Before each driver call, over compact_at (0.6) of the prompt budget, the
same model is asked for notes over the very request it was about to get,
plus one message, so the server's cache covers all but the notes. The notes,
with code's list of the calls they replace, become a compaction event that
renders in place of its range, keeping word for word what started the work
and the last few turns. Rare, large steps keep the prompt's start unchanged
between them. A later compaction folds in the earlier; one from an earlier
run's view is hidden.
Failed notes are logged and fitting takes over. Fitting now applies
compactions before it measures, records what it cuts (Trim) and logs each
cut as context.trimmed. Wired into both agents' drivers and the benchmark;
nisse-dev's helper loop is a follow-up.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01J5Utyfzw7gF7UhjRDEtjTk
Shared parts, not a shared scheduler (agreed): nisse-review will have its
own small scheduler; a common one waits until two exist to compare.
- nisse_core.loop.driver: the driver run, moved; the agent names its
waiting tools per run (nisse-dev's are WAITING_TOOLS), and execution
goes to any Executor (ToolOutcome moves with it).
- nisse_core.loop.tools: the tool helper and the tools any agent offers
(notes, memory, memory_done, ask_expert); nisse_core.loop.expert.
- nisse_core.knowledge.handlers: Knowledge, the vault and memory as
control tools, the index and prompt text, out of LoopAgent.
- nisse_core.config: file reading, references, people, tuning sections of
the agent's own, LLM numbers, the vault.
- New core tests: a driver run with a non-workspace executor and the
agent's own waiting set; Knowledge alone; the shared config pieces.
Docs: the Development Plan marks M7.2 done; Architecture Part L and the
nisse-core README name the new parts.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01J5Utyfzw7gF7UhjRDEtjTk
What isn't the dev agent's moves to nisse-core, unchanged in behaviour,
with its tests: canonical, render, render_anthropic, llm, catalog, probe,
context, views, state, reconcile, worker, cancel, dump, transport, matrix,
execution, testing, vault, memory, oplog, envfile and gitea (whole).
- nisse_core.state.StateStore holds the tables every agent has; an
agent's subclass adds its tables (version, migrations), roles and
views. nisse_dev.state.StateStore adds the build VM blocks (v5).
- views.project takes what each role is shown, rather than a fixed table.
- Vault takes the agent's name (agent=), for provenance.
- A test that nisse_core never imports an agent.
- The loop, config, git, delegation, expert, tools and evals stay in
nisse-dev until nisse-ops has a loop to share.
Docs: the Development Plan marks M7.1 done and notes that Phase 6 waits on
the agent host and Terraform; Architecture Part L names the package.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01J5Utyfzw7gF7UhjRDEtjTk
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
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
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
Calibration showed Haiku fixing the right constant but keeping only four
decimal places, which passes the visible test and still loses the JPY
rate's precision. An explicit plan should say so.
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>
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>
dongarandClaude Opus 5.5