Repository navigation
Releases
0.9.1
[0.9.1] — 2026-11-08
Changed
- The README is a landing page again. The long design document moved to
docs/DESIGN.md; the README now leads with the one-command install, the
measured A/B, the two routes and the full command list. It still names no
concrete model: the model is a fact about your account, not a recommendation
that can be shipped. - The README carries the A/B numbers directly, including the result where
the kit loses: on a task that fits in one read it uses more tokens and takes
longer for the same answers. - A recorded demo and a social preview ship under
docs/assets/. The demo
shows a realinitagainst a free model, the ten lanes, and the install
verifier agreeing with the profile.
0.9.0
[0.9.0] — 2026-11-05
Changed
- No model profiles ship. The kit is the harness, not a list of models:
profiles/index.jsonis empty anddefaultisnull. The old profiles
moved totest/fixtures/profiles/and are test data, not choices. - No model retention entries ship either.
retention.ymlnow carries only
theunknowndefault and documentation; a person's verdicts live in
~/.config/opencode/retention.jsonand are written by
use <provider>/<model> --retention <level>. Fixture verdicts moved to
test/fixtures/profiles/retention.jsonso the suite still exercises them. - The README documents two routes, free and paid/upgraded, and names no
concrete model. A model id in the README would be a recommendation that goes
stale the day the route is withdrawn; the test suite now fails if one is
added. oc-harness use <provider>/<model> --retention <level>creates a
profile. It reads the model from the live catalogue, writes
~/.config/opencode/profiles/<id>.json, and records the retention verdict
in~/.config/opencode/retention.json.--idoverrides the generated name.oc-harness init --model <provider>/<model>creates the profile and then
installs and verifies it. Without a terminal or--model,initprints the
command instead of guessing a model.- Profiles are discovered in the config directory first, then in the
package directory, so a user profile and a fixture profile cannot shadow
each other silently. list,auto,upgrade,verifyandlanessay when there is no profile
yet and print theusecommand that creates one, instead of failing on an
empty selection.- No runtime path names a model. The invariant is enforced against the
fixture ids and models, so a future edit that puts a concrete choice in
bin/,src/, the prompts, templates, skills, schema or README fails the
suite. Historical measurement docs may keep their names; nothing executable
may recommend one. doctorwith no profiles asks for a model instead of printing "all
profiles match" over an empty list, and<command> --helpprints the help
rather than starting the command and failing for want of a terminal.- CI and the V2 gate use the test fixtures explicitly instead of relying on
profiles that no longer ship, andselfcheckcopies the fixture profiles
into a temporary directory before installing, so it never writes test data. - Project metadata:
SECURITY.mdnow lists the supported versions for the
0.8.x and 0.7.x lines, the Contributor Covenant 2.1 applies to every
community space, and.mailmapfolds the historical author spellings into
one maintainer identity.
0.8.0
[0.8.0] — 2026-11-04
The loop becomes something you can leave running. run now fixes a failed
verify or review instead of stopping at the first one, work is the one
command that plans, builds, verifies, reviews and finishes, and the run
records enough for a script - or the person after you - to see what happened.
The loop is now observable and resumable rather than only restartable: a run
leaves evidence, an event log and a record a script can read; a worktree run
can be found after the terminal closes; and a route that died is not retried
blind.
Added
-
oc-harness work "<request>". Plan when there is no executable plan,
then build, verify, review, final verify, final review and summary. Flags:
--plan-only,--dry-run,--branch,--commit,--worktree,
--allow-dirty,--max-tokens,--max-task-minutes,--final-verify,
--final-review. It never pushes, opens a pull request or publishes. -
oc-harness plan "<request>". The planner's answer is validated before
it is written: every task needsfiles:,change:and averify:that can
fail. A rejected plan never replaces an existing one.--printshows it
without writing;--forcereplaces an existing plan. -
oc-harness status [--json]. The last run from.space/run.json:
status, reason, per-task attempts, review rounds and checkpoint commits. -
Plan preflight.
runchecks the whole plan before the first agent is
called: no tasks, a missing verify, a verify that cannot fail, or two tasks
the parser cannot tell apart all stop the run at exit 2. -
Fix rounds. A failed verify is handed back to the build agent with its
real output; a failed review is handed back with its blockers. Default 2
rounds each (--max-fix-rounds,--review-fix-rounds). -
Final checks.
--final-verifyruns one command after the last task;
--final-reviewreviews the whole diff from the run's base commit.
.space/summary.mdis written only when both pass. -
Checkpoints.
--branchand--commitcommit each finished task and
only the paths it touched.--worktreeruns in.space/worktrees/<id>so
the project itself is untouched while the run happens. -
Route failover inside a run. A spent or unreachable route no longer
marks the task blocked;workretries once per other route in the lane
chain usingOC_HARNESS_MODEL, and never on the route that failed. -
oc-harness lanes --suggest. Names lanes the telemetry says are not
earning a slot. It reads; it never editslanes.jsonor a profile. -
A local profile,
ollama-qwen3-coder. Qwen3 Coder 30B served by Ollama
on localhost: zero egress, zero-retention by construction, no per-token
cost. Pull it withollama pull qwen3-coder:30b. -
npm run test:loop, and a named CI step for it: end-to-end scenarios
through the real CLI with a stand-inopencode, covering a fix round, a
review fix, an out-of-scope edit, a vacuous verify, a token budget and a
failed final verify. -
oc-harness autopilot: from an empty directory, initialise a blank
track, tick the charter, run research/design/decompose, then call
work --project. It refuses an existing.spacewithout atrack.json
instead of overwriting a hand-written track.--jsonkeeps stdout to one
object;--dry-runstill runs the model-backed spec steps and only makes the
final work step a preflight. -
Milestone gates in the built-in templates.
blank,inventory,
billingandexistingnow put- gate: {{test_command}}on M1 and M2, so
spec decompose --milestone M1works on a freshly initialised project. The
last milestone may omit its gate because there is no next milestone to gate. -
oc-harness finishruns the real completion gate. It refuses missing or
empty plans and unfinished tasks with exit 2, runs the track's acceptance
command (exit 4 on failure), then the install check (exit 1 on drift), and
reports the dirty/publish boundary without publishing. -
resolve: lane models drawn from the catalogue instead of written down.
A profile may declare"resolve": { "from": "catalogue", "free": true, "chainDepth": 3 }and carry noroleModelsat all. Atusetime the kit
keeps the models that are free, at or below the profile's retention ceiling,
and able to serve the lane —visionneeds images,schemaneeds structured
output — and records the result in<config dir>/lane-resolve.json.
--offlinereplays that lock and refuses when there is none or when it
belongs to a different profile, so an offline install is reproducible rather
than freshly guessed. A lane with both a hand-written chain and aresolve
block is refused: one lane, one source of models. -
big-pickle-auto, the first profile built onresolve. Ten lanes, base
modelbig-pickle,may-train, refreshed at install time. Opt-in, and not
the default. -
oc-harness lanesnow answers the two questions a reader had to open the
profile index for: whether a lane costs money (free) and who trains on it
(retention). A footer names where the models came from — the profile file,
or a lock and how old it is. -
The
specworkflow: a validated plan before the loop starts.oc-harness spec initlays a track down from a template (inventory,billing,
existing, orblank),spec checkproves the charter still has its locked
sections and a ticked## Ready, andspec research,spec designand
spec decomposemake exactly one model call each and write what it
returns into.space/. Each step is budget-checked before the call and
size-checked after it, is refused rather than retried when the model is
offline, times out or returns junk, and cannot be re-run without--force;
--printand--dry-runnever reach the model at all, so a template is data
and the data commands cost nothing. Readiness lives in.space/track.json
(charter_ready→research_ready→design_ready→plan_ready) and
setting a key clears every later one, so a plan cannot stay "ready" after the
design it was built on changed.spec decomposeruns the generated
tasks.mdthrough the same parserworkuses, and refuses a plan whose task
has noverify, whose[REQ-NNN]tag is not in the charter, or whose verify
command is not runnable. Every refusal is exit2with a JSON body on
stderr and an empty stdout, listed indocs/EXIT-CODES.md. -
oc-harness work --project: run a planned track and nothing else. It
takes no positional request — the request is.space/tasks.md— and refuses
before it looks foropencodewhen the directory has no track or
plan_readyis nottrue.--strictis forced off, because strict mode
exists to make an under-specified request fail and a planned track is the
output of the steps that remove that ambiguity. The refusal prints to stdout
with exit2, matching plainwork --json, and--dry-runruns the same
preflight without writing a run record. -
A ceiling on what one session may spend (
effort.sessionTokens). The kit
bounded one request (compaction at 130k) and one task (40 calls for a build),
and neither bounds a session: the measured one made 769 requests, every one
of them under the ceiling, and spent 67.93M prompt tokens until the route
answered "Free usage limit reached".fleet-safenow sets 6,000,000 tokens;
the telemetry hook charges each finished message (input + cache read + output, counted once per message id), and the guard throws a three-line
message at 70% and refuses every tool call at 100%. The message is prose, not
a rule map, because opencode echoes a rule map back as the refusal and the
measured one was 11,041 characters.0turns the ceiling off, which is what
every install had before this; a value under one 200,000-token window is
rejected as a profile that would stop a session before its first request.
test/budget.test.jscovers the arithmetic, both thresholds, the
once-per-session warning, and the guard actually refusing a call - the last
one is why the test bundles the guard and the counter into one graph. -
eval/token-burn.mjs(npm run token-burn). The reading that was missing
when a session on a free route spent 67.93M prompt tokens, peaked at 169,597
tokens in one request and ended with "Free usage limit reached": requests,
prompt and output tokens, the worst request and when it happened, how many
tokens were re-sent in refusals, and what the tool output weighs — per day,
from the database the session already wrote. It does not report prune
coverage, because opencode stores a part unpruned and the pruner rewrites the
outgoing request: a stored prune marker means the session was resumed from a
pruned transcript, not that pruning ran. The gate is
the fleet's own compaction threshold plus the size of one tool result, judged
only on requests made after the running config was installed, so the history
that motivated the fix does not fail every run forever. -
test/token-burn.test.js, the token-burn suite. Eleven tests, each
pinning one factor ofrequests x context: compaction at 130k rather than
176k, 40 calls per build rather than 120, no lane narrowing bash with a
catch-all rule map, no lane prompt carrying a copy of the allowlist, the
refusal text under its measured size,tool_outputwide enough that aread
is not truncated, a denial ending the turn, and compaction running the pruner.
The last test runs the audit itself against a database it builds, including
the case where the gate fails. -
oc-harness report [--json]. One run as numbers: phases and their
tokens, verify count and failures, routes used, recorded error codes, and
the next action. -
oc-harness resume. Continues the recorded run, keeps its id, tokens,
attempts and budget, and refuses a finished run or an edited plan. ...
0.7.3
[0.7.3] — 2026-10-26
Documentation only. No code, config or behaviour changed.
The 0.7.2 tarball was built from 8bf0283, before the README sentence that
describes npm run selfcheck was corrected on main. It said the release
check ran last; it runs right after the build, against the tarball that build
produced. 0.7.2 users have the old sentence in node_modules; this release
republishes the corrected README.
0.7.2
[0.7.2] — 2026-10-26
A patch about the release process itself: the checks that decide whether a
release is coherent now live in one command and run in CI.
Added
npm run release:check. It comparespackage.jsonwith both versions
inpackage-lock.json, the first release heading in this file, an optional
tag, and the filesnpm pack --dry-runwould publish.--releaserequires
the CHANGELOG heading;--tag v<version>requires the tag. It is part of
npm run selfcheck, runs in the test job without--release, and runs in
the publish job with the pushed tag.
Fixed
CONTRIBUTING.mdasked for two commits and two tags.npm version
already makes both by default, and the steps after it made a second one. The
release steps now usenpm version --no-git-tag-version, verify the tree
before committing, and tag once.models.ymlclaimed a check it did not run. The scheduled job reads
catalogue lifecycle; it cannot see a route that answers 200 with no text.
The comment now says so and nameseval/models.mjs --free --check.- README named
doctoras the pre-release check. It is the online check.
npm run selfcheckis the release gate, and the README now says which is
which.
Documentation
MIGRATION.mdcovers 0.4.x → 0.7.x. The config file format survives, but
the compiler now owns more keys:runis build + verify,--reviewexits 5
on a rejected review, thebuildcap is 120,docandverifyno longer
open bash with"*": "allow", denied tool calls are not retried, the
reviewer moved off the retirednemotron-3-ultra-free, anddoctor --check-modelschecks lifecycle rather than presence.
0.7.1
[0.7.1] — 2026-10-26
A release about checks that were green while the thing they check was not:
a withdrawn model reported healthy, a lane on a model that cannot do the
lane's job, a config from the previous kit reported as matching the current
profile, and a free route that answered nothing still recommended as a
candidate.
Fixed
doctor --check-modelsread presence, not lifecycle. models.dev keeps
a withdrawn model and marks itdeprecated; the check only asked whether a
record existed, so a pulled base or fallback rung was reported healthy.
It now asksisWithdrawn, and the message says withdrawn-or-removed.- A lane's route was never checked for the job it does. A live model that
cannot call tools, avisionlane on a text-only model, a missing context
or output ceiling, and an effort level the model cannot express all passed
doctor. Each is now reported by lane, against the first live rung of the
chaininstallwould use. A level that snaps down is still allowed, because
that is the documented behaviour; a level that resolves tonullis not. verifycompared a hand-written key list.harness.kithad been added
to the compiler and not to the list, so the live install - compiled by
0.6.4, still carrying 25 ECC subagents - passed as matching the profile.
The check now recompiles the profile over the installed file and compares
every key the compiler owns. A config from an older kit is a note;
verify --strictmakes it a failure.- The probe could not tell an answer from a spent route.
eval/models.mjs
now classifiesok,empty,timeoutanderror, treats a finished step
with no text asempty, and exits non-zero under--check. - A free route that answered nothing was still recommended. Two live
probes ofnemotron-3-ultra-freeproduced no step and no text. The route
is recorded ineval/unusable-models.json;--freeandresearchskip it
until it is re-probed, and the standalone profile and its reviewer rung are
retired. The reviewer lane now runs on GLM 5.3 Flash.
Added
eval/models.mjs --free --listselects live free routes without spending a
request and names the ones a probe rejected.eval/graph.mjschecks the import graph and prints the routing graph:
--checkfails on an unresolved import, a module cycle, or a source file no
entry or test reaches;--routingprints profile -> lane -> model with the
catalogue's verdict on each rung.--jsonand--dotare for a machine.npm run graph,npm run test:coverage(85% lines, 70% branches, 80%
functions), andnpm run selfcheck, which runs build, typecheck, the suite,
the graph,premises, and a fresh install in a temporary config directory
ending inverify --strict.test/effort.test.jscovers the config mutation directly: the router's
session model, a lane's preserved model and prompt, anullvariant
deleting a stale one,harnesskeys surviving, and health entries keyed by
provider.test/trust.test.jscovers a replacement that fails after the
revoke and has to say the package now has no publisher.- CI runs
graph,test:coverageandpremisesbeside the suite, and the
install smoke test usesverify --strict. A dailymodels.ymljob asks the
live catalogue whether every shipped model is still live.
Changed
fleet-safeandfleet-allare refreshed against the live catalogue:
fleet-safepublishes three effort levels (low,high,max) and
fleet-allno longer claims image attachments on a model that does not
read them.- README and
docs/ROUTING.mdquote the current fleet: reviewer on GLM 5.3
Flash, build and longform on Flint, and the seven lanes that move under the
zero-retention ceiling.
0.7.0
[0.7.0] — 2026-10-25
oc-harness run now stops for the reasons it already claimed to stop for,
the read-only lanes no longer have a write-capable shell, and the token budget
counts the context a request reads instead of only its uncached part.
Fixed
- A denied tool call was fed back as another turn, and the retry spent the
window.experimental.continue_loop_on_denywastrue, so opencode
handed the guard's refusal back to the model and continued the loop. The
measured session made 138 requests and 7,559,016 context tokens in 29
minutes on sixshared token windowdenials, each retry re-sending the
whole prompt. The generated config now sets it tofalse, and the guard
blocks a single request that re-sends more thanOC_HARNESS_TURN_STOP
(default 300k) with "do not retry". - The in-event shared-window stop did not refuse anything. When the
step-finishsum crossedOC_HARNESS_BUDGET_WINDOW_STOP, the guard set
windowSpentbut never setwindowWhy, sosharedWindowSpent()returned
an empty reason and every tool call was allowed. The reason is set where
the sum crosses, and the stop is written to the ledger with
recordStop()so a new session cannot outlive the provider's window. - A corrupt ledger locked the machine for six hours, or vanished. It is
now moved tobudget.json.corruptand the window starts empty; the broken
bytes are kept as evidence.recordSpend()andrecordStop()repair
before writing instead of reading the corruption as an empty window. - The bundled ECC subagents were loaded on every request. Their
descriptions came fromprompts/agents/*.txt, were not lanes, and so were
not covered by the stale-lane rule. They are removed from the generated
agent map when that is the prompt, and named in the config header; a
hand-written agent that reuses that prompt path is removed like the
bundle's own, while any other prompt is left alone. - The token budget undercounted a cached session five to forty times, so it
never fired. Astep-finishpart reportsinput,cache.read,
cache.write,outputandreasoning. The guard summedinputalone -
the part of the prompt that was not cached - while the cached prefix, which
is what a long conversation actually re-reads, was ignored. One measured
stream reported 844 uncached tokens against 36,480 cached ones on the same
step, then 391 against 9,088. The 2M warn and 5M stop were therefore five
to forty times the real spend, and the provider's window ran out before
either fired.requestTokens()insrc/budget.jsnow countsinput + cache.read + cache.writefor both the session total and the shared-window ledger; the
toasts and refusals say "context tokens" because that is what they count. - The A/B's
--budgethad the same undercount. The suite cap summed
inputTokens, so 120 runs reported about 1.3M of spend while their streams
carried 7.1M.parseStreamnow also returnscontextTokens, the cap and the
report use it, and--rescorere-reads the old streams to put the corrected
number on them.inputTokensis still recorded so a table can be compared
with the ones written before this. - The test suite wrote to the real window ledger.
guard-hook.test.js
exercised the guard with the defaultOC_HARNESS_BUDGET_FILE, so every test
run added its 1,500 and 600 tokens to
~/.local/state/opencode-harness-kit/budget.json- the live ledger showed
twelve test entries and nothing else. The tests now point the ledger at
their own temporary directory. --rescoregraded a timeout as a wrong answer. A run that hit the
ceiling has a stream that stops mid-sentence, which reads exactly like an
answer that cited nothing; only the sidecar knows which it was. Four runs
that timed out at 12 steps were re-scored as four wrong answers, and the
table then showed 18/23 where the honest count is 18/19.- A resume crashed the report on a sidecar from an older version. The
re-score readsescaped,toolsandtextfrom every result; a sidecar
written before those fields existed threw aTypeErrorafter every run had
already been paid for. Missing fields are read as empty, and the corrected
token count is written back so a second resume does not depend on the stream
still being there. - The A/B's preflight said "ok" to a run that had failed. It checked only
the timeout and the quota, so a run that reported an error event and exited
in three seconds passed. The first suite with that preflight spent eighteen
runs learning that every kit-off arm had failed with
ProviderModelNotFoundError, scored as "no output, not scored". It now
reads the stream, and any error event stops the suite with exit 2 and the
provider's own message. An exit 0 with zero steps and no text is the same
shape - a spent route - and is refused the same way. - The kit-off arm could not resolve the model it was told to use. opencode
keeps its model catalogue in~/.cache/opencode/models.json, and the arm's
temporaryHOMEhad no cache, so a baremodel:string was a model the
installation had never heard of. The cache is opencode's, not the kit's, so
it is symlinked into the temporary home the same way~/.local/share/opencode
already was. - An error in a measured run was counted as a wrong answer. A run that
produced no tool call and no text, but did report an error, now stops the
suite at once with that error instead of waiting for--fail-fastand then
scoring the refusal. - The lane counter's "no telemetry" test failed on a machine that had one
two directories up. The lookup walks up to the project root by design, so
/tmp/.space/telemetry.jsonl- written by anyopencode runstarted in
/tmp- made it find a file and fail for a reason that was not the code's.
The test now runs twelve levels down, past the walk-up's ceiling, so nothing
above its own tree is looked at; the walk-up itself is unchanged. - The A/B was governed by the kit's own budget. The kit-on arm runs on the
real config, so the guard's session and window thresholds applied to it: 120
runs spend more than the window's stop, and from the run that crossed it the
guard refused tool calls - every later run would be scored as a wrong answer
for a reason the suite is not measuring. The suite now sets all four
thresholds to0for its runs and pointsOC_HARNESS_BUDGET_FILEat the run
directory, so a benchmark neither trips the guard nor charges the account's
real window. The suite's own cap is--budget. - An empty plan reported success. A plan with no tasks printed
"every task is done (0/0)" and exited 0. Nothing was done, so it now stops
with exit 2 and says the plan has no tasks. - A blocked task did not stop the tasks after it. The loop marked the
blocker and carried on, then reported it at the end. It now runs in file
order and stops at the first blocked task, so pending work before the
blocker still lands and work after it no longer builds on a failure. - The task file was written back from the copy read at the start. A
human edit or an agent deletion made mid-run was overwritten, and a deleted
task file was recreated. Write-back re-reads the file, refuses to write a
stale copy, and goes throughrenameso a reader never sees half a plan. - A task was matched by title and position only. With two same-titled
tasks, removing or renaming the first made the loop mark the second one
done. Matching now uses title +change+files, and an ambiguous match
stops the run instead of guessing. verify: trueandverify: echo okwere treated as checks. Both exit
0 whatever the agent did. A verify line made only of no-ops now blocks the
task before the agent runs; a no-op before a real command
(true && npm test) is still accepted.docandverifyopened their bash maps with"*": "allow". The
read-only lanes could write throughsed -i,awkorpython3 -c
regardless of theiredit: "deny".docnow defaults to deny andverify
defaults to ask, each with the commands its job actually needs.- A run that hit the lane's step cap reported only "verify failed". At the
cap OpenCode exits 0 with a text-only answer, so the loop saw a failed
check and blamed the change, while the log held the real reason.runnow
countsstep_finishevents in the transcript and, when the count reaches
the role's configured cap, says so in the stop message and in the task's
recorded output, with the setting to raise. - An agent that answered nothing was reported as a failed change. A spent
model route shows up in one of two shapes. Some providers exit 0 with an
empty stdout and the reason on stderr: the loop then ran the verify against
an untouched tree and blocked the task for work the model never attempted.
Others hang until the ceiling kills them, which at the default budget is two
hours spent on one task with nothing to show for it.
runnow stops with exit 3 when stdout is empty without running the verify,
and a timeout names the route and the limit instead of saying the agent
"failed". The task is marked blocked either way, with the stderr and the
advice in its output. - The A/B ground through a quota error as if it were an answer. The free
tier can print{"type":"error"}with "Free usage limit reached" on stdout
and exit 0; the suite read that as an empty stream and needed--fail-fast
runs before stopping. It now stops on the first one with exit 2 and the
provider's own words, because the refusal measures the provider and not the
kit. - The kit-off arm made paid title calls.
makeNokitHomepinnedmodel
but notsmall_model, so the title agent used opencode's default paid
model. A real run made 1,073 title calls togpt-5.4-nano; every one was
refused for insufficient...
0.6.4
[0.6.4] — 2026-10-23
Fixed, each one found by re-reading what 0.6.3 published, not by running
anything new:
- The install command in the README named a package that does not exist.
The Install section saidnpx oc-harness init"from the registry", and
oc-harnessis this kit's binary name, not a package anyone can fetch: the
command every new reader was pointed at returnsnpm error 404 'oc-harness' is not in this registry. Every install path in the README now reads
npx -y opencode-harness-kit@latest <cmd>, bareoc-harnessis defined
once as shorthand for one of the two real forms, and a test pins every
npxreference in the README to this package. Released because the npm
page is a snapshot: the README on npmjs.com renders from the published
tarball, so fixing it in git changes nothing for the next visitor. profiles/index.jsonlistedflint-40b-freetwice, and the README's
shipped-profiles table repeated it with a second, differently-worded
note.oc-harness listprinted twelve rows,oc-harness initoffered
twelve choices, flint twice in both, and the README counted twelve
shipped profiles. Eleven profiles ship; one entry remains, carrying the
honest note of both. Tests now diff the index, the table and the loader,
with no Set to collapse the difference.retention.ymlrepeatedopencode/flint-40b-free, and the last key
won. The hand-rolled parser overwritesmodels[id]on every repeat, so
the vague 2026-10-22 stub silently replaced the 2026-10-23 record that
quotes the provider's terms verbatim - same retention class, wrong date,
lost detail. A test now scans the file for repeated keys.installfrom npx symlinked into the npx cache. The installer always
symlinked the plugin, prompts, skills and GLOBAL.md back to the package
root - right from a git checkout ("edit here, live everywhere"), visible
breakage from an npx cache thatnpm cache clean --forcedeletes. From a
checkout nothing changes; from npx or any npm install, where.gitis
absent,installnow copies, so the harness on disk does not dangle the
day the cache is cleaned.verifyanduninstalllearn that a
byte-identical copy of the kit's file is the kit's, and a test installs
from a copy of the package with no.gitinto a temp HOME, asserts real
files where links used to go, and runsverifygreen.
No runtime change: dist/, the plugin and every model-facing artifact are
byte-for-byte what 0.6.3 shipped.
0.6.3
[0.6.3] — 2026-10-23
Nothing here changes the runtime. The plugin, the CLI and dist/ are
byte-for-byte what 0.6.2 shipped; this release is the front page and the
packaging metadata catching up with decisions already made.
Removed
- The demo media is gone from the repository.
docs/demo/pitch.mp4
(12.5 MB) anddocs/demo-loop.gifare deleted, and the README no longer
embeds them. The recording lives on in the private
pham-canh/videorepo. Two reasons
it had to leave this one: the.mp4was never shipped by npm (files
coversdocs/*.mdonly), so the README rendered on npmjs.com carried a
<video>tag whose source could not exist — a 404 embed on the public
page; and a repo whose pitch is "the metadata is the truth" was weighting
its own checkout with twelve megabytes nobody could watch.
Fixed
- The lockfile answered for a version two releases behind.
package-lock.jsonstill said0.5.8whilepackage.jsonsaid0.6.2,
and carried apeerDependenciesblock the manifest no longer has. Both are
synced now, by hand rather than a regeneratingnpm install, so the locked
dependency versions are untouched.
0.6.2
[0.6.2] — 2026-10-23
Published. Tag v0.6.2.
Fixed
- Transient catalogue errors stopped benching live routes. When
oc-harness doctoror the router asked models.dev and the request failed
for a reason other than "this model is withdrawn" — a timeout, a rate
limit, a network blip — the failure was percolating into the route health
as if the model itself were dead, and a live route got benched on evidence
that was about the network, not the model. The distinction now matters:
transient errors are retried and reported as transient; only a confirmed
status: "deprecated"out of the catalogue marks a route unhealthy.