nisse-dev now runs only the conversational loop. The gated state machine
it was built alongside is gone:
- Removed control.py, planner.py, executor.py, approvals.py, gates.py
(command_form moves to loop/reader.py), publish.py and poller.py, the
old assembly in agent.py (its shared helpers move to loop/build.py),
the store's state-machine view (get_state, transition), their tests,
the old end-to-end and live planning/milestone/gate tests, and
specs/orchestrator.qnt with its scenarios.
- Roles are driver, helper, expert and classifier; driver is required.
loop:, roles.planner, roles.executor and tuning.executor are refused
with what to use instead; tuning.driver takes the turn cap and failure
limit.
- The benchmark and the capability matrix measure the driver: the
loop's own system prompt and tools, the task's plan as the agreed
restatement. Task plans end with finish_milestone, so earlier runs
show as (old); stored runs' executor tuning reads as the driver's.
- A delegation on the loop records what its helper's alias resolved to
(L9), as the old path's did.
- The dump shows a work item's revision and what it waits for.
- Behaviour the old tests covered and the loop shares stays covered:
driver runs that fail on a model call, stop mid-turn, reset the
failure count; go-aheads ignoring case and punctuation; a request too
big to fit is never sent; the assembly tests, now on build_loop_agent.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
- A near miss for abandon ("abort", "cancel", "drop it", ...) in an item's
thread is asked about, and changes nothing.
- The driver is told only the operator stops or closes an item, so it
doesn't claim to have done either.
- An issue assigned again after its item closed starts a new item that
notes the one before (Gitea's timeline gives the assignment time);
item_for_issue returns the latest item.
- Abandoning an issue Gitea won't close because of open dependencies says
so, naming them.
- An agreed item isn't offered restate.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
Found live on 2026-10-07 (bring-up on nisse-testing-operator/nisse-bringup,
Opus as driver):
- The driver's prose nudge went into the system prompt, so the retry ended
on the model's own reply. Opus 5.5 refuses that ("does not support
assistant message prefill. The conversation must end with a user
message"). The nudge is now a harness fact in the conversation. The
delegation harness's prose path did the same and is fixed likewise.
- A helper's first request ended with its own delegation_started event,
rendered as an assistant message: every delegation would have failed on
Opus. The helper now sees it as the harness handing over its task.
- model_contract checks that every request ends on the user's side, so
scripted tests catch this from now on (each fix is mutation-checked).
- `status` (and `details`) in an item's thread were taken as replies: they
woke the driver and cancelled a bug's countdown. They now answer at once
and change nothing.
- The driver had no way to answer a remark without changing the item, so it
answered in prose. Every run woken by a message (or a review comment) now
offers `reply`, which answers where asked and returns to the very wait
that was interrupted.
Also: scripts/bringup-repo.sh makes the bring-up's test repository
(make_repo/delete_repo, split out of fresh_repo), and made repositories get
the agent's labels (bug, discuss, design).
Tests: test_loop_live_fixes.py (one per finding); the walk sends `status`
into threads (status_changes_nothing) and its scripted driver sometimes
answers remarks with reply.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
- Every PR: lint, unit tests across the runner's cores (pytest-xdist), 16
property walks from a seed that moves with the run number, integration
tests, and the end-to-end smoke tests (the slice, a conversation filed
as work, a question; `smoke` marker).
- A PR labelled e2e:full, a push to main, or a manual run: all 48 walks and
the whole end-to-end suite. Adding e2e:full to a PR runs it; any other
label runs nothing.
- A published release runs the full set before its image is built.
- The walk's seeds come from NISSE_WALK_SEEDS / NISSE_WALK_FIRST (48 from 0
by default).
Measured here: the unit tier on 2 workers took 3m49s against 6-8 minutes
serially.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
A 502, 503 or 504 from the LiteLLM proxy, or a dropped connection, is
retried after 5, 15 and 45 seconds before the call fails. Twice on
2026-10-07 a short burst of 502s stopped work outright: the integration
tier at 16:39 and an e2e conversation at 20:43, where the item waited on
the operator after one refused call.
Not retried: a 500 (it may be the request's fault), a timeout (it already
had its request_timeout_s), or a budget refusal. Rate limits keep their
single 20-second retry. The time spent waiting is reported with the rate
limit's (last_rate_limit_wait_s), so it isn't counted as the model's.
Tests: an outage waited out, one that outlasts the backoff, a dropped
connection, a 500 and a timeout not retried, a budget refusal not waited
on. The offline agent in test_agent_build skips the backoff (its proxy
doesn't exist).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
- Questions never change state (L5, #77): a message ending in "?" to a
waiting project item starts a question run (reply and look-only tools,
an instruction to change nothing), and the item goes back to the very
wait it was in: a restatement up stays up and a go binds to it; the
agent's own question stays open.
- Explanations: a restatement gives its reasons and the main alternative;
a revision says what changed (restate(changed), sent back without it);
a milestone's summary says what was done, verified and left open.
- Progress in prose (#42): the driver's and a helper's own words for a
turn are posted, trimmed, at most once a minute per item; `details`
shows the last tool calls; a message arriving mid-run is acknowledged at
once ("Noted; ...").
- CI: the e2e job's comments and the plan no longer claim a newer push
cancels the run it supersedes (Gitea here doesn't); the concurrency
block goes.
Tests: conversation unit tests (questions in a go-ahead and an answer wait,
an unanswerable question, revisions, progress rate and trimming, details,
mid-run notes), the helper's progress, the walk asking questions in every
state with a questions_change_nothing property (routing questions the old
way fails 32 of 48 seeds), and an e2e question while a restatement is up.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
- roles.driver (unset: the planner's model) and roles.classifier (unset:
the expert's); the mapping is pinned when an item opens, and the driver
and the go-ahead reader run on the item's own (L9).
- profiles: in configuration (tools, strengths, weaknesses), used in
briefings: the driver's instructions say what its helper and expert are
good at, the helper's brief who reads its handback; never a model name.
- delegate(task) and ask_expert(question) for the driver. A helper works
on a branch off the milestone's; on a handback its work comes back as
uncommitted changes, otherwise it is discarded, so harness branches
never hold commits only the clone has. ask_expert keeps its budget.
- Triggers: failure_limit failed commands in a row (Driver's new "stuck"
outcome) or the turn cap, mid-milestone, call a helper in once per
milestone; if it fails too, the item waits on the operator. A
delegation cut off by a crash is discarded at boot.
- Evaluation capture: delegations and ask_expert as before, and the
reader's verdicts as classifier pairs.
Tests: help tests (merge-back, discard, needs a milestone, the stuck
trigger and its once-per-milestone limit, the expert and its budget,
pinning, briefings by profile, boot discarding a cut-off delegation),
config tests for roles and profiles, and the walk delegating with a
scripted helper that sometimes fails, with a delegations_settle property.
The walk found helper commits stranded on a merged PR's branch, fixed by
returning a helper's work uncommitted.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
Found by the first local e2e run: qwen3.8-27b put a new project's code in
the same PR as its ARCHITECTURE.md, with no build plan. D1 says a `design`
item agrees its design before any code, and that rested on wording only.
- Until a PR of a `design` item has merged, finish_milestone is sent back
if the milestone changes anything but ARCHITECTURE.md and docs/, naming
the paths to take out. The merge is recorded (design_merged) and code
may follow.
- The label's instruction and the new project's brief say so plainly:
documents first, in a PR of their own; code after it has merged.
Tests: a design milestone with code is sent back and only the documents
are published, then code after the merge; the walk's model follows the
rule.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
- propose_project(name, purpose, decisions) in a conversation goes up for
a go-ahead (checked: a slug, not an existing project, no trailing
question).
- On go, the provisioning checklist (D1): a repository with a README,
protection on its default branch, the agent as a collaborator, a build
VM, a deploy key, and the project entry as YAML with <...> to fill in.
- On "done", the agent reads its configuration again and adds the new
project live (build_loop_agent(load=...); only additions), then checks
what was provided (D4, verified, not assumed): permissions and
protection as it sees them, then on a throwaway branch with an empty
commit, a branch push must succeed and a push to the default branch and
merging its own PR must be refused. The check's PR is closed and its
branch deleted. Failures are said per item; nothing starts.
- Checked: the tracking issue (the distilled decisions, carrying the new
item's ID), and the project's item, agreed and labelled `design`, whose
first milestone is ARCHITECTURE.md and the build plan. A conversation
waiting on provisioning doesn't close as idle.
- The walk replays exactly from its seed (item IDs too), and event IDs
stay unique across simulated restarts, as a server's are.
- The test config holds output to 8192 tokens, under the test key's
per-request cost cap.
Tests: new-project unit tests (checklist and its YAML parse as config, done
before the entry, what isn't set up, a checked project's record and first
PR, an existing name), protection enforced by a hook on the fake forge,
the walk proposing projects and saying done, and an e2e new project in a
repository the test operator provisions from the agent's checklist.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
- A top-level message that isn't a command is a conversation: an item in
the `general` scope (L1), threaded under the operator's message, with
`reply`, `ask_human` and `propose_work` and no workspace or execution.
`general` is reserved as a project name.
- propose_work(project, title, summary) goes up as a revision for a
go-ahead (or a plain yes). On go, under a file_issue intent (L8), the
record issue is opened in that project, assigned to the agent and
carrying the new item's ID; its item starts agreed, with the summary as
its brief. Intake skips any issue carrying an ID (D1); boot adopts a
filed issue or puts the proposal up again.
- A conversation closes after tuning.loop.general_idle_s (default a day);
a reply in its thread then starts a new conversation that continues it.
Tests: conversation unit tests (threading, filing, the marker, an unknown
project, changes, idle close and continuation, a plain yes, both crash
cases, commands at top level); the walk with conversations, proposals and
idle time, and records_have_one_item / conversations_do_no_work; an e2e
conversation that becomes filed work.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
- Config: `projects: {name: entry}` with repo, base_branch, workspace,
build_host, push credential, labels (defaults added to every issue),
secrets by ${VAR} reference with visibility (commands | harness), and
egress (recorded, not enforced). Every problem in every entry is
reported at once; an inline secret, a bad visibility, a repeated repo or
a non-slug name is an error.
- The old loop runs exactly one project (only_project) and refuses more.
- The new loop is per project: each its own clone, workspace (with its
commands-visible secrets, redacted on return), push credential and
milestones; intake polls every project's repository; items are scoped
to their project for every run, publish and PR watch. Issues are named
project#14 when there is more than one project, and a bare #14 that
matches two is asked about. Switching projects parks the one worked on
last (all others after a boot).
- A restatement that ends in a question is sent back to the driver, not
posted: the e2e driver kept asking "Is this the right interpretation?",
which a plain yes can't answer.
Tests: config entries and errors, two-project assembly, two-project loop
tests (intake, own workspaces and repos, #14 in both, do-next, label
defaults, parking across projects), the property walk over two projects
with a new own_repository_only property (an item using the first
project's repo fails 15 of 16 seeds), and the restatement check.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
- Every milestone branch is cut from the latest default branch: the next
milestone builds on the merged previous one, and a redo after a closed PR
drops the rejected work. Uncommitted work comes along.
- While the item's PR is open, start_milestone is refused, with the PR's
state read from Gitea at that moment, not the last watch round.
- finish_milestone(final): earlier PRs say "Part of #N", the last "Closes
#N"; unsaid, it's final, as before.
- close_item with the item's PR still open waits for the merge (L2: a
closed item leaves nothing open).
- The driver's instructions cover ARCHITECTURE.md and build plans under
docs/plans/ (D2); the harness never parses a plan.
- Parking stashes uncommitted work and pushes it to <item>-parked, put back
on resume, instead of committing it onto the item's branch. The walk
found a half-done review fix parked onto a PR that had merged meanwhile,
where it was lost. Architecture D4's parking line says so.
Tests: unit tests for fresh milestone branches, the merge wait (including
a merge not yet watched), redo after a closed PR, Part of/Closes, close
waiting on an open PR, and the parked-fix regression; the walk's new
milestones_start_fresh property (fork-from-current fails 9 seeds); an e2e
two-milestone feature.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
- Commands that name their item ("go #14", "abandon #14", "stop #14")
apply to it wherever they're said; go, approve and abandon alone at top
level are asked about (L5). stop is never taken as an answer.
- A plain-language reply to a restatement is read by GoAheadReader, one
model call that agrees only on a clear yes; questions never reach it. The
agent says "taking that as a go-ahead on rev-N" first, bound to the
revision on screen at arrival (L3, L5, D1).
- Labels steer (D1): `bug` goes ahead after tuning.loop.bug_grace_s
(default 600 s) unless the operator replies; `discuss` isn't offered
restate until the operator has spoken; `bug`, `discuss` and `design` add
to the driver's instructions. A restatement asks no questions.
Tests: label and command unit tests, the property walk with labelled
issues, plain yeses, named gos and time passing (grace_only_unanswered,
discuss_talks_first), and e2e tests for a bug proceeding after its grace
period and a plain-language go-ahead.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
Several work items on the new loop (L1, D4):
- Switching parks the item on the workspace: its work is committed and its
branch pushed; resuming checks the item's branch back out. Branches of
open items are never set aside by a fresh workspace.
- Next item: the operator's "do #N next" first, otherwise the oldest.
- Issues with open Gitea dependencies are blocked until they close, checked
at intake and on every round; status shows blocked items and the line.
Tests: queue unit tests, the property walk over several issues with
dependencies and a new parked_work_safe property (it catches parking
disabled), and an end-to-end test with two issues where B depends on A.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
From the new loop's first live run (sandbox #255, PR #256):
- A question asked on the PR is answered on the PR (an inline one quoted
with its file and line, since the API can't reply in its thread), with
a pointer in the item's Matrix thread; one asked in Matrix stays there.
- wait_for_review: when a review needs no change, or only the operator's
merge is left, the item waits on the merge instead of on an answer.
- finish_milestone takes the PR's title from the driver.
The e2e test's merge waits out Gitea's brief 'try again later' after a
push.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
The first live run of the new loop's e2e test found that a fact recorded
while start_milestone was in flight landed between the tool call and its
result; both API formats reject that, so every later request in the run
failed. Views now keep each result right after its call. Scripted test
models check the APIs' ordering rules, so this can't pass unit tests
again.
The e2e tests no longer depend on the sandbox's history: each run asks for
a uniquely named function and works against a base branch of its own,
deleted afterwards. The review step asks for a change that can't already
be there. The scripted operator gives up after five fruitless answers, and
each run keeps its state store for inspection. The driver's prompt says
review fixes stay in the current milestone.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
test_loop_properties.py walks the new loop with a seeded mix of operator
and Gitea actions (go-aheads, stale go-aheads, answers, comments, merges,
the comment-then-merge race, closes, abandon, a crash and restart) and an
adaptive scripted model, checking every loop.qnt property after every
step. It found two bugs, both fixed with regression tests: abandoning an
item left its open PR open (the model's fix 1, now in code), and after a
review fix became a follow-up PR, the next fix went to the merged branch.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
A work item starts from a fresh workspace: fetch, set aside (never
discard) leftover changes, branches and unpushed commits, a clean default
branch, and a branch of its own (D3, #62). start_milestone cuts the
milestone's branch; finish_milestone commits only the work, pushes and
opens a PR that closes the issue, under an intent, adopting an open PR and
treating a closed one as the operator's answer (D4). Between runs the
loop watches each item's PR: operator comments and inline review notes
(#69) reach the item, a merge or close hands back to the driver. A review
fix is checked after it's pushed and goes out as a new PR if the PR merged
first (D5). At boot an interrupted publish is settled against Gitea (L8).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
nisse_dev.loop runs alongside control.py, chosen with `loop: new`.
Issues assigned to the agent (by an operator, or filed by one) become work
items, each with a Matrix thread. A driver run is one model with tools
until it calls a waiting tool: restate puts rev-N up for a go-ahead, bound
to the revision on screen when the reply arrived; ask_human waits for an
answer; close_item closes the item and its issue. Before the go-ahead the
driver can look but not change or close anything. One item is active at a
time; replies in an item's thread go to it. On boot, an item cut off
mid-run waits on the operator, and waiting points are put up again.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
Transport.post(text, thread=...) returns the event ID, which can root a
thread; a threaded post carries m.thread with the reply fallback for
clients without threads. Inbound messages carry their event ID and thread
root: a message in a thread gives its root, a plain reply joins the
thread of the message it replies to when known, and the quoted lines of a
rich reply are no longer part of the text. The in-memory transport plays
threads for tests.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
A failing CI test printed a test key through SSHKey's default repr. Push
credentials, AgentConfig's tokens and MatrixConfig's token are now
excluded from their reprs, so no failure or log line can print them. In
CI's container killed ssh-agents linger as zombies, so the tests count
only live ones.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
The clone gets the standard build artifacts in its own exclude file, so
they never show as changes, even to a model's `git add -A`. The
harness's commits leave the same patterns out even when the project
tracks them, as the sandbox's main did. What was left out is reported in
the publish note and the PR description. Found again in bring-up run 3
(sandbox PR #208).
Fixes#80.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
Live run 3: the executor ran `git checkout -b <task id>` itself, and
publishing then failed, because git can't create '<task>/milestone-1'
while a branch named '<task>' exists. Milestone and delegation branches
are now '<task>-milestone-N' and 'delegation-milestone-N-F'. D4's
harness-owned branches (M2.1) remove the cause; this removes the
collision on the current loop.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
The push credential is either a push-only account's token, handed to git
through a credential helper that names environment variables, never
values, or a deploy key in a short-lived ssh-agent. The remote URL
follows the credential. Configured as gitea.push_user and push_token, or
gitea.push_key.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
execution.SSHHost.execute is now the only way the agent runs anything
(L13). Commands travel as a script on SSH's stdin, so the secrets handed
to a call are never on a command line, and every value handed out is
redacted from the output. Timeouts kill the remote process group, and
an unreachable host is reported as such, not as a failed command.
Workspace and GitRepo run on the build VM. The clone uses the
repository's SSH URL, so no credential is stored in it (#89); fetches
and pushes get the push key over stdin, in a short-lived ssh-agent,
and PRs are still opened from the agent's host with its API token.
The bench runs on the build host too.
Tests reach "the build VM" through a throwaway local sshd (nisse_dev
.testing); CI installs openssh-server. The live push test now pushes
with the sandbox's deploy key (NISSE_PUSH_KEY).
Fixes#89.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
A model no longer reads another role's turns as its own (#78). The log
stays one record per work item; what a role is shown is projected from
it by Transcript.view() (views.py), which the planner, executor and
helper now build their context from:
- own events unchanged;
- operator messages labelled with their channel, e.g.
[operator, Matrix] or [operator, comment on PR #165];
- orchestrator facts (a new `fact` event) as [orchestrator] statements;
- the executor sees a helper's handback as [helper report], not its turns;
- a helper sees the approved plan and the executor's milestone work as
one-line narration;
- the planner sees no other role's work at all.
The orchestrator now writes facts at the points the models must know
about: work item opened, plan approved or rejected, milestone started,
finished or stopped, published (with PR number and URL) or not, review
comment answered, next milestone, task completed or abandoned.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
Development Plan M1.1, as discussed.
- Schema v4: a work_items table (id, scope, title, status, outcome,
position, data) and an agent row whose active_item points at one
(NULL: idle). Canonical events and intents carry their work item, and
events their actor. Older stores are refused with a clear message:
there was no production data, so no migration (the mechanism stays).
- The log is reached through a Transcript bound to one work item and
one actor: store.transcript(item, actor) reads only that item and
writes with both. The planner, executor, delegation and expert desk
get one per call from the orchestrator instead of holding the store,
so a component can neither read another task's events (the planner
read every earlier task's, #78) nor write without attribution.
store.all_events() is kept for readers that span items (evaluation
capture, dumps). Intents are begun through the transcript or with
work_item=, so each belongs to its item.
- The orchestrator opens a work item on 'use', closes it with its
outcome (completed or abandoned), and carries what the operator said
while idle into the new item as copies with their origin, so a task
sent before 'use' is still planned from (D1). control.py keeps its
state machine through get_state()/transition(), now over the active
work item, until M2.6 retires it.
- nisse_dev.dump shows work items and each event's item and actor.
Tests: test_work_items.py (scoping, attribution, intent ownership,
restart, outcomes, carried messages, a second task never seeing the
first); the suites updated for the per-call transcript.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
Architecture L9's role names, for the two roles with a direct
counterpart: the delegation role is now 'helper' and the ask_expert role
'expert', in RoleMap, configuration (roles.helper, roles.expert),
recorded events and tests. 'planner' and 'executor' stay distinct for
now; they become 'driver' when the new loop merges them (Development
Plan M2.3).
Configurations must rename their keys: nisse-dev.example.yaml and
tests/integration/nisse-test.yaml are updated here.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
Development Plan M0.2.
- scripts/bringup.sh: start [REF] sets the previous run aside under
/root/nisse-e2e/runs/, exports the ref under test and starts the agent
as the transient systemd unit nisse-bringup, so it outlives the shell
that started it; stop (SIGTERM, clean shutdown), status, logs, and
collect (the journal log and a readable transcript).
- nisse_dev.dump (python -m nisse_dev.dump STATE_DB [--full]): a readable
dump of a state store's state, pending intents and canonical log. Opens
the store read-only, so it is safe against a live agent.
- The agent runs as root on this development VM, as it will on its own
VM. Architecture D3/D6/D7 corrected to match the operator's intent:
installing packages and toolchains on its own VM is class 2, recorded
so the environment can be rebuilt (it was class 3 with provisioning an
open question); egress names OS package mirrors. The plan's M0.2, M4.1,
M4.2 and bring-up guards follow.
Acceptance: a smoke run started, answered status in the test room,
stopped cleanly, and collect produced the log and transcript.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
Development Plan M0.1. Every section reference in code, tests, specs/,
the example config, the tuning guide, and the CLAUDE.md and README files
now cites docs/Architecture.md (P, L, D, O, X, Q) instead of the old
spec's A-H, and docs/Architecture-Build-Spec.md is deleted (git keeps
it). The plan's old-to-new table goes with it.
Mostly mechanical, with a per-site choice where one old section split:
A4 became L2 where it meant stop/abandon/the loop and L3 where it meant
approvals; A10 became L6 in the context-budget code and L7 elsewhere; B2
became D4 for outbound Gitea and publishing and D5 for review intake;
'A3 step N' became the matching step of L8's boot sequence. Comments
describing behaviour the architecture drops (planning always escalating,
thresholds tuned at an old build step) now describe the current code and
point at the Development Plan.
docs/Architecture.md gains one line the rewrite missed: intake outages
delay but never fail work (old B2), now in L5, where the poller and
Matrix code cite it.
Comments and docs only; no behaviour change.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
Fixes counterexamples 6 and 7 from specs/orchestrator.qnt, the last two.
6. A crash while the executor answered a review comment lost it: its
cursor is saved before the work (at most once), and a restart in
PR_OPEN resumed quietly. Now pr.reviewing marks the comment until
the executor returns (every way it returns, but not a crash), and
boot resumes the watch just before it, with a notice. The mark stays
until it is answered, so a second crash cannot lose it either.
7. 'project complete' left the plan issue open. It now closes it with a
note, after IDLE is saved, as abandon does (shared helper).
The model marks and redelivers, and closes the issue; scenarios 6 and 7
now expect their properties to hold. A mutant without redelivery fails
scenario 6. All eight properties held together over 300,000 traces.
No xfail reproductions remain.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
Fixes counterexample 5 from specs/orchestrator.qnt. A crash after the
PR was opened but before PR_OPEN was recorded left the agent in
EXECUTING; the restart question's 'continue' and 'start' then published
again, Gitea refused a second PR, and there was no way back to PR_OPEN.
On the live run the executor model happened to notice the existing
work; recovery must not depend on that.
- Boot (A3 step 4): in EXECUTING, if Gitea shows the milestone's PR
open, adopt it, watch it from the start, and resume in PR_OPEN with a
notice. Gitea unreachable or the PR closed: the restart question, as
before.
- MilestonePublisher.publish is idempotent: after pushing, an open PR
for the branch is adopted (and so updated) rather than a second one
opened. This also covers a PR opened whose response was lost.
- The model adopts too, gains the lost-response outcome, and now counts
PRs opened per milestone (one_pr_per_milestone was a flag the fixed
model could never set). A mutant without adoption fails the new
scenario; the properties held over 200,000 traces each.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
Fixes counterexamples 1-4 from specs/orchestrator.qnt: a repeated
'approve', 'approve rev-1' meant for the architecture, approving on both
channels, and 'approve' plus an afterthought all landed on the next
plan, one the operator had not seen.
- Revisions are numbered across the work item (A4 already required
monotonic tags; the code numbered per gate), so rev-N names one plan.
- Each reply is stamped on arrival with the revision that was up: in
Matrix the one whose approval request had been sent (a plain value
the transport thread reads, restored on boot), in Gitea the latest
plan comment older than the reply (board.posts).
- Gate.evaluate applies a reply that names no revision only if its stamp
is the current revision. Otherwise an approval is not applied and
feedback rejects nothing (it is still in the transcript), and the
agent says which revision is up. 'approve rev-N' binds by its number.
- Spec A4 states both rules. The model applies them; scenarios 1-4 now
expect the properties to hold, and both held over 200,000 random
traces and Apalache's exhaustive check to depth 8.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
specs/orchestrator.qnt models control.py: states, gate, pending
questions, comment cursors, publish intents and the worker queue, with
operator messages, Gitea comments, work outcomes (exceptions included)
and a crash at any point interleaving freely. Eight properties.
Seven counterexamples, each written out as a trace in
specs/orchestrator_scenarios.qnt and reproduced against the real code in
nisse-dev/tests/test_quint_counterexamples.py (xfail, strict, until
fixed):
1-4 an approval or feedback lands on a plan the operator had not seen:
a double 'approve', 'approve rev-1' meant for the architecture,
approving on both channels, and 'approve' plus an afterthought;
5 a crash after the PR opens but before PR_OPEN publishes again and
cannot get back to PR_OPEN;
6 a crash while answering a review comment loses it silently;
7 'project complete' leaves the plan issue open.
no_stranding, await_input_has_a_question and execution_needs_approval
held over 200,000 random traces of 40 steps, and under Apalache's
exhaustive check to depth 8, which does find counterexample 1 at that
depth. No behaviour changes here.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
A phone keyboard's "Approve" did not match the approval pattern, so it
counted as rejection feedback and threw the plan away; "Stop" did not
reach the cancel flag; other capitalised commands got a hint instead of
acting.
gates.command_form() is how a message reads as a command: case,
surrounding whitespace, inner runs of whitespace and a trailing '.' or
'!' do not matter. The approval gate and every command match in the
orchestrator (stop, status, abandon, use, start milestone, next
milestone, project complete, the restart choices) use it. Only matching
changes: the project name in 'use <name>' keeps its case, and feedback
and answers are recorded and passed on as written. A stale revision is
still refused, and approval still needs the whole message to be the
command.
Closes#64.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
A4 already names abandon as one of the ways out of a halt (continue,
re-plan, or abandon) and 0.2 closes a work item as completed or
abandoned, but no command did it: from a gate or a question the only way
to drop a task was to edit the state database.
- 'abandon' ends the task from any waiting state and returns to IDLE.
The move is made durable first; then the plan issue and any open
milestone PR are closed with a note (branches kept), and any
unresolved intent is aborted so boot reconciliation does not gate on
work dropped on purpose. A Gitea failure is reported, not fatal.
- Like 'status' it is a command in every state, never read as an answer
or feedback. While work runs, 'stop' comes first, as before.
- The restart / stranded question and the PR_OPEN hints mention it.
- Spec: B1 gains the abandon edge and B2 the command.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
- Record a comment cursor per thread (board.last_comment, the field that
already existed unused; pr.last_comment), advanced durably when a
comment is taken up. CommentPoller.resume_after() skips comments at or
below it, so re-attaching needs no request and cannot replay.
- Boot re-attaches the plan watch at an open gate and the review watch in
PR_OPEN from their cursors, catching up on comments posted while the
agent was down (#59). Data saved without a cursor skips what is there,
so an old approval cannot approve the current revision.
- A restart in PR_OPEN resumes with a notice instead of the restart
question (#58). A stored restart question for PR_OPEN returns to it on
either reply instead of re-planning the milestone.
- A stopped or failed review response gets its own pending kind; the
reply returns to PR_OPEN and resumes the watch from the cursor, so
comments that arrived meanwhile are delivered. It used to return to
milestone sign-off, where starting the milestone would open a second PR.
- Comments ignored because of the state leave the cursor alone, so they
are delivered once the agent is back where they apply.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
The first live run found four dead ends one at a time: states where no
operator message could move the agent. Most had one shape: work ended,
by an exception or by a path that never transitioned, and left the agent
in PLANNING_* or EXECUTING, where every message is answered "busy".
The worker only prints an exception, so that would keep happening.
- Every job the orchestrator runs for an operator message or a Gitea
comment now goes through _settled(). If the job leaves a working state
behind, whatever the reason, the agent moves to AWAIT_INPUT with the
continue / re-plan choice a restart offers, and says why. An exception
in a waiting state is reported and the state kept, since its commands
still lead on.
- A sign-off state whose gate record is missing asks to plan again
instead of crashing on every reply.
- The stub configurations (no executor, no publisher) complete the
milestone and wait in PR_OPEN instead of resting in EXECUTING.
test_no_stranding.py drives a whole two-milestone project with a fault
injected at each planner, executor and publisher call, and asserts after
every message that the agent is not in a working state and that the
project still reaches 'project complete'.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
On the first live run the executor committed its milestone itself, as
its plan asked. The publisher only published a commit it made, found a
clean tree, reported "no changes to publish", and left the agent in
EXECUTING, where every message is answered "busy".
- publish(): with nothing left to commit, publish anyway when the branch
has commits origin/<base> lacks; report no changes only when it has
none.
- push_followup(): same for review fixes, comparing HEAD with what the
remote branch holds (ls-remote; pushes go by URL, so there is no
tracking ref).
- Nothing to publish now gates to AWAIT_INPUT, and the reply returns to
the still-approved milestone plan.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
Clients put the addressee's name in front of a message that mentions
it ("Nisse Dev Testing: re-plan"), and the transport passed that text
on whole, so no command matched: the live run's re-plan was asked again,
and an addressed 'approve' would have counted as rejection feedback.
- A leading mention of the agent (an HTML pill, its user ID, or its
display name followed by ':') is removed before commands are matched.
The display name is learnt from the agent's own member event.
- A message whose leading mention is someone else is skipped: in a
shared room it is addressed to another agent or person.
- Only the start counts; a mention later in the text is left alone.
- The transcript records the message as sent (InboundMessage.original).
The sender is unaffected: it still comes from the event's homeserver-set
sender field, never from the text.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A restart in AWAIT_INPUT replaced the saved pending item with a generic
"continue or re-plan?" whose answer nothing handled: the reply cleared
it and left the agent in AWAIT_INPUT with nothing to resume, where 'use'
was refused too. The first live run hit this after a planning failure
and could not be moved by any message.
- Waiting on the operator is a committed state (A3 step 8): on boot, ask
the saved question again instead of replacing it. _await_input keeps
the prompt with the pending item for this.
- The restart question (A3 step 7) records the state it interrupted, and
'continue' / 're-plan' now do what they offer: re-run the interrupted
planning turn, or return an interrupted milestone to its approved plan.
Any other reply asks again rather than being swallowed.
- With nothing pending, 'use <project>' starts over, as the reply
already advised.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
It led the executor benchmark (PR #55): 37/40 over eight tasks and 17/20
on the harder four, ahead of claude-haiku-4-5's 15/20. local-fast, the
previous example, scored 0/8 on the harder four. Model selection is on
the dev executor alone for now; the shared-host question (spec Part H)
stays open.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
All seven newer local models passed the original four tasks 20 of 20,
so the benchmark no longer told them apart. The new tasks keep plans
explicit, as the planner is told to write them, and make the work
itself harder:
- trace-bug: a wrong multi-currency total whose cause is a shared
rate helper; discounts and tax go through it too, so a fix at the
call site fails the check.
- refactor-rename: rename load_config across a ten-module package,
leaving load_config_defaults and reload_config alone.
- spec-parser: implement parse_duration from an eight-rule SPEC.md
given three examples, and write tests for every rule.
- long-milestone: eight dependent steps adding due dates to a task
tracker, from the model through storage versions to the CLI.
Self-tests prove that a call-site fix, a blind text replace and a
partial rename all fail their checks.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The report grouped runs by model, serving note and task version, so
runs made at different tuning were silently averaged together. Tuning
is now part of the key, and a tuning column shows how a row's settings
differ from today's defaults; it appears only when some row needs it.
tuning.executor.max_turns goes from 24 to 40. On the harder benchmark
tasks Haiku, the calibration ceiling, hit 24 on both refactor-rename
runs and both long-milestone runs, one of which had finished the work.
The cap guards against a model trying forever; it isn't a budget.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Two setup failures found while calibrating the new benchmark tasks,
both of which would also fail real milestones.
Devstral's chat template (Mistral) rejected any turn in which the model
wrote some text before a tool call: the OpenAI renderer made one message
per event, so the turn became a text message and a tool-call message in
a row, and llama.cpp answered 500. The renderer now folds consecutive
user or assistant messages into one, which is the OpenAI format for a
turn anyway. Checked against the live server: the captured failing
request no longer occurs, over six real runs.
LiteLLMClient gave up on any request after a fixed 120 s. A dense 27B
model writing a whole module in one turn takes longer, so the executor
escalated work that was going well. llm.request_timeout_s now sets it
for both clients, default 600.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The client waits RATE_LIMIT_RETRY_S and retries once on a 429, silently.
A benchmark run that waited looked slower than the model is: on
2026-09-27, fast models under the unit-test key's 200k tokens/minute
limit had 20-45 s runs that were probably waits.
Both clients now report last_rate_limit_wait_s after every call, even
one that then fails. The benchmark sums it per run into
rate_limit_wait_s, the report's time column leaves it out, and a new
429 column counts the runs that waited. Results stored before this
load unchanged.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Choosing a local executor model, and tuning its knobs, needs measured
success rates across repeated runs, comparable with every model
measured before.
- Four canned milestones under nisse_dev/bench/tasks: write a function,
fix a bug a failing test exposes, thread an option through several
files, and find one failure in output far past max_output_chars. Each
is an approved plan over a seed project, judged by a hidden check the
model never sees; a reference solution proves each is solvable and
that its check tells the difference
- Runs use the agent's own client wiring, now shared as
make_llm_clients, with delegation and ask_expert off: an escalation is
a failure of the model under test
- Each run is stored as one JSON line in bench/results/, committed, with
the model's backing name and declared facts, a note on how it was
served, the task's content hash, the agent's commit and tuning, the
outcome, turns, misbehaviour counts, context trims, tokens, time and
LiteLLM's cost; a fallback answer can never count as a success
- The report rebuilds the comparison from every stored run, with Wilson
95% intervals, marking runs against older task versions
- scripts/benchmark.py runs and reports; scripts/evaluate_model.sh takes
a new model through the context probe, the capability matrix and the
benchmark in one command
- Clients keep LiteLLM's per-call cost, and BudgetedClient counts the
requests it had to trim
- The tuning guide gains "Benchmarking models"
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The executor ended a milestone on any reply without a tool call, and
nothing held a model to the offered tools. Measured on the 4070:
tool_choice "required" is enforced by vLLM (local-fast) and Anthropic,
but qwen3-14b (local-coder) under it cannot stop and writes until the
output ceiling, a known model bug. So enforcement is asked for per
model, and the loop holds the contract itself everywhere.
- The catalog reads supports_tool_choice from LiteLLM's model_info;
BudgetedClient sends tool_choice only to models declared true, as
"required" on the normalized route and {"type": "any"} on the
Anthropic passthrough, and only when tools are offered
- The executor gains an explicit finish tool, the only way to end a
milestone. Finishing with nothing changed is refused once; a prose
reply gets one nudge, then escalates; a call to a tool not offered
counts toward failure_limit, like a failed command; calls after
finish are recorded as not run
- Clients record the model that actually answered. ModelFacts.answered
recognises a proxy fallback, which once credited Claude's answers to
a local model that could not call tools at all
- The context probe reports a fallback instead of trusting it; rerun,
both local models now refuse over-long prompts, as they should
- The capability matrix checks every answer came from the model asked,
that models call only offered tools, and that a declared
supports_tool_choice is true
- Spec A15 and the tuning guide say how enforcement is requested,
declared and backed up
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Every turn re-sent the whole transcript with nothing checking it against
the model's window. On a local model that fails silently: Ollama drops
the start of an over-long prompt, which is where the system prompt and
the plan are. Measured on the Jetson's llama3.2, cut at 4096 tokens with
no error.
- The model catalog reads max_input_tokens and max_output_tokens from
LiteLLM's model_info, and startup refuses a role on a model with no
declared window
- BudgetedClient wraps both routes. It fits the canonical events before
rendering: older tool results cut to head and tail, then the oldest
turns dropped with a note, keeping calls with their results; a request
that still cannot fit fails with ContextOverflow and is never sent.
The output ceiling is the smaller of config and the model's limit
- Tokens are estimated from characters, pessimistically, and the
estimate is corrected from reported prompt sizes, only ever upward
- The executor takes its scope from the log: the latest plan, which is
the approved milestone plan, goes in its system prompt, and it sees
only events after it
- Tool output keeps its head and, mostly, its tail, so a test runner's
failure summary survives truncation
- tuning.context.keep_recent_turns and safety_margin are new knobs
- scripts/probe_context.py checks a declared window against what the
server does; it confirmed llama3.2 cuts at the declared 4096
- The tuning guide covers the budget, declaring a local model's limits,
and the probe
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
With no production instance yet, there is no reason to carry two ways
of configuring the agent. Every environment variable that duplicated a
config key is gone; the environment now supplies only the secrets the
file references by ${…}, plus NISSE_CONFIG and NISSE_ENV to find the
files.
- AgentConfig loads from a file (load, from_file) or a parsed tree
(from_tree, for tests); a missing file is an error naming the example
- The from_env constructors on LiteLLMClient, AnthropicPassthroughClient,
GiteaClient, MatrixConfig and RoleMap are removed, with the env
reading for max output tokens; clients take their settings from config
- The live integration tests share a committed agent config,
tests/integration/nisse-test.yaml, aimed at test resources, and one
conftest that loads it and builds clients. NISSE_LIVE_PLANNER_MODEL
becomes its planner role; NISSE_CAPABILITY_MODELS stays as a per-run
switch
- A missing secret now fails the live tests instead of skipping them
- CI passes only the three tokens
- .env.example is secrets only; spec A5 says config comes from the file
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
dongarandClaude Opus 5.5