v0.12.0

What's Changed

• fix(release): resolve Homebrew formula prefix portably by @forrestbthomas in #170
• feat(sessions): pi-run sessions fleet view + --heal connection-flap events (W11) by @forrestbthomas in #171
• docs(backlog): re-open COST-1 as static task->tier router (EPIC-3) by @forrestbthomas in #172
• chore(pi): add pi-teams 0.9.14 (agent-teams try-out) by @forrestbthomas in #173
• v0.12.0: pi-bulletin wiring + sessions --heal + SEC-1/HEAL-6 (reconcile + release prep) by @forrestbthomas in #176

Full Changelog: v0.11.1...v0.12.0

v0.11.1

What's Changed

• fix(ci): install linuxbrew before the REL-3 brew-verify step by @forrestbthomas in #167
• fix(ollama): install the provider extension globally via pi-run setup by @forrestbthomas in #168
• release(v0.11.1): record Ollama global-extension fix by @forrestbthomas in #169

Full Changelog: v0.11.0...v0.11.1

v0.11.0

What's Changed

• feat(eval-13): flake-aware cost gate — single-run cost spike is a flake, not a failure by @forrestbthomas in #105
• feat(eval-14): benchmark provenance parity — ci-benchmark scorecard carries live-schema provenance by @forrestbthomas in #106
• feat(eval-15): split-seam verification — seam contract doc + hermetic dry-run by @forrestbthomas in #107
• feat(eval-8): judge stabilization — deterministic conversion + majority-of-3 + judge-pass bugfix by @forrestbthomas in #108
• feat(rel-3/rel-4): post-release brew verify + Node-drift guard by @forrestbthomas in #109
• docs(tax-2): flag/env-var prune audit — no dead surfaces, watchdog env vars added to README table by @forrestbthomas in #110
• feat(eval-6): agentic tool-using case family — exercise the harness tool surface by @forrestbthomas in #111
• feat(gov-1): PM-layer drift guard — EPICS/BACKLOG/tag/changelog self-audit by @forrestbthomas in #112
• docs(audit): Wave 1 P0 — truth restoration (count sweep, guards, ghost purge, spec statuses, PM reconciliation) by @forrestbthomas in #113
• docs(audit): Wave 2 P1 — status & sequencing truth by @forrestbthomas in #114
• docs(audit): Wave 3+4 P2/P3 — hygiene (README Skills/Layout, requirements comment, stamps, PR template, AGENTS polish) by @forrestbthomas in #115
• docs(scope): close docs-audit contract — 3 waves logged by @forrestbthomas in #116
• docs(readme): audit leftovers — real task.json keys, pi-mcp-adapter package, eval/CI knob rows by @forrestbthomas in #117
• docs(backlog): re-rank queue — shipped items out of open ranks by @forrestbthomas in #118
• docs(governance): GOV-2 relocation + spec archive by @forrestbthomas in #119
• test(docs-drift): cwd-independent live-results assertion + GOV-3 backlog item by @forrestbthomas in #120
• docs(scope): close + archive GOV-2 SCOPE.md by @forrestbthomas in #121
• docs(contributing): OSS-2 contributor on-ramp v2 by @forrestbthomas in #122
• ci(python-quick): wire docs/pm drift guards into CI (GOV-3) by @forrestbthomas in #123
• docs(pm): GOV-3 reconciliation — shipped row, STATUS regen, CHANGELOG by @forrestbthomas in #125
• docs(scope): close + archive GOV-3 SCOPE.md by @forrestbthomas in #126
• docs(scope): archived GOV-3 contract status is CLOSED by @forrestbthomas in #127
• docs(pm): apply whats-next debate change-set (BACKLOG/EPICS/STATUS) by @forrestbthomas in #128
• feat(eval): EVAL-16 pilot — score-delta module + report-only CI wiring by @forrestbthomas in #129
• docs(pm): EVAL-16 pilot reconciliation by @forrestbthomas in #130
• fix(eval): judge-case metrics wiring + nightly honesty fixes by @forrestbthomas in #131
• feat(eval): coding-055 — debugging/root-cause eval case (dataset 54→55) by @forrestbthomas in #132
• fix(eval): coding-052 self-healing grader (derived count) by @forrestbthomas in #133
• fix(security): bump .pi/npm transitive deps — 16/17 dependabot alerts closed by @forrestbthomas in #134
• chore(cleanup): remove lore tooling — .lore.md/.lore.json + references by @forrestbthomas in #135
• feat(harness): remove @loreai/pi — closes the last dependabot alert by @forrestbthomas in #136
• docs(backlog): add MEM-1 — memory-engine hook spike + decision by @forrestbthomas in #137
• docs(decision): MEM-1 — memory-engine spike closed as decided (no build) by @forrestbthomas in #138
• fix(eval): judge-graded cases no longer gate as incomplete by @forrestbthomas in #139
• data(live-baseline): EVAL-12 re-baseline — 17 → 55 cases (0 unbaselined) by @forrestbthomas in #140
• fix(eval): EVAL-13 cost gate is median-shift-only by @forrestbthomas in #141
• docs(pm): EVAL-12 reconciliation — re-baseline shipped, queue re-ranked by @forrestbthomas in #142
• docs(readme): minimal front-door rewrite — differentiators + real scorecard + visuals by @forrestbthomas in #143
• docs(readme): drop the private context-engine mention from the memory statement by @forrestbthomas in #144
• docs(hygiene): stale counts + quick references by @forrestbthomas in #145
• docs(readme): fix mermaid rendering (quoted labels) by @forrestbthomas in #146
• feat(eval): EVAL-18 — variance-aware gate (run-step band) by @forrestbthomas in #147
• data(baseline): coding-055 re-baselined at its true 10-run rate (0.2) by @forrestbthomas in #148
• feat(eval): coding-055 prompt scaffold — measured score 0.2 → 0.8 by @forrestbthomas in #149
• docs: v0.11.0 release prep — governance snapshot + release runbook by @forrestbthomas in #150
• docs: honest reframe — no-moat positioning, per 7-persona debate by @forrestbthomas in #151
• docs: qualify 'yours' — provider-neutral, NOT runtime-neutral by @forrestbthomas in #152
• docs: dogfood posture — personal agent-improvement loop, public as demo by @forrestbthomas in #153
• data(baseline): coding-049 cost re-baselined — cost-of-correct, not a regression by @forrestbthomas in #154
• feat(eval): EVAL-17 — chat-session runner + multi-turn cases (056/058), 057 parked by @forrestbthomas in #155
• fix(eval): chat cases live in their own jsonl surface — restore manifest parity by @forrestbthomas in #156
• data(baseline): coding-017 + coding-052 re-baselined (stale snapshots) by @forrestbthomas in #158
• chore(ci): raise the nightly live job timeout 60 → 90 min by @forrestbthomas in #159
• fix(provider): derive the Ollama /model picker from the local daemon catalog by @forrestbthomas in #160
• docs(hygiene): Ollama keyless row + superpowers skill check path/count fixes by @forrestbthomas in #165
• release(v0.11.0): changelog flip + gate re-typed by recorded owner decision by @forrestbthomas in #166

Full Changelog: v0.10.0...v0.11.0

v0.10.0

What's Changed

• ci(release): tag must be ancestor of main — guard + ritual + script by @forrestbthomas in #71
• docs: high-level roadmap workflow — STATUS.md snapshot, cycle ritual, traceability by @forrestbthomas in #72
• docs(backlog): RICE prioritization pass — 12 candidates scored by @forrestbthomas in #73
• docs: codify git hygiene — ff-only sync, never routine reset --hard by @forrestbthomas in #74
• feat(doctor): verify non-interactive launch env (backlog #1) by @forrestbthomas in #75
• docs(roadmap): promote BACKLOG #1 to W5 (upstream per-tool-call timeout) by @forrestbthomas in #76
• feat(w5): adopt upstream run-level timeout machinery — pin pi-subagents 0.48.0 by @forrestbthomas in #77
• docs: agent navigation map, post-merge docs-drift guard, cruft cleanup by @forrestbthomas in #79
• docs(backlog): watchdog liveness heartbeat idea by @forrestbthomas in #80
• docs(roadmap): reconcile W5 status after upstream submission by @forrestbthomas in #81
• docs(roadmap): W5 Part B merged upstream by @forrestbthomas in #82
• feat(eval): surface PI_SELF_HEAL events in scorecard + enable in CI (W6) by @forrestbthomas in #83
• docs(roadmap): W6 shipped — close DoD after manual nightly verification by @forrestbthomas in #84
• docs(pm): add EPICS layer + EPIC-1 eval-suite backlog by @forrestbthomas in #85
• docs(pm): consolidate all planned work into 5 epics / 22 backlog items by @forrestbthomas in #86
• feat(eval): evidence artifacts on every outcome + flake-aware gate (W7) by @forrestbthomas in #87
• docs(roadmap): W7 shipped — close DoD by @forrestbthomas in #88
• feat(eval): dataset versioning + scorecard provenance (W8) by @forrestbthomas in #89
• docs(roadmap): W8 shipped — close DoD by @forrestbthomas in #90
• feat(eval): self-heal events in provider scorecard (W9) by @forrestbthomas in #91
• docs(roadmap): W9 shipped — close DoD; W10 (EVAL-5) active by @forrestbthomas in #92
• feat(eval): dataset growth Batch A — 20 to 35 cases (W10) by @forrestbthomas in #93
• feat(eval): dataset growth Batch B — 35 to 50 cases (W10) by @forrestbthomas in #94
• feat(eval): benchmark batch — 3 edit-based Docker tasks (W10) by @forrestbthomas in #95
• docs(backlog): add EVAL-12 — live re-baseline follow-up (EPIC-1) by @forrestbthomas in #96
• fix(eval): size live-suite timeouts for the 50-case dataset (EVAL-12) by @forrestbthomas in #97
• docs(charter): define project boundary — one product, the harness (scope debate) by @forrestbthomas in #98
• docs(changelog): record project charter under Unreleased by @forrestbthomas in #99
• feat(cut-list): remove consumerless MCP server, OTel exporter, pdf2txt (scope debate) by @forrestbthomas in #100
• docs(pm): apply maturity-scan change-set to EPICS/BACKLOG/STATUS by @forrestbthomas in #101
• feat(oss-1): canonical module path + identity alignment + install contract check by @forrestbthomas in #102
• docs(pm): add release milestones v0.10.0 -> v1.0.0 by @forrestbthomas in #103
• release(v0.10.0): prep — version-truth stamp, ledger repair, changelog cut by @forrestbthomas in #104

Full Changelog: v0.9.2...v0.10.0

v0.9.2

What's Changed

• chore: session finalization — docs reality pass, live-eval baseline, W2/W3 closures, changelog + nightly env by @forrestbthomas in #64
• feat(security): pin pi-subagents + owner-only artifact perms (backlog #2/#3) by @forrestbthomas in #65

Full Changelog: v0.9.1...v0.9.2

What's Changed

• chore: session finalization — docs reality pass, live-eval baseline, W2/W3 closures, changelog + nightly env by @forrestbthomas in #64
• feat(security): pin pi-subagents + owner-only artifact perms (backlog #2/#3) by @forrestbthomas in #65
• docs: v0.9.2 release notes (CHANGELOG + ROADMAP W2 note) by @forrestbthomas in #70

Full Changelog: v0.9.1...v0.9.2

v0.9.1

What's Changed

• fix: nightly live-eval attribution — --cost-mode + explicit judge model by @forrestbthomas in #50
• fix: missing global Pi settings is not a config-check failure by @forrestbthomas in #51
• ci: pin nvm Node 22.19.0 (nvm install latest fails on runners) by @forrestbthomas in #52
• fix: verified model-tier ids + honest run_pi_print failure detection by @forrestbthomas in #53
• fix: GEval requires evaluation_params in deepeval 4.1.7 by @forrestbthomas in #54
• fix: de-brittle E2E factorial heuristic (live nightly) by @forrestbthomas in #55
• fix: self-contained eval report writer (replaces broken pytest-json-report) by @forrestbthomas in #56
• fix: install real pi coding agent + robust graders (nightly was grading '3') by @forrestbthomas in #57
• fix: pin scoped pi-coding-agent package (ledger lane) by @forrestbthomas in #58
• feat: prevent agent hang on interactive git editors/pagers by @forrestbthomas in #59
• docs: project-management layer — ROADMAP, BACKLOG, PM skills by @forrestbthomas in #60
• docs(spec): self-healing agent runs — W1 design by @forrestbthomas in #61
• feat: self-healing agent runs — watchdog, process-group kill, git recovery, escalation by @forrestbthomas in #62

Full Changelog: v0.9.0...v0.9.1

What's Changed

• fix: nightly live-eval attribution — --cost-mode + explicit judge model by @forrestbthomas in #50
• fix: missing global Pi settings is not a config-check failure by @forrestbthomas in #51
• ci: pin nvm Node 22.19.0 (nvm install latest fails on runners) by @forrestbthomas in #52
• fix: verified model-tier ids + honest run_pi_print failure detection by @forrestbthomas in #53
• fix: GEval requires evaluation_params in deepeval 4.1.7 by @forrestbthomas in #54
• fix: de-brittle E2E factorial heuristic (live nightly) by @forrestbthomas in #55
• fix: self-contained eval report writer (replaces broken pytest-json-report) by @forrestbthomas in #56
• fix: install real pi coding agent + robust graders (nightly was grading '3') by @forrestbthomas in #57
• fix: pin scoped pi-coding-agent package (ledger lane) by @forrestbthomas in #58
• feat: prevent agent hang on interactive git editors/pagers by @forrestbthomas in #59
• docs: project-management layer — ROADMAP, BACKLOG, PM skills by @forrestbthomas in #60
• docs(spec): self-healing agent runs — W1 design by @forrestbthomas in #61
• feat: self-healing agent runs — watchdog, process-group kill, git recovery, escalation by @forrestbthomas in #62
• docs: v0.9.1 release notes by @forrestbthomas in #63

Full Changelog: v0.9.0...v0.9.1

v0.9.0

What's Changed

• docs: deterministic eval hardening design spec by @forrestbthomas in #40
• docs: live agent evaluation v2 design spec by @forrestbthomas in #41
• feat: deterministic scorecard artifact + drift guards (eval hardening Go lane) by @forrestbthomas in #42
• test: Python contract tests for the real pi-run binary (eval hardening Python lane) by @forrestbthomas in #43
• ci: python-contract job + nightly build step + CLI-owned scorecard-latest by @forrestbthomas in #44
• feat: cost-per-task scorecard schema + --cost-mode (live-eval v2 Go lane) by @forrestbthomas in #45
• feat: live eval dataset v2 — 20 tasks, graders, references, schema lint by @forrestbthomas in #46
• feat: score_run.py baseline gate + task manifest + benchmark taxonomy (live-eval v2 scorer lane) by @forrestbthomas in #47
• ci: nightly live eval two-job split + live suite/metrics (live-eval v2 nightly lane) by @forrestbthomas in #48
• docs: v0.9.0 release notes + README pass by @forrestbthomas in #49

Full Changelog: v0.8.0...v0.9.0

v0.8.0

What's Changed

• fix(setup): resolve eval paths absolute from repo root by @forrestbthomas in #31
• docs: P2 research synthesis + context-engine decision by @forrestbthomas in #32
• feat: pi-run project-understand (Kiro-style project docs) by @forrestbthomas in #33
• feat: pi-run mcp-server (read-only MCP over stdio) by @forrestbthomas in #34
• feat: env-gated OTel GenAI agent telemetry export by @forrestbthomas in #35
• docs: cost-aware model routing (--model-tier) design spec by @forrestbthomas in #36
• docs: Agent Plugins 1.0 positioning + example manifest (P3) by @forrestbthomas in #37
• feat: cost-aware model routing (--model-tier fast|balanced|cheap) by @forrestbthomas in #38
• docs: v0.8.0 release notes + README pass by @forrestbthomas in #39

Full Changelog: v0.7.0...v0.8.0

v0.7.0

What's Changed

• feat: P1 permissions, hooks, and provider breadth by @forrestbthomas in #29
• docs(changelog): cut v0.7.0 release notes by @forrestbthomas in #30

Full Changelog: v0.6.0...v0.7.0

v0.6.0

What's Changed

• docs: competitive gap analysis + v0.5.0 feature specs by @forrestbthomas in #24
• feat: benchmark runner + cost tracking (v0.5.0) by @forrestbthomas in #25
• docs: revise gap analysis + CI provider scorecard spec by @forrestbthomas in #26
• feat: provider scorecard in CI (pi-run ci-benchmark, v0.6.0) by @forrestbthomas in #27
• docs(changelog): cut v0.6.0 release notes by @forrestbthomas in #28

Full Changelog: v0.4.3...v0.6.0