loki-mode 11.119.0 → 11.120.1
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.
- package/SKILL.md +4 -13
- package/VERSION +1 -1
- package/autonomy/completion-council.sh +3 -4
- package/autonomy/lib/backlog.py +107 -37
- package/autonomy/lib/contract-scaffold/README.md +2 -2
- package/autonomy/lib/proof-generator.py +1 -2
- package/autonomy/lib/scaffold-hook.sh +3 -7
- package/autonomy/loki +5 -3
- package/autonomy/notification-checker.py +2 -2
- package/autonomy/run.sh +28 -57
- package/dashboard/__init__.py +1 -1
- package/dashboard/api_v2.py +1 -1
- package/docs/ARCHITECTURE.md +4 -4
- package/docs/CLI-REFERENCE.md +1 -1
- package/docs/COMPONENTS.md +1 -3
- package/docs/INSTALLATION.md +1 -1
- package/docs/dev/architecture-reference.md +8 -12
- package/docs/trackers.md +12 -0
- package/docs/v10/GUIDE.md +3 -3
- package/loki-ts/data/child-env.json +2 -2
- package/loki-ts/data/finding-schema.json +2 -2
- package/loki-ts/dist/loki.js +1526 -1516
- package/mcp/__init__.py +1 -1
- package/mcp/lsp_proxy.py +3 -3
- package/package.json +2 -2
- package/packages/control-plane/dist/server.js +4947 -4750
- package/plugins/loki-mode/.claude-plugin/plugin.json +1 -1
- package/skills/model-selection.md +1 -1
- package/skills/quality-gates.md +8 -38
- package/skills/testing.md +1 -1
- package/skills/troubleshooting.md +0 -1
- package/tools/calibration-audit.py +1 -1
- package/web-app/dist/assets/{AdminPage-BI9sK7Ww.js → AdminPage-CSwrd2UA.js} +1 -1
- package/web-app/dist/assets/{Avatar-CqStW3hh.js → Avatar-B6SmKtXt.js} +1 -1
- package/web-app/dist/assets/{Badge-Sqo523qW.js → Badge-CR4U3l95.js} +1 -1
- package/web-app/dist/assets/{Button-65BGFYzo.js → Button-DmcHmlo6.js} +1 -1
- package/web-app/dist/assets/{CockpitPage-DLppS7i4.js → CockpitPage-CR_GmqRg.js} +1 -1
- package/web-app/dist/assets/{ComparePage-y-To7RDW.js → ComparePage-DH9gwrob.js} +1 -1
- package/web-app/dist/assets/{ErrorBoundary-BRfjLxKD.js → ErrorBoundary-23WE8Kb8.js} +1 -1
- package/web-app/dist/assets/{EvidenceReceiptPanel-BZYXpZlj.js → EvidenceReceiptPanel-dFODI74J.js} +1 -1
- package/web-app/dist/assets/{GitHubIssuesPanel-D5VQeqef.js → GitHubIssuesPanel-CudHcPXf.js} +1 -1
- package/web-app/dist/assets/{GitHubPRsPanel-B-ZTugBP.js → GitHubPRsPanel-0WcASaRW.js} +1 -1
- package/web-app/dist/assets/{HomePage-3O1Ixdaj.js → HomePage-CQL8xADH.js} +2 -2
- package/web-app/dist/assets/{LoginPage-D_c2Q5Yd.js → LoginPage-BRfdWXgI.js} +1 -1
- package/web-app/dist/assets/{MagicPage-CWoBDJIL.js → MagicPage-CaaPZyuJ.js} +1 -1
- package/web-app/dist/assets/{MetricsPage-D_8_KsDG.js → MetricsPage-BjQ0sz0F.js} +1 -1
- package/web-app/dist/assets/{NotFoundPage-Ck7EB21i.js → NotFoundPage-BNnrwX96.js} +1 -1
- package/web-app/dist/assets/{ProjectPage-DjzxAUpk.js → ProjectPage-pDqfWbBG.js} +3 -3
- package/web-app/dist/assets/{ProjectsPage-IiL9eZZl.js → ProjectsPage-CiHCZYVm.js} +1 -1
- package/web-app/dist/assets/{SettingsPage-B21XsJXj.js → SettingsPage-B-1-R56L.js} +1 -1
- package/web-app/dist/assets/{ShowcasePage-Dw-qMqRX.js → ShowcasePage-BCa643R5.js} +1 -1
- package/web-app/dist/assets/{SystemSettingsPage-DRKM2q7Y.js → SystemSettingsPage-BUHVmvMB.js} +1 -1
- package/web-app/dist/assets/{TeamsPage-cE3eF_mQ.js → TeamsPage-CG-DqMjn.js} +1 -1
- package/web-app/dist/assets/{TemplatesPage-2m8EFD1w.js → TemplatesPage-CO_kXHRt.js} +1 -1
- package/web-app/dist/assets/{TerminalOutput-BVj57oRn.js → TerminalOutput-C_iYHB-7.js} +1 -1
- package/web-app/dist/assets/{activity--3RdDrr_.js → activity-Cp6ZIiZc.js} +1 -1
- package/web-app/dist/assets/{bell-Cf1aNwyn.js → bell-BZf6hsa9.js} +1 -1
- package/web-app/dist/assets/{bot-C9eepiBo.js → bot-CGmNbYLx.js} +1 -1
- package/web-app/dist/assets/{check-C0eAyl30.js → check-CMDDpUDY.js} +1 -1
- package/web-app/dist/assets/{chevron-left-CQzEDgKz.js → chevron-left-DTtdfTHr.js} +1 -1
- package/web-app/dist/assets/{circle-alert-C45LOFIn.js → circle-alert-Ch6h3j52.js} +1 -1
- package/web-app/dist/assets/{clock-DRQ4_6O5.js → clock-Ce1nf2ai.js} +1 -1
- package/web-app/dist/assets/{cloud-D2YwYtSX.js → cloud-p2TFekpg.js} +1 -1
- package/web-app/dist/assets/{code-xml-C0kAguzE.js → code-xml-Kyu4vmM1.js} +1 -1
- package/web-app/dist/assets/{database-Dc-69TwF.js → database-B4HrqS9b.js} +1 -1
- package/web-app/dist/assets/{dollar-sign-D80FbCfO.js → dollar-sign-Cw-uyKAg.js} +1 -1
- package/web-app/dist/assets/{file-code-corner-D0vcShuD.js → file-code-corner-C8N3vuAd.js} +1 -1
- package/web-app/dist/assets/{file-plus-D7P7v4hD.js → file-plus-Ddd367Rd.js} +1 -1
- package/web-app/dist/assets/{globe-DGnqby6a.js → globe-BW_Nzcek.js} +1 -1
- package/web-app/dist/assets/{hammer-BvafjTXC.js → hammer-rS0zwOgP.js} +1 -1
- package/web-app/dist/assets/{index-1Kvtppgn.js → index-BFtUVrg2.js} +3 -3
- package/web-app/dist/assets/{layers-7f3JxJPK.js → layers-ChTtsWHz.js} +1 -1
- package/web-app/dist/assets/{loader-circle-Dsmov_Fj.js → loader-circle-Dl2j378Y.js} +1 -1
- package/web-app/dist/assets/{lock-CYPcR4L9.js → lock-H0jwxcaH.js} +1 -1
- package/web-app/dist/assets/{package-CDRe5Fid.js → package-Cn0OyUWi.js} +1 -1
- package/web-app/dist/assets/{plus-BwmTPRgM.js → plus-c4Cn8Hg_.js} +1 -1
- package/web-app/dist/assets/{refresh-cw-DxlbreLM.js → refresh-cw-CvtIp31h.js} +1 -1
- package/web-app/dist/assets/{rotate-ccw-CpTwximO.js → rotate-ccw-CACNVELf.js} +1 -1
- package/web-app/dist/assets/{scroll-text-CgcWfO8K.js → scroll-text-wcpAN4zz.js} +1 -1
- package/web-app/dist/assets/{server-CBVNgMbf.js → server-C8tiYYoH.js} +1 -1
- package/web-app/dist/assets/{shield-alert-CL8lOpTJ.js → shield-alert-D2EqAn5F.js} +1 -1
- package/web-app/dist/assets/{thumbs-up-B245iaZM.js → thumbs-up-DoxGQgUE.js} +1 -1
- package/web-app/dist/assets/{trash-2-D_LCAVug.js → trash-2-HbLfFoN4.js} +1 -1
- package/web-app/dist/assets/{trending-down-DQn3R8k7.js → trending-down-Djd4tkxi.js} +1 -1
- package/web-app/dist/assets/{trending-up-BuXPO-tN.js → trending-up-B9Nco_QQ.js} +1 -1
- package/web-app/dist/assets/{upload-DNc_smXL.js → upload-Bu_bR4UG.js} +1 -1
- package/web-app/dist/assets/{usePolling-CwMs9ya4.js → usePolling-D6jTTEO8.js} +1 -1
- package/web-app/dist/assets/{user-DupXJP9t.js → user-3vyW2Iwk.js} +1 -1
- package/web-app/dist/index.html +1 -1
package/SKILL.md
CHANGED
|
@@ -3,7 +3,7 @@ name: loki-mode
|
|
|
3
3
|
description: Runs an autonomous software factory that knows what it is supposed to deliver and proves it did, with a signed receipt. Use when the user says Loki Mode or asks to build, fix or verify software autonomously.
|
|
4
4
|
---
|
|
5
5
|
|
|
6
|
-
# Loki Mode v11.
|
|
6
|
+
# Loki Mode v11.120.1
|
|
7
7
|
|
|
8
8
|
**Current entry points (Loki 10 engine):** `loki "<task>"`, `loki owner/repo#N` and `loki quick "<task>"` run the Loki 10 engine and end with a signed receipt (`loki verify`). `loki backlog owner/repo --all` runs many issues in parallel. Guide: `docs/v10/GUIDE.md`. The skill instructions below, and `loki start`, describe the previous bash loop. PRD-file starts (`loki start ./prd.md`) and the opencode provider still run on it until FC38-SWEEP ports them; the legacy-engine command and the LOKI_ENGINE switch remain removed. Prefer `loki owner/repo#N` for issues.
|
|
9
9
|
|
|
@@ -395,15 +395,12 @@ Two completion-trust features extend the verification gates. Full details in `sk
|
|
|
395
395
|
|
|
396
396
|
## Harness intelligence (v8.0.0)
|
|
397
397
|
|
|
398
|
-
|
|
398
|
+
Three measured-harness disciplines layered onto the existing trust core. None of
|
|
399
399
|
them can weaken a gate: each either adds verification or saves budget on work
|
|
400
400
|
that cannot succeed.
|
|
401
401
|
|
|
402
402
|
| Env Var | Default | Effect |
|
|
403
403
|
|---------|---------|--------|
|
|
404
|
-
| `LOKI_CONFIDENCE_SPIKE=0` | on | Disable the confidence-spike re-check |
|
|
405
|
-
| `LOKI_CONFIDENCE_SPIKE_DELTA` | `40` | Confidence jump (points) that counts as a spike |
|
|
406
|
-
| `LOKI_CONFIDENCE_SPIKE_MIN` | `90` | Absolute level that counts as a spike on first arrival |
|
|
407
404
|
| `LOKI_GOAL_SCORING=0` | on | Disable the goal-measurability advisory |
|
|
408
405
|
| `LOKI_SMART_RETRY=0` | on | Retry every failure, including non-retryable ones |
|
|
409
406
|
| `LOKI_SIMPLE=1` | off | Strip the coaching half of the system prompt (-78%, ~1562 tokens/iteration). Experimental ablation arm. |
|
|
@@ -413,18 +410,12 @@ that cannot succeed.
|
|
|
413
410
|
`[CACHE_BREAKPOINT]`; the SDK judge path applies `cache_control` on that split.
|
|
414
411
|
Any new always-on instruction belongs in the prefix, or it busts the cache
|
|
415
412
|
every iteration.
|
|
416
|
-
- **Confidence-spike re-check.** A jump to near-maximal self-reported confidence
|
|
417
|
-
forces ONE extra verification before the done-signal valve force-stops a run.
|
|
418
|
-
Strictly additive: a spike can only ADD a verification pass, never skip,
|
|
419
|
-
shorten, or satisfy a gate. It cannot delay the stagnation valve, and the
|
|
420
|
-
delay is one-shot, so a repeatedly-spiking run cannot postpone the valve
|
|
421
|
-
indefinitely.
|
|
422
413
|
- **Hill-climbable goal scoring.** A `COMPLETION_PROMISE` with no measurable
|
|
423
414
|
target (no number, comparator, named metric, or verifiable artifact) gets a
|
|
424
415
|
prompt advisory asking for a checkable success condition. Advisory only: it
|
|
425
416
|
never blocks a build and never rewrites the goal. Suppressed for an absent
|
|
426
417
|
goal and in perpetual mode, where open-endedness is the chosen configuration.
|
|
427
|
-
|
|
418
|
+
Implemented in the bash loop (`autonomy/run.sh`).
|
|
428
419
|
- **Smart retry.** A positively-identified permanent failure (bad credentials,
|
|
429
420
|
unknown model, exhausted quota) stops early instead of burning the retry
|
|
430
421
|
budget on guaranteed-identical failures. Fail-safe: an unrecognized error
|
|
@@ -473,4 +464,4 @@ Implemented, planned and deprecated items, with the version each landed in: `CHA
|
|
|
473
464
|
|
|
474
465
|
---
|
|
475
466
|
|
|
476
|
-
**v11.
|
|
467
|
+
**v11.120.1 | [Autonomi](https://www.autonomi.dev/) flagship product | core skill**
|
package/VERSION
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
11.
|
|
1
|
+
11.120.1
|
|
@@ -2031,7 +2031,7 @@ runner = d.get('runner', 'none')
|
|
|
2031
2031
|
passed = d.get('pass')
|
|
2032
2032
|
status = d.get('status', '')
|
|
2033
2033
|
# BACKLOG 89: a recorded failure count above zero is a failure whatever "pass"
|
|
2034
|
-
# says
|
|
2034
|
+
# says: a numeric
|
|
2035
2035
|
# failed_count wins, the legacy numeric failed is the fallback, and anything
|
|
2036
2036
|
# else (null, missing, a bool) is unmeasured, never a failure and never 0.
|
|
2037
2037
|
def _count(v):
|
|
@@ -3295,8 +3295,7 @@ ISSUES: CRITICAL:description (optional, one per line per issue)"
|
|
|
3295
3295
|
# output; the downstream parse already greps VOTE/REASON/ISSUES.
|
|
3296
3296
|
# CAVEMAN_DEFAULT_MODE=off suppression is preserved (see above).
|
|
3297
3297
|
# bash-F3: timeout-guard the provider subcall so a hung CLI can
|
|
3298
|
-
# not stall the whole council. Default 600s
|
|
3299
|
-
# (council.ts LOKI_COUNCIL_TIMEOUT_MS=600000).
|
|
3298
|
+
# not stall the whole council. Default 600s.
|
|
3300
3299
|
# bash-F4 (WAVE10 SAFE-DEFAULT): capture the subcall exit code so a
|
|
3301
3300
|
# timeout (124) is NOT silently routed into council_heuristic_review.
|
|
3302
3301
|
# The heuristic fallback defaults to APPROVE on benign evidence, so
|
|
@@ -3502,7 +3501,7 @@ REASON: one-line justification citing path:line evidence"
|
|
|
3502
3501
|
# v7.41.3 BUG A: full capture, no tail-truncation (see member
|
|
3503
3502
|
# subcall note). CAVEMAN_DEFAULT_MODE=off suppression preserved.
|
|
3504
3503
|
# bash-F3: timeout-guard the devil's-advocate subcall (same 600s
|
|
3505
|
-
# default as the member subcalls
|
|
3504
|
+
# default as the member subcalls). An empty
|
|
3506
3505
|
# verdict on timeout hits the conservative REJECT fallback below.
|
|
3507
3506
|
verdict=$(echo "$prompt" | timeout "${LOKI_COUNCIL_REVIEW_TIMEOUT:-600}" env "SHELL=$_co_shell" CAVEMAN_DEFAULT_MODE=off claude "${_co_argv[@]}" -p 2>/dev/null)
|
|
3508
3507
|
fi
|
package/autonomy/lib/backlog.py
CHANGED
|
@@ -3,11 +3,14 @@
|
|
|
3
3
|
|
|
4
4
|
loki backlog owner/repo --all | --label X | --issues 1,2,3
|
|
5
5
|
[--concurrency N] [--slots N] [--dry-run]
|
|
6
|
+
loki backlog jira:PROJ | linear:TEAM (same options; Jira and Linear items)
|
|
6
7
|
|
|
7
8
|
Each issue gets its own git worktree and branch (loki/backlog-N) beside the
|
|
8
9
|
checkout; the launcher runs inside it. Success = the child exited 0 (the engine's
|
|
9
|
-
ladder: 0 VERIFIED, 1 FAILED, 3 BUDGET_STOP, 4 BLOCKED, 5 STALLED). A
|
|
10
|
-
from a failed run is still FAILED
|
|
10
|
+
ladder: 0 VERIFIED, 1 FAILED, 3 BUDGET_STOP, 4 BLOCKED, 5 STALLED, 6 PARTIAL). A
|
|
11
|
+
draft PR from a failed run is still FAILED; a PARTIAL run counts only with its PR.
|
|
12
|
+
A split issue prints one Outcome and one Cost line for all its slices, so its
|
|
13
|
+
spend reaches the day ledger like any run. Exit: 0 all ok, 3 only budget stops, 1 other
|
|
11
14
|
failures, 2 usage/config error.
|
|
12
15
|
|
|
13
16
|
Unit mode (D61 slice 10): --dag FILE [--group G] also runs the units of a
|
|
@@ -21,14 +24,22 @@ Passing unit branches are kept for the integrator; issue mode is unchanged.
|
|
|
21
24
|
|
|
22
25
|
Triage: before anything runs, every issue is classified by the triage router
|
|
23
26
|
(loki-ts/src/features/triage.ts, via `loki internal triage`) as implement (run
|
|
24
|
-
it
|
|
25
|
-
|
|
26
|
-
with
|
|
27
|
+
it; a large issue, by the one rule in features/decompose.ts deliverables, runs
|
|
28
|
+
and the engine splits it into slices), spec (vague or ambiguous: draft
|
|
29
|
+
.loki/specs/<slug>.md with `loki plan --spec`, run nothing, print the command to
|
|
30
|
+
run it after review) or human (never run headless; listed with the reason). --dry-run prints the class and reason per issue. A label
|
|
27
31
|
loki:implement, loki:spec or loki:human forces the class; LOKI_TRIAGE=0 turns
|
|
28
32
|
the router off (every issue runs, as before). The class and reason reach each
|
|
29
33
|
child run as LOKI_TRIAGE_CLASS / LOKI_TRIAGE_REASON, which the engine records in
|
|
30
34
|
the run events, the receipt and the PR body.
|
|
31
35
|
|
|
36
|
+
Jira and Linear: `jira:PROJ` or `linear:TEAM` in place of owner/repo lists that
|
|
37
|
+
project's open items with `loki internal tracker-list` (features/tracker_intake.ts,
|
|
38
|
+
the same credentials as `loki jira:PROJ-1`), triages them the same way and runs
|
|
39
|
+
each as `loki jira:PROJ-N` / `loki linear:TEAM-N` from the current checkout.
|
|
40
|
+
--issues takes item numbers (12 means PROJ-12). A large Jira or Linear item runs
|
|
41
|
+
as one run: the engine splits only GitHub issues today.
|
|
42
|
+
|
|
32
43
|
Test seam: LOKI_BACKLOG_LAUNCHER replaces the real launcher (called with the
|
|
33
44
|
issue ref, cwd = the issue worktree).
|
|
34
45
|
"""
|
|
@@ -51,6 +62,10 @@ import worktree_prep # noqa: E402
|
|
|
51
62
|
|
|
52
63
|
REPO_ROOT = os.path.abspath(os.path.join(HERE, "..", ".."))
|
|
53
64
|
REPO_RE = re.compile(r"^[A-Za-z0-9_.-]+/[A-Za-z0-9_.-]+$")
|
|
65
|
+
TRACKER_RE = re.compile(r"^(jira|linear):([A-Za-z][A-Za-z0-9_]*)$")
|
|
66
|
+
# A tracker item key names worktree, branch and log paths: only PROJECT-<number> of the listed project is accepted.
|
|
67
|
+
ITEM_KEY_RE = r"^%s-\d+$" # %s: the validated, uppercased project key
|
|
68
|
+
TRACKER_SECRETS = {"jira": ("JIRA_API_TOKEN", "JIRA_TOKEN", "LOKI_JIRA_TOKEN"), "linear": ("LINEAR_API_KEY", "LOKI_LINEAR_TOKEN")} # each tracker's token and its aliases
|
|
54
69
|
|
|
55
70
|
|
|
56
71
|
def say(msg):
|
|
@@ -76,9 +91,41 @@ def origin_slug(cwd):
|
|
|
76
91
|
return m.group(1) if m else ""
|
|
77
92
|
|
|
78
93
|
|
|
79
|
-
def _issue(i):
|
|
94
|
+
def _issue(i, repo):
|
|
95
|
+
"""One work item. key: the job-table key; label: how it prints; ref: what the launcher runs; slug: branch suffix."""
|
|
80
96
|
labels = [l.get("name", "") if isinstance(l, dict) else str(l) for l in i.get("labels") or []]
|
|
81
|
-
|
|
97
|
+
n = i["number"]
|
|
98
|
+
return {"number": n, "key": n, "label": "#%d" % n, "ref": "%s#%d" % (repo, n), "slug": str(n), "split": True,
|
|
99
|
+
"title": i.get("title") or "", "body": i.get("body") or "", "labels": labels}
|
|
100
|
+
|
|
101
|
+
|
|
102
|
+
def list_tracker(spec, label, numbers, env):
|
|
103
|
+
"""Jira/Linear items via `loki internal tracker-list` (features/tracker_intake.ts). Raises RuntimeError."""
|
|
104
|
+
source, project = spec.split(":", 1)
|
|
105
|
+
cmd = [os.path.join(REPO_ROOT, "bin", "loki"), "internal", "tracker-list", spec]
|
|
106
|
+
if label:
|
|
107
|
+
cmd += ["--label", label]
|
|
108
|
+
if numbers:
|
|
109
|
+
cmd += ["--issues", ",".join(str(n) for n in numbers)]
|
|
110
|
+
r = subprocess.run(cmd, capture_output=True, text=True, env=env, stdin=subprocess.DEVNULL)
|
|
111
|
+
try:
|
|
112
|
+
rows = json.loads(r.stdout) if r.returncode == 0 else None
|
|
113
|
+
if not isinstance(rows, list):
|
|
114
|
+
raise ValueError("not a list")
|
|
115
|
+
out = [{"number": int(x["number"]), "key": str(x["key"]), "label": str(x["key"]), "ref": "%s:%s" % (source, x["key"]),
|
|
116
|
+
"slug": str(x["key"]), "split": False, "title": str(x.get("title") or ""), "body": str(x.get("body") or ""),
|
|
117
|
+
"labels": [str(l) for l in x.get("labels") or []], "source": source} for x in rows]
|
|
118
|
+
except (ValueError, TypeError, KeyError):
|
|
119
|
+
why = (r.stderr.strip().splitlines() or ["no usable output"])[-1]
|
|
120
|
+
raise RuntimeError("listing %s failed (%s)" % (spec, why))
|
|
121
|
+
key_re = re.compile(ITEM_KEY_RE % re.escape(project))
|
|
122
|
+
for i in out:
|
|
123
|
+
if not key_re.fullmatch(i["key"]) or not str(i["number"]) == i["key"].rsplit("-", 1)[1]:
|
|
124
|
+
raise RuntimeError("listing %s returned an invalid item key %r; refusing to use it in a path" % (spec, i["key"][:40]))
|
|
125
|
+
missing = sorted(set(numbers or []) - {i["number"] for i in out})
|
|
126
|
+
if missing:
|
|
127
|
+
raise RuntimeError("%s has no item %s" % (spec, ", ".join("%s-%d" % (project.upper(), n) for n in missing)))
|
|
128
|
+
return out
|
|
82
129
|
|
|
83
130
|
|
|
84
131
|
def list_issues(repo, label, env):
|
|
@@ -89,7 +136,7 @@ def list_issues(repo, label, env):
|
|
|
89
136
|
r = subprocess.run(cmd, capture_output=True, text=True, env=env)
|
|
90
137
|
if r.returncode != 0:
|
|
91
138
|
raise RuntimeError("gh issue list failed: " + (r.stderr.strip().splitlines() or ["unknown error"])[0])
|
|
92
|
-
return sorted((_issue(i) for i in json.loads(r.stdout or "[]")), key=lambda i: i["number"])
|
|
139
|
+
return sorted((_issue(i, repo) for i in json.loads(r.stdout or "[]")), key=lambda i: i["number"])
|
|
93
140
|
|
|
94
141
|
|
|
95
142
|
def view_issues(repo, numbers, env):
|
|
@@ -99,7 +146,7 @@ def view_issues(repo, numbers, env):
|
|
|
99
146
|
capture_output=True, text=True, env=env)
|
|
100
147
|
if r.returncode != 0:
|
|
101
148
|
raise RuntimeError("gh issue view %d failed: %s" % (n, (r.stderr.strip().splitlines() or ["unknown error"])[0]))
|
|
102
|
-
out.append(_issue(json.loads(r.stdout or "{}")))
|
|
149
|
+
out.append(_issue(json.loads(r.stdout or "{}"), repo))
|
|
103
150
|
return out
|
|
104
151
|
|
|
105
152
|
|
|
@@ -107,12 +154,12 @@ TRIAGE_CLASSES = ("implement", "spec", "human")
|
|
|
107
154
|
|
|
108
155
|
|
|
109
156
|
def triage(issues, top, env):
|
|
110
|
-
"""-> {
|
|
157
|
+
"""-> {key: (class, reason)} from `loki internal triage`; raises RuntimeError (fail closed: nothing untriaged runs)."""
|
|
111
158
|
if os.environ.get("LOKI_TRIAGE") == "0":
|
|
112
|
-
return {i["
|
|
159
|
+
return {i["key"]: ("implement", "LOKI_TRIAGE=0: router off") for i in issues}
|
|
113
160
|
if not issues:
|
|
114
161
|
return {}
|
|
115
|
-
items = [{"id": i["
|
|
162
|
+
items = [{"id": i["key"], "title": i["title"], "body": i["body"], "labels": i["labels"], "split": i["split"]} for i in issues]
|
|
116
163
|
r = subprocess.run([os.path.join(REPO_ROOT, "bin", "loki"), "internal", "triage", "--repo-dir", top],
|
|
117
164
|
input=json.dumps(items), capture_output=True, text=True, env=env)
|
|
118
165
|
try:
|
|
@@ -120,12 +167,17 @@ def triage(issues, top, env):
|
|
|
120
167
|
got = {row["id"]: (row["class"], row["reason"]) for row in rows if row.get("class") in TRIAGE_CLASSES}
|
|
121
168
|
except (ValueError, TypeError, KeyError, AttributeError):
|
|
122
169
|
got = None
|
|
123
|
-
if not got or set(got) != {i["
|
|
170
|
+
if not got or set(got) != {i["key"] for i in issues}:
|
|
124
171
|
why = (r.stderr.strip().splitlines() or ["no usable output"])[-1]
|
|
125
172
|
raise RuntimeError("triage failed (%s); fix it, or set LOKI_TRIAGE=0 to run every issue untriaged" % why)
|
|
126
173
|
return got
|
|
127
174
|
|
|
128
175
|
|
|
176
|
+
def without(env, keys):
|
|
177
|
+
"""A copy of env with keys removed."""
|
|
178
|
+
return {k: v for k, v in env.items() if k not in keys}
|
|
179
|
+
|
|
180
|
+
|
|
129
181
|
def draft_spec(issue, top, env):
|
|
130
182
|
"""Draft .loki/specs/<slug>.md with `loki plan --spec` (one plan session). -> (ok, status line)."""
|
|
131
183
|
task = issue["title"] + ("\n\n" + issue["body"] if issue["body"].strip() else "")
|
|
@@ -180,10 +232,13 @@ def add_spend(usd):
|
|
|
180
232
|
|
|
181
233
|
def parse_result(rc, text):
|
|
182
234
|
"""-> (ok, status line text, cost_usd or None)."""
|
|
183
|
-
pr = re.search(r"^PR:\s+(https
|
|
235
|
+
pr = re.search(r"^PR:\s+((?:https?|local)://\S+)", text, re.M) # local://: published to a local bare origin (engine10-push.sh)
|
|
184
236
|
why = re.search(r"^Reason:\s+(.+)$", text, re.M)
|
|
185
237
|
cost = re.search(r"^Cost:\s+(?:partial:\s+)?\$([0-9]+(?:\.[0-9]+)?)", text, re.M)
|
|
186
238
|
usd = float(cost.group(1)) if cost else None
|
|
239
|
+
plan = re.search(r"^Split planning:\s+(?:partial:\s+)?\$([0-9]+(?:\.[0-9]+)?)", text, re.M) # a refused split's planning sessions
|
|
240
|
+
if plan:
|
|
241
|
+
usd = (usd or 0.0) + float(plan.group(1))
|
|
187
242
|
reason = why.group(1).strip() if why else ""
|
|
188
243
|
if rc == 0:
|
|
189
244
|
return True, ("PR " + pr.group(1)) if pr else "VERIFIED", usd
|
|
@@ -191,6 +246,8 @@ def parse_result(rc, text):
|
|
|
191
246
|
return False, "BLOCKED: " + (reason or "see the run log"), usd
|
|
192
247
|
if rc == 3:
|
|
193
248
|
return False, "BUDGET_STOP: " + (reason or "run cost cap reached"), usd
|
|
249
|
+
if rc == 6: # PARTIAL: some proof missing; counts as done only with its (draft) PR
|
|
250
|
+
return bool(pr), ("PARTIAL, PR " + pr.group(1)) if pr else "PARTIAL: " + (reason or "not every check proven"), usd
|
|
194
251
|
return False, "FAILED: " + (reason or "exit %d" % rc), usd
|
|
195
252
|
|
|
196
253
|
|
|
@@ -283,8 +340,14 @@ def main(argv):
|
|
|
283
340
|
ap.add_argument("--group", help="unit group id (default: the DAG's group, else g1)")
|
|
284
341
|
a = ap.parse_args(argv)
|
|
285
342
|
|
|
286
|
-
|
|
287
|
-
|
|
343
|
+
tracker = TRACKER_RE.match(a.repo)
|
|
344
|
+
if tracker: # jira:proj is jira:PROJ: tracker project keys are uppercase
|
|
345
|
+
a.repo = "%s:%s" % (tracker.group(1), tracker.group(2).upper())
|
|
346
|
+
tracker = TRACKER_RE.match(a.repo)
|
|
347
|
+
if tracker and os.environ.get("LOKI_TRACKER_INTAKE") == "0":
|
|
348
|
+
return die("LOKI_TRACKER_INTAKE=0 disables Jira and Linear; unset it to run %s" % a.repo)
|
|
349
|
+
if not (REPO_RE.match(a.repo) or tracker):
|
|
350
|
+
return die("repo must be owner/repo, jira:PROJ or linear:TEAM, got %r" % a.repo)
|
|
288
351
|
if not (a.all or a.label or a.issues or a.dag):
|
|
289
352
|
return die("choose one of --all, --label X, --issues 1,2,3, --dag FILE")
|
|
290
353
|
group, dag_units = None, []
|
|
@@ -329,7 +392,7 @@ def main(argv):
|
|
|
329
392
|
top = git("rev-parse", "--show-toplevel").stdout.strip()
|
|
330
393
|
if not top:
|
|
331
394
|
return die("run this from a checkout of %s" % a.repo)
|
|
332
|
-
if origin_slug(top).lower() != a.repo.lower():
|
|
395
|
+
if not tracker and origin_slug(top).lower() != a.repo.lower():
|
|
333
396
|
return die("this checkout's origin is %r, not %s; cd into a checkout of %s"
|
|
334
397
|
% (origin_slug(top) or "unknown", a.repo, a.repo))
|
|
335
398
|
|
|
@@ -345,8 +408,11 @@ def main(argv):
|
|
|
345
408
|
u["base"] = sha
|
|
346
409
|
|
|
347
410
|
try:
|
|
348
|
-
|
|
349
|
-
|
|
411
|
+
if tracker:
|
|
412
|
+
issues = list_tracker(a.repo, a.label, numbers, env) if (numbers or a.all or a.label) else []
|
|
413
|
+
else:
|
|
414
|
+
issues = (view_issues(a.repo, numbers, env) if numbers
|
|
415
|
+
else list_issues(a.repo, a.label, env) if (a.all or a.label) else [])
|
|
350
416
|
klass = triage(issues, top, env)
|
|
351
417
|
except RuntimeError as e:
|
|
352
418
|
return die(redact(str(e), token), 1)
|
|
@@ -354,13 +420,14 @@ def main(argv):
|
|
|
354
420
|
say("backlog: %s concurrency %d, %d issue(s)" % (a.repo, conc, len(issues))
|
|
355
421
|
+ (", %d unit(s) in group %s" % (len(dag_units), group) if dag_units else ""))
|
|
356
422
|
for i in issues:
|
|
357
|
-
c, why = klass[i["
|
|
358
|
-
say("backlog:
|
|
423
|
+
c, why = klass[i["key"]]
|
|
424
|
+
say("backlog: %s %-9s %s | %s" % (i["label"], c, i["title"], why))
|
|
359
425
|
if a.dry_run:
|
|
360
|
-
n_of = lambda c: sum(1 for i in issues if klass[i["
|
|
426
|
+
n_of = lambda c: sum(1 for i in issues if klass[i["key"]][0] == c) # noqa: E731
|
|
427
|
+
n_large = sum(1 for i in issues if i["split"] and klass[i["key"]][0] == "implement" and klass[i["key"]][1].startswith("large ("))
|
|
361
428
|
if issues:
|
|
362
|
-
say("backlog: dry run: %d would run, %d would get a drafted spec, %d need a human; nothing launched"
|
|
363
|
-
% (n_of("implement"), n_of("spec"), n_of("human")))
|
|
429
|
+
say("backlog: dry run: %d would run%s, %d would get a drafted spec, %d need a human; nothing launched"
|
|
430
|
+
% (n_of("implement"), " (%d split into slices)" % n_large if n_large else "", n_of("spec"), n_of("human")))
|
|
364
431
|
for u in dag_units:
|
|
365
432
|
say("backlog: would run unit %s-%d (%s) after: %s" % (group, u["n"], u["id"], ", ".join(u["deps"]) or "nothing"))
|
|
366
433
|
return 0
|
|
@@ -395,26 +462,29 @@ def main(argv):
|
|
|
395
462
|
day_cap = cfg.get("budgets", {}).get("per_day_usd")
|
|
396
463
|
|
|
397
464
|
# Triage: only implement issues run; spec issues get a drafted spec, human issues are listed.
|
|
398
|
-
routed = {} #
|
|
465
|
+
routed = {} # key -> (ok, status) for issues that do not run
|
|
466
|
+
labels = {i["key"]: i["label"] for i in issues}
|
|
399
467
|
for i in issues:
|
|
400
|
-
c, why = klass[i["
|
|
468
|
+
c, why = klass[i["key"]]
|
|
401
469
|
if c == "human":
|
|
402
|
-
routed[i["
|
|
470
|
+
routed[i["key"]] = (True, "HUMAN: " + why)
|
|
403
471
|
elif c == "spec":
|
|
404
|
-
ok, status = draft_spec(i, top, child_env)
|
|
405
|
-
routed[i["
|
|
472
|
+
ok, status = draft_spec(i, top, without(child_env, sum(TRACKER_SECRETS.values(), ())))
|
|
473
|
+
routed[i["key"]] = (ok, "%s (%s)" % (status, why))
|
|
406
474
|
if c != "implement":
|
|
407
|
-
say("backlog:
|
|
475
|
+
say("backlog: %s %s" % (i["label"], routed[i["key"]][1]))
|
|
408
476
|
|
|
409
477
|
# One job table for issues and units: they share one slot pool.
|
|
410
478
|
spec = {}
|
|
411
479
|
for i in issues:
|
|
412
|
-
n = i["
|
|
480
|
+
n = i["key"]
|
|
413
481
|
if n in routed:
|
|
414
482
|
continue
|
|
415
|
-
spec[n] = {"label": "
|
|
416
|
-
"cmd": [launcher, "
|
|
417
|
-
"env": {"LOKI_TRIAGE_CLASS": klass[n][0], "LOKI_TRIAGE_REASON": klass[n][1]}
|
|
483
|
+
spec[n] = {"label": i["label"], "branch": "loki/backlog-%s" % i["slug"], "wt": "issue-%s" % i["slug"],
|
|
484
|
+
"cmd": [launcher, i["ref"]], "deps": [], "unit": False,
|
|
485
|
+
"env": {"LOKI_TRIAGE_CLASS": klass[n][0], "LOKI_TRIAGE_REASON": klass[n][1]},
|
|
486
|
+
# a child keeps only its own tracker's token (intake and writeback need it); the other tracker's never reaches it
|
|
487
|
+
"drop": [k for src, keys in TRACKER_SECRETS.items() if src != i.get("source") for k in keys]}
|
|
418
488
|
key_of = {u["id"]: "unit-%s-%d" % (group, u["n"]) for u in dag_units}
|
|
419
489
|
for u in dag_units:
|
|
420
490
|
k = key_of[u["id"]]
|
|
@@ -423,7 +493,7 @@ def main(argv):
|
|
|
423
493
|
"deps": [key_of[d] for d in u["deps"]],
|
|
424
494
|
"env": {"LOKI_GROUP_ID": group, "LOKI_UNIT_ID": u["id"], "LOKI_UNIT_N": str(u["n"]),
|
|
425
495
|
"LOKI_UNIT_DEPS": ",".join(u["deps"]), "LOKI_UNIT_WRITE_SET": "\n".join(u["write_set"])},
|
|
426
|
-
"done": u["done"], "base": u["base"]}
|
|
496
|
+
"done": u["done"], "base": u["base"], "drop": sum(TRACKER_SECRETS.values(), ())} # a unit is a text task: no tracker token
|
|
427
497
|
if u["spec"]: # frozen group acceptance for unit_mode (wall.ts)
|
|
428
498
|
spec[k]["env"].update(LOKI_UNIT_SPEC=u["spec"], LOKI_SPEED="1")
|
|
429
499
|
state = {k: {"status": "queued", "ok": None, "cost": None, "ran": False} for k in spec}
|
|
@@ -531,7 +601,7 @@ def main(argv):
|
|
|
531
601
|
return
|
|
532
602
|
log = os.path.join(wt_root, "logs", s["wt"] + ".log")
|
|
533
603
|
with open(log, "w") as lf:
|
|
534
|
-
p = subprocess.Popen(s["cmd"], cwd=wt, env=dict(child_env, **s["env"]), stdout=lf,
|
|
604
|
+
p = subprocess.Popen(s["cmd"], cwd=wt, env=without(dict(child_env, **s["env"]), s.get("drop", ())), stdout=lf,
|
|
535
605
|
stderr=subprocess.STDOUT, stdin=subprocess.DEVNULL, start_new_session=True)
|
|
536
606
|
running[k] = (p, log, wt)
|
|
537
607
|
state[k].update(status="running", ran=True)
|
|
@@ -589,7 +659,7 @@ def main(argv):
|
|
|
589
659
|
say("Summary: %s" % a.repo)
|
|
590
660
|
say("%-14s %-10s %s" % ("ITEM", "COST", "RESULT"))
|
|
591
661
|
for n, s in state.items():
|
|
592
|
-
say("%-14s %-10s %s" % (spec[n]["label"] if n in spec else
|
|
662
|
+
say("%-14s %-10s %s" % (spec[n]["label"] if n in spec else labels[n], ("$%.2f" % s["cost"] if s["cost"] is not None else "unmeasured" if s["ran"] else "-"), s["status"]))
|
|
593
663
|
measured = sum(s["cost"] or 0 for s in state.values())
|
|
594
664
|
say("Total measured cost: $%.2f%s" % (measured, " (%d run(s) unmeasured, not counted as $0)" % unmeasured if unmeasured else ""))
|
|
595
665
|
bad = [s for s in state.values() if not s["ok"]]
|
|
@@ -33,7 +33,7 @@ knowledge of any specific PRD** (anti-teaching-to-the-test).
|
|
|
33
33
|
This is a capability module, like FV-1 was. It changes zero existing build
|
|
34
34
|
behavior. **Wiring it into the default build lane is the gated next step** (the
|
|
35
35
|
M1->"M1-wire" fork, analogous to FV-1->FV-2): that step touches the parity-locked
|
|
36
|
-
`run.sh
|
|
36
|
+
`run.sh` build_prompt core, reclassifies what every build produces, and
|
|
37
37
|
requires bash+bun parity work + council review + founder sign-off. Do NOT wire it
|
|
38
38
|
in as a solo/rushed action.
|
|
39
39
|
|
|
@@ -79,5 +79,5 @@ UI -- the spine working as designed).
|
|
|
79
79
|
## Next (wiring, see the plan)
|
|
80
80
|
|
|
81
81
|
- Wire M1+M2+M3 into the build lane (GATED: touches parity-locked run.sh/
|
|
82
|
-
|
|
82
|
+
run.sh build_prompt core -> council + founder sign-off).
|
|
83
83
|
- Wire the FV harness into the completion verdict (FV-2, founder-gated).
|
|
@@ -3,7 +3,6 @@
|
|
|
3
3
|
|
|
4
4
|
Single implementation called by both routes:
|
|
5
5
|
- bash: autonomy/run.sh generate_proof_of_run() via python3
|
|
6
|
-
- Bun: loki-ts/src/runner/proof.ts via spawn
|
|
7
6
|
|
|
8
7
|
Assembles the frozen proof.json schema v1.0 from .loki/ state, runs the
|
|
9
8
|
redaction chokepoint exactly once, computes an integrity hash, and writes
|
|
@@ -341,7 +340,7 @@ def _test_results_fresh(quality_dir):
|
|
|
341
340
|
test-results.json together with .test-results.iter; they are this
|
|
342
341
|
iteration's evidence only when that marker holds this iteration's number.
|
|
343
342
|
Both loop callers always pass the iteration as ITERATION_COUNT (run.sh
|
|
344
|
-
generate_proof_of_run
|
|
343
|
+
generate_proof_of_run); a set but empty value is not
|
|
345
344
|
fresh.
|
|
346
345
|
ponytail: ITERATION_COUNT absent altogether means a hand run outside any
|
|
347
346
|
loop, with no iteration to compare against, so the pre-82 reading applies
|
|
@@ -14,13 +14,9 @@
|
|
|
14
14
|
# lets the normal build proceed -- it can only ADD a starting skeleton, never
|
|
15
15
|
# remove or overwrite existing code.
|
|
16
16
|
#
|
|
17
|
-
#
|
|
18
|
-
#
|
|
19
|
-
#
|
|
20
|
-
# only (not behavior-mirrored in the bun runner). Because it is DEFAULT-OFF, the
|
|
21
|
-
# bun default path is unaffected. When the bun runner (loki-ts autonomous.ts,
|
|
22
|
-
# currently a skeleton port) reaches run_autonomous behavior parity, it should
|
|
23
|
-
# mirror this opt-in gate.
|
|
17
|
+
# This is a bash-only ORCHESTRATION seam in run_autonomous, not part of
|
|
18
|
+
# build_prompt(). It follows the LOKI_SENTRUX_GATE precedent, which is likewise
|
|
19
|
+
# bash-orchestration only. It is DEFAULT-OFF.
|
|
24
20
|
|
|
25
21
|
# Is the target dir greenfield (no meaningful source yet)? Ignores .loki, .git,
|
|
26
22
|
# the PRD, and dotfiles. Echoes "yes"/"no".
|
package/autonomy/loki
CHANGED
|
@@ -14873,6 +14873,7 @@ cmd_backlog() {
|
|
|
14873
14873
|
echo -e "${BOLD}loki backlog${NC} - run a v10 issue-mode run for every matching open issue"
|
|
14874
14874
|
echo ""
|
|
14875
14875
|
echo "Usage: loki backlog owner/repo (--all | --label X | --issues 1,2,3) [--concurrency N] [--dry-run]"
|
|
14876
|
+
echo " loki backlog jira:PROJ | linear:TEAM (same options; --issues 12 means PROJ-12)"
|
|
14876
14877
|
echo ""
|
|
14877
14878
|
echo " --all Every open issue"
|
|
14878
14879
|
echo " --label X Open issues with label X"
|
|
@@ -14880,8 +14881,10 @@ cmd_backlog() {
|
|
|
14880
14881
|
echo " --concurrency N Max runs in parallel (default: loki.yaml concurrency, else 2)"
|
|
14881
14882
|
echo " --dry-run Print each issue's triage class (implement, spec, human) and reason; launch nothing"
|
|
14882
14883
|
echo ""
|
|
14883
|
-
echo "Run from a checkout of owner/repo
|
|
14884
|
-
echo "
|
|
14884
|
+
echo "Run from a checkout of owner/repo (for Jira or Linear, the checkout the work lands in)."
|
|
14885
|
+
echo "Each issue runs in its own worktree on loki/backlog-N. Jira and Linear use the credentials in docs/trackers.md."
|
|
14886
|
+
echo "Triage: implement issues run, and a large one is split into slices that each run on their own;"
|
|
14887
|
+
echo "vague or ambiguous (spec) issues get .loki/specs/<slug>.md and the command to run it;"
|
|
14885
14888
|
echo "human issues are listed with the reason. Label loki:implement|loki:spec|loki:human forces a class;"
|
|
14886
14889
|
echo "LOKI_TRIAGE=0 runs every issue untriaged."
|
|
14887
14890
|
echo "Config: loki.yaml (repo root, then ~/.loki/loki.yaml); see docs/loki.yaml.example."
|
|
@@ -36886,7 +36889,6 @@ except Exception:
|
|
|
36886
36889
|
#
|
|
36887
36890
|
# Concretely, this is what reported 2 CRITICAL findings on the v8 PR:
|
|
36888
36891
|
# sk-ant-xxx (loki-ts/test/runner/oauth_dev.test.ts)
|
|
36889
|
-
# sk-ant-test-not-called (loki-ts/tests/runner/quality_gates.test.ts)
|
|
36890
36892
|
# Both exist to prove the SDK path fails closed without a real key. Flagging
|
|
36891
36893
|
# them trains people to ignore CRITICAL, which is how a real one gets missed.
|
|
36892
36894
|
local _ci_secret_deny='(test|example|sample|fixture|dummy|fake|placeholder|redacted|changeme|your[-_]|xxx+|not[-_]called|\$\{|\$[A-Za-z_]|process\.env|os\.(environ|getenv)|<[^>]*>)'
|
|
@@ -265,7 +265,7 @@ def check_quality_gate(trigger, loki_dir, iteration, notifications):
|
|
|
265
265
|
defect as the budget key, in the same file.
|
|
266
266
|
|
|
267
267
|
gate-failures.txt is the canonical failure source used by
|
|
268
|
-
proof-generator.py:413 and
|
|
268
|
+
proof-generator.py:413 and run.sh build_prompt, so this
|
|
269
269
|
reuses a proven contract rather than introducing a second one.
|
|
270
270
|
"""
|
|
271
271
|
failures_file = Path(loki_dir) / "quality" / "gate-failures.txt"
|
|
@@ -273,7 +273,7 @@ def check_quality_gate(trigger, loki_dir, iteration, notifications):
|
|
|
273
273
|
try:
|
|
274
274
|
if failures_file.exists():
|
|
275
275
|
# Cap the read: a pathological file must not be loaded whole
|
|
276
|
-
# (same reason as
|
|
276
|
+
# (same reason as proof-generator.py:413).
|
|
277
277
|
raw = failures_file.read_text(errors="replace")[:8000]
|
|
278
278
|
seen = set()
|
|
279
279
|
for name in raw.replace("\n", ",").split(","):
|