buildlog 0.8.0__py3-none-any.whl → 0.10.0__py3-none-any.whl

buildlog/constants.py

@@ -0,0 +1,121 @@
+"""Shared constants for buildlog, including the CLAUDE.md integration section."""
+
+CLAUDE_MD_BUILDLOG_SECTION = """
+## buildlog Integration
+
+buildlog is configured as an MCP server. Use the tools below to maintain
+a learning loop: write entries, extract skills, run gauntlet reviews, and
+track experiments.
+
+### Always On: Commit -> Gauntlet -> Learn Loop
+
+After every significant commit or feature completion:
+
+1. **Create/update entry**: `buildlog_entry_new(slug="what-you-built")`
+2. **Run gauntlet review**: `buildlog_gauntlet_rules()` to load rules, then review code against them
+3. **Process findings**: `buildlog_gauntlet_issues(issues=[...])` to categorize and persist learnings
+4. **Fix criticals/majors**, re-run until clean or accept risk: `buildlog_gauntlet_accept_risk(remaining_issues=[...])`
+5. **Log reward**: `buildlog_log_reward(outcome="accepted")` when work is approved
+
+### Skill Extraction & Promotion Workflow
+
+1. `buildlog_status()` — see extracted skills by category
+2. `buildlog_diff()` — see skills pending review
+3. `buildlog_promote(skill_ids=[...], target="claude_md")` — surface to agent rules
+4. `buildlog_reject(skill_ids=[...])` — mark false positives
+
+### Gauntlet Review Loop Workflow
+
+1. `buildlog_gauntlet_rules()` — load reviewer persona rules
+2. Review code against rules, collect issues
+3. `buildlog_gauntlet_issues(issues=[...])` — categorize, persist learnings, get next action
+4. If action="fix_criticals": fix and re-run
+5. If action="checkpoint_majors"/"checkpoint_minors": ask user
+6. `buildlog_gauntlet_accept_risk(remaining_issues=[...])` — accept remaining
+7. `buildlog_learn_from_review(issues=[...])` — persist learnings
+
+### Reward Signal / Thompson Sampling Workflow
+
+1. `buildlog_log_reward(outcome="accepted"|"revision"|"rejected")` — explicit feedback
+2. `buildlog_rewards()` — view reward history and statistics
+3. `buildlog_bandit_status()` — see which rules the bandit favors
+
+### Session Tracking & Experiment Workflow
+
+1. `buildlog_experiment_start(error_class="missing_test")` — begin tracked session
+2. `buildlog_log_mistake(error_class="...", description="...")` — log mistakes
+3. `buildlog_experiment_end()` — end session, calculate metrics
+4. `buildlog_experiment_metrics()` — per-session or aggregate stats
+5. `buildlog_experiment_report()` — comprehensive report
+
+### Tool Reference (29 tools)
+
+**Commit & Entries:**
+- `buildlog_commit(message, git_args, slug, no_entry)` — git commit with auto buildlog entry
+- `buildlog_entry_new(slug, entry_date, quick)` — create entry
+- `buildlog_entry_list()` — list all entries
+- `buildlog_overview()` — project state at a glance
+
+**Skill Management:**
+- `buildlog_status(min_confidence="low")` — extracted skills
+- `buildlog_promote(skill_ids, target="claude_md")` — promote to agent rules
+- `buildlog_reject(skill_ids)` — reject false positives
+- `buildlog_diff()` — pending skills
+- `buildlog_distill(since, category)` — extract patterns from entries
+- `buildlog_skills(since, min_frequency)` — generate skill set from entries
+- `buildlog_stats(since, detailed)` — buildlog statistics and insights
+
+**Gauntlet Review:**
+- `buildlog_gauntlet_prompt(target, personas)` — generate review prompt with rules
+- `buildlog_gauntlet_loop(target, personas, max_iterations, stop_at, auto_gh_issues)` — full loop config
+- `buildlog_gauntlet_rules(persona, format)` — load reviewer rules
+- `buildlog_gauntlet_issues(issues, iteration)` — process findings
+- `buildlog_gauntlet_accept_risk(remaining_issues)` — accept risk
+- `buildlog_gauntlet_list_personas()` — list available reviewer personas
+- `buildlog_gauntlet_generate(source_text, persona, dry_run)` — generate rules from source text
+
+**Review Learning:**
+- `buildlog_learn_from_review(issues, source)` — persist review learnings
+
+**Reward & Bandit:**
+- `buildlog_log_reward(outcome, rules_active)` — log feedback
+- `buildlog_rewards(limit)` — reward history
+- `buildlog_bandit_status(context)` — bandit state
+
+**Experiments:**
+- `buildlog_experiment_start(error_class)` — start session
+- `buildlog_experiment_end()` — end session
+- `buildlog_log_mistake(error_class, description)` — log mistake
+- `buildlog_experiment_metrics(session_id)` — metrics
+- `buildlog_experiment_report()` — full report
+
+**Project Setup:**
+- `buildlog_init(no_mcp, no_claude_md)` — initialize buildlog in project
+- `buildlog_update()` — update buildlog template to latest
+
+### When to Use Each Tool
+
+- **At session start**: `buildlog_overview()` for context
+- **During active dev**: `buildlog_entry_new()` to document work
+- **After commits**: `buildlog_commit()` or `buildlog_gauntlet_prompt()` + review + `buildlog_gauntlet_issues()`
+- **After review approval**: `buildlog_log_reward(outcome="accepted")`
+- **For learning**: `buildlog_distill()`, `buildlog_skills()`, `buildlog_stats()`
+- **For skill promotion**: `buildlog_status()` -> `buildlog_diff()` -> `buildlog_promote()`
+- **To accept risk**: `buildlog_gauntlet_accept_risk()`
+- **For experiments**: `buildlog_experiment_start()` -> work -> `buildlog_experiment_end()`
+
+### Integration with Commits
+
+`buildlog commit -m "message"` wraps git commit and auto-logs to today's entry.
+
+### Reference Files
+
+- `buildlog/.buildlog/promoted.json` — promoted skill IDs
+- `buildlog/.buildlog/rejected.json` — rejected skill IDs
+- `buildlog/.buildlog/review_learnings.json` — review-based learnings
+- `buildlog/.buildlog/reward_events.jsonl` — reward signals
+- `buildlog/.buildlog/sessions.jsonl` — session tracking
+- `buildlog/.buildlog/mistakes.jsonl` — mistake tracking
+- `buildlog/.buildlog/active_session.json` — current session
+- `buildlog/bandit_state.jsonl` — Thompson Sampling state
+"""

buildlog/core/__init__.py

@@ -1,14 +1,23 @@
 """Core operations for buildlog skill management."""
 
 from buildlog.core.operations import (
+    CommitResult,
+    CreateEntryResult,
     DiffResult,
     EndSessionResult,
     GauntletAcceptRiskResult,
+    GauntletGenerateResult,
+    GauntletLoopConfigResult,
     GauntletLoopResult,
+    GauntletPromptResult,
+    GauntletRulesResult,
+    InitResult,
     LearnFromReviewResult,
+    ListEntriesResult,
     LogMistakeResult,
     LogRewardResult,
     Mistake,
+    OverviewResult,
     PromoteResult,
     RejectResult,
     ReviewIssue,
@@ -19,22 +28,33 @@ from buildlog.core.operations import (
     SessionMetrics,
     StartSessionResult,
     StatusResult,
+    UpdateResult,
+    commit,
+    create_entry,
     diff,
     end_session,
     find_skills_by_ids,
     gauntlet_accept_risk,
+    gauntlet_generate,
+    gauntlet_loop_config,
     gauntlet_process_issues,
+    generate_gauntlet_prompt,
     get_bandit_status,
     get_experiment_report,
+    get_gauntlet_rules,
+    get_overview,
     get_rewards,
     get_session_metrics,
+    init_buildlog,
     learn_from_review,
+    list_entries,
     log_mistake,
     log_reward,
     promote,
     reject,
     start_session,
     status,
+    update_buildlog,
 )
 
 __all__ = [
@@ -58,6 +78,11 @@ __all__ = [
     # Gauntlet loop
     "GauntletLoopResult",
     "GauntletAcceptRiskResult",
+    # Entry & overview
+    "GauntletRulesResult",
+    "OverviewResult",
+    "CreateEntryResult",
+    "ListEntriesResult",
     "status",
     "promote",
     "reject",
@@ -76,4 +101,23 @@ __all__ = [
     # Gauntlet loop operations
     "gauntlet_process_issues",
     "gauntlet_accept_risk",
+    # Entry & overview operations
+    "get_gauntlet_rules",
+    "get_overview",
+    "create_entry",
+    "list_entries",
+    # P0: Gauntlet loop
+    "CommitResult",
+    "GauntletPromptResult",
+    "GauntletLoopConfigResult",
+    "commit",
+    "generate_gauntlet_prompt",
+    "gauntlet_loop_config",
+    # P2: Nice-to-have
+    "GauntletGenerateResult",
+    "InitResult",
+    "UpdateResult",
+    "gauntlet_generate",
+    "init_buildlog",
+    "update_buildlog",
 ]