
+
+**buildlog** makes a falsifiable claim:
+
+> **H₀ (Null Hypothesis):** buildlog makes no measurable difference to agent behavior.
+>
+> **H₁ (Alternative):** Agents using buildlog-learned rules have lower Repeated Mistake Rate than baseline.
+
+We provide the infrastructure to **reject or fail to reject** this hypothesis with your own data.
+
+If buildlog doesn't work, the numbers will show it. That's the point.
+
+---
+
+## The Metric: Repeated Mistake Rate (RMR)
+
+```
+RMR = (Mistakes that match previous mistakes) / (Total mistakes logged)
+```
+
+A mistake "matches" if it has the same semantic signature—same error class, similar description, same root cause showing up again.
+
+**Why RMR?**
+- **Observable**: You can count it
+- **Attributable**: Lower RMR after rule injection = signal
+- **Meaningful**: Repeating mistakes is the actual pain point
+
+RMR is not the only metric that matters. But it's one we can measure, and measurement is where science starts.
+
+---
+
+## The Mechanism
+
+buildlog uses **contextual bandits** to select which rules to surface.
+
+```
+┌─────────────────────────────────────────────────────────────────┐
+│                    CONTEXTUAL BANDIT SETUP                      │
+├─────────────────────────────────────────────────────────────────┤
+│                                                                 │
+│  Context (c):     Error class, file type, task category        │
+│  Arms (a):        Candidate rules to surface                   │
+│  Reward (r):      Human feedback (👍 helpful / 👎 not helpful) │
+│                                                                 │
+│  Policy π(c) → a:  Thompson Sampling with Beta-Bernoulli       │
+│                                                                 │
+│  Objective:       Minimize regret = Σ(r* - r_chosen)           │
+│                                                                 │
+└─────────────────────────────────────────────────────────────────┘
+```
+
+**Arms** = learned rules (from buildlog entries, code reviews, explicit teaching)
+
+**Context** = what kind of problem you're working on
+
+**Reward** = did surfacing this rule actually help?
+
+The system explores (tries uncertain rules) and exploits (uses proven rules) based on accumulated evidence. Thompson Sampling provides theoretical guarantees: O(√(KT log K)) regret bounds.
+
+This isn't magic. It's a well-understood framework with decades of research. We're applying it to agent rule selection.
+
+---
+
+## The Pipeline
+
+buildlog captures signal at every stage:
+
+```mermaid
+flowchart LR
+    A["Work Sessions"] --> B["Structured Entries"]
+    B --> C["Extracted Rules"]
+    C --> D["Bandit Selection"]
+    D --> E["Rule Surfaced"]
+    E --> F["Human Feedback"]
+    F --> G["Posterior Update"]
+    G --> D
+
+    style F fill:#ff6b6b,color:#fff
+    style G fill:#4ecdc4,color:#fff
+```
+
+### Stage 1: Capture
+Document your work. Include the fuckups—they're the most valuable signal.
+
+```bash
+buildlog new auth-api
+# Edit the markdown, document what happened
+```
+
+### Stage 2: Extract
+Pull structured rules from your entries.
+
+```bash
+buildlog distill    # Extract patterns
+buildlog skills     # Deduplicate into rules
+```
+
+### Stage 3: Promote
+Surface rules to your agent via CLAUDE.md, settings.json, or Agent Skills.
+
+```bash
+buildlog promote --target skill
+```
+
+### Stage 4: Measure
+Track what happens when rules are active.
+
+```bash
+buildlog experiment start --error-class "type-errors"
+# ... work session ...
+buildlog experiment log-mistake --error-class "type-errors" \
+  --description "Forgot to handle null case"
+buildlog experiment end
+buildlog experiment report
+```
+
+### Stage 5: Learn
+Log reward signals when rules help (or don't).
+
+```python
+# Via MCP
+buildlog_log_reward(
+    skill_id="arch-123",
+    reward=1,           # 1 = helped, 0 = didn't help
+    context="type-errors",
+    outcome="Caught the bug before commit"
+)
+```
+
+---
+
+## Experiment Infrastructure
+
+buildlog ships with infrastructure to run actual experiments:
+
+```bash
+# Start a tracked session
+buildlog experiment start --error-class "api-design"
+
+# Log mistakes as they happen
+buildlog experiment log-mistake \
+  --error-class "api-design" \
+  --description "Returned 200 for error case"
+
+# End session
+buildlog experiment end
+
+# Get metrics
+buildlog experiment metrics
+
+# Full report across all sessions
+buildlog experiment report
+```
+
+The report includes:
+- Total sessions, total mistakes
+- Repeat rate (RMR)
+- Mistakes by error class
+- Rules that correlate with corrections
+
+This is the data you need to make claims.
+
+---
+
+## Quick Start
+
+```bash
+# Install
+pip install buildlog
+
+# Initialize
+buildlog init
+
+# Create your first entry
+buildlog new my-feature
+
+# After a few entries, extract rules
+buildlog distill
+buildlog skills
+
+# Start measuring
+buildlog experiment start
+# ... work ...
+buildlog experiment end
+buildlog experiment report
+```
+
+### MCP Server (Claude Code Integration)
+
+```bash
+pip install buildlog[mcp]
+```
+
+Add to `~/.claude/settings.json`:
+
+```json
+{
+  "mcpServers": {
+    "buildlog": {
+      "command": "buildlog-mcp"
+    }
+  }
+}
+```
+
+Available tools:
+| Tool | Purpose |
+|------|---------|
+| `buildlog_status` | View rules by category and confidence |
+| `buildlog_promote` | Surface rules to agent |
+| `buildlog_reject` | Mark false positives |
+| `buildlog_diff` | Rules pending review |
+| `buildlog_learn_from_review` | Extract rules from code review |
+| `buildlog_log_reward` | Record reward signal for bandit |
+| `buildlog_rewards` | View reward history |
+| `buildlog_start_session` | Begin tracked experiment session |
+| `buildlog_end_session` | End session |
+| `buildlog_log_mistake` | Record mistake during session |
+| `buildlog_session_metrics` | Get session statistics |
+| `buildlog_experiment_report` | Full experiment report |
+
+---
+
+## What This Is Not
+
+**This is not AGI.** This is not "agents that truly learn." This is not a revolution.
+
+This is:
+- A structured way to capture engineering knowledge
+- A bandit framework for rule selection
+- Infrastructure to measure whether it works
+
+Boring? Maybe. But boring things that work beat exciting things that don't.
+
+---
+
+## The Falsification Protocol
+
+Want to test whether buildlog actually helps? Here's the protocol:
+
+1. **Baseline**: Run N sessions without buildlog rules active. Log mistakes.
+2. **Treatment**: Run N sessions with buildlog rules active. Log mistakes.
+3. **Compare**: Calculate RMR for both conditions.
+4. **Statistical test**: Two-proportion z-test or chi-squared.
+5. **Report**: Effect size, confidence interval, p-value.
+
+If p > 0.05, we fail to reject the null. buildlog didn't help. That's a valid outcome.
+
+If p < 0.05, we have evidence of an effect. How big? Check the effect size.
+
+This is how you know. Not vibes. Data.
+
+---
+
+## Theoretical Foundations
+
+For the technically curious:
+
+| Concept | Application in buildlog |
+|---------|------------------------|
+| **Thompson Sampling** | Rule selection under uncertainty |
+| **Beta-Bernoulli model** | Posterior updates from binary reward |
+| **Contextual bandits** | Context-dependent rule selection |
+| **Regret bounds** | O(√(KT log K)) theoretical guarantee |
+| **Semantic hashing** | Mistake deduplication for RMR |
+
+We're not inventing new math. We're applying proven frameworks to a new domain.
+
+---
+
+## Honest Limitations
+
+Things we don't have figured out yet:
+
+- **Credit assignment**: When multiple rules are active, which one helped?
+- **Non-stationarity**: Developer skill changes over time
+- **Cold start**: New rules have high uncertainty
+- **Context representation**: What features actually matter?
+
+These are hard problems. We have directional ideas, not solutions. If you're a researcher working on bandit algorithms or causal inference, we'd love to talk.
+
+---
+
+## Philosophy
+
+1. **Falsifiability over impressiveness** - If you can't prove it wrong, it's not a claim
+2. **Measurement over intuition** - "Feels better" is not evidence
+3. **Mechanisms over magic** - Explain how it works or admit you don't know
+4. **Boring over exciting** - Proven frameworks beat novel demos
+5. **Honesty over marketing** - State limitations. Invite scrutiny.
+
+---
+
+## Contributing
+
+We're especially interested in:
+- Better context representations for the bandit
+- Credit assignment approaches
+- Statistical methodology improvements
+- Real-world experiment results (positive or negative)
+
+```bash
+git clone https://github.com/Peleke/buildlog-template
+cd buildlog-template
+pip install -e ".[dev]"
+pytest
+```
+
+---
+
+## License
+
+MIT License — see [LICENSE](./LICENSE)
+
+---
+
+<div align="center">
+
+**"Agent learning" without measurement is just prompt engineering with extra steps.**
+
+**buildlog is measurement.**
+
+[Back to top](#buildlog)
+
+</div>

{buildlog-0.4.0.dist-info → buildlog-0.5.0.dist-info}/RECORD

@@ -1,30 +1,30 @@
 buildlog/__init__.py,sha256=FrgjyZhC19YyB40rOXHJWTA4xKWx2Yn2heIhVraaZ7A,90
-buildlog/cli.py,sha256=cmg77_RVJx8mdtStApS1KXxBUUB8Id6psHZjtHo33iE,14350
+buildlog/cli.py,sha256=UY95mCJNCdN7KRkUJiebSFbVvh0q7hk3eeom0mGhLz0,28433
 buildlog/confidence.py,sha256=EOkPxIH1_y7k6B3Hl7Wn0iR2qK_lumvOyyyqUdafXVY,9382
 buildlog/distill.py,sha256=fqXW_YyBFIFhwIWhnR-TQ7U65gypqG-mcAzNBr-qaag,11262
 buildlog/embeddings.py,sha256=vPydWjJVkYp172zFou-lJ737qsu6vRMQAMs143RGIpA,12364
 buildlog/skills.py,sha256=C_tqspj9lVRkjF-g-KpKt8WoQ-Xq0e0l1VtXly_Naks,30055
 buildlog/stats.py,sha256=2WdHdmzUNGobtWngmm9nA_UmqM7DQeAnZL8_rLQN8aw,13256
-buildlog/core/__init__.py,sha256=ifByeEE3sM13EHWIKNAOjNc3zx_4qYEfyXdB4Woumuw,594
-buildlog/core/operations.py,sha256=Q9wX8kfnpDBAC8ySHYXZi3vA9w0vuOyzOYl6JYB6YyM,22006
+buildlog/core/__init__.py,sha256=U7FS1CLUfnJ4AHGlgc3-F5KS7bepRMFD7TTvp8vKfQM,1301
+buildlog/core/operations.py,sha256=iXrhBd7cbr9wMgpiIl4Q-7WIQOgQCWBepQibzgoY0Gg,51751
 buildlog/mcp/__init__.py,sha256=jCLNUkYFrDcPd5dY9xbaaVANl-ZzdPim1BykgGY7f7U,334
-buildlog/mcp/server.py,sha256=vlfXlRvGfoXBhkHv2y5e1jc7JH2Hkv7G1obSdxlQ14w,583
-buildlog/mcp/tools.py,sha256=Rfk3dS4bLflvaUa5O2YLdvBx354ZQ_Lfy-Uoaz_AjGo,4397
+buildlog/mcp/server.py,sha256=9AOvweRIQ-BJVbNc7OA6iRQ0lC9OjbUFsCHhU9FvJgk,1064
+buildlog/mcp/tools.py,sha256=Zz_y4IwquSe1jwIb9EFlNO_Xc8GgR8f-OBj-yXdWUKk,12228
 buildlog/render/__init__.py,sha256=VxJWEmcX7pSiC3W-ytsHv9gNVUr4hJrVHBW084kEnAI,1993
 buildlog/render/base.py,sha256=gQfvOsH1zttAo10xtEyNsAbqZ4NRSPiDihO-aiGgTsw,533
 buildlog/render/claude_md.py,sha256=Z_E6MbJyVM_hJSoB4KL2rvbt5UEQHekTpJijj106lsI,2624
 buildlog/render/settings_json.py,sha256=4DS5OWksPrFCa7MIgWIu0t4rxYmItpMdGfTqMX3aMNs,2495
 buildlog/render/skill.py,sha256=_7umIS1Ms1oQ2_PopYueFjX41nMq1p28yJp6DhXFdgU,5981
 buildlog/render/tracking.py,sha256=6O0RIU-1gjVG-_S5dmXLz6RCMsQoHOR2u5___UpqXEo,1294
-buildlog-0.4.0.data/data/share/buildlog/copier.yml,sha256=A-1JKV59kOe0BQosGUBgRCg7iQozP_qyA3zfoHwpBKY,927
-buildlog-0.4.0.data/data/share/buildlog/post_gen.py,sha256=XFlo40LuPpAsBhIRRRtHqvU3_5POss4L401hp35ijhw,1744
-buildlog-0.4.0.data/data/share/buildlog/template/buildlog/.gitkeep,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-buildlog-0.4.0.data/data/share/buildlog/template/buildlog/2026-01-01-example.md,sha256=7x9sKmydfmfKyNz9hV7MtYnQJuBwbxNanbPOcpQDDZQ,7040
-buildlog-0.4.0.data/data/share/buildlog/template/buildlog/BUILDLOG_SYSTEM.md,sha256=osclytWwl5jUiTgSpuT4cT3h3oPvCkZ5GPCnFuJZNcY,3802
-buildlog-0.4.0.data/data/share/buildlog/template/buildlog/_TEMPLATE.md,sha256=CUvxgcx1-9XT_EdQ8e_vnuPq_h-u1uhXJgForJU2Pso,2932
-buildlog-0.4.0.data/data/share/buildlog/template/buildlog/assets/.gitkeep,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-buildlog-0.4.0.dist-info/METADATA,sha256=adRpA8HpJlbAFLk7UCbHPqhWXNU7Xp9WEfYo-e4GqXI,30019
-buildlog-0.4.0.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
-buildlog-0.4.0.dist-info/entry_points.txt,sha256=BMFclPOomp_sgaa0OqBg6LfqCMlqzjZV88ww5TrPPoo,87
-buildlog-0.4.0.dist-info/licenses/LICENSE,sha256=fAgt-akug9nAwIj6M-SIf8u3ck-T7pJTwfmy9vWYASk,1074
-buildlog-0.4.0.dist-info/RECORD,,
+buildlog-0.5.0.data/data/share/buildlog/copier.yml,sha256=A-1JKV59kOe0BQosGUBgRCg7iQozP_qyA3zfoHwpBKY,927
+buildlog-0.5.0.data/data/share/buildlog/post_gen.py,sha256=XFlo40LuPpAsBhIRRRtHqvU3_5POss4L401hp35ijhw,1744
+buildlog-0.5.0.data/data/share/buildlog/template/buildlog/.gitkeep,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+buildlog-0.5.0.data/data/share/buildlog/template/buildlog/2026-01-01-example.md,sha256=7x9sKmydfmfKyNz9hV7MtYnQJuBwbxNanbPOcpQDDZQ,7040
+buildlog-0.5.0.data/data/share/buildlog/template/buildlog/BUILDLOG_SYSTEM.md,sha256=osclytWwl5jUiTgSpuT4cT3h3oPvCkZ5GPCnFuJZNcY,3802
+buildlog-0.5.0.data/data/share/buildlog/template/buildlog/_TEMPLATE.md,sha256=CUvxgcx1-9XT_EdQ8e_vnuPq_h-u1uhXJgForJU2Pso,2932
+buildlog-0.5.0.data/data/share/buildlog/template/buildlog/assets/.gitkeep,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+buildlog-0.5.0.dist-info/METADATA,sha256=KTL5nTO8ikp5IhtMgCzVO-1KYVXw-iyZ_oo4RVDk2nU,13210
+buildlog-0.5.0.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
+buildlog-0.5.0.dist-info/entry_points.txt,sha256=BMFclPOomp_sgaa0OqBg6LfqCMlqzjZV88ww5TrPPoo,87
+buildlog-0.5.0.dist-info/licenses/LICENSE,sha256=fAgt-akug9nAwIj6M-SIf8u3ck-T7pJTwfmy9vWYASk,1074
+buildlog-0.5.0.dist-info/RECORD,,