M2.8: switch over to the new loop

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

scripts/benchmark.py

@@ -1,4 +1,4 @@
-"""Benchmark executor models on canned milestones, and compare the results.
+"""Benchmark driver models on canned milestones, and compare the results.
 
     uv run python scripts/benchmark.py run --models A,B [--runs 5] [--tasks T,...]
         [--note "vLLM 0.29, AWQ, fp8 KV"] [--config PATH]
@@ -14,13 +14,13 @@ import os
 import sys
 from pathlib import Path
 
-from nisse_dev.agent import build_host, make_llm_clients
 from nisse_dev.bench.results import append, load, render, results_file, summarise
 from nisse_dev.bench.runner import agent_commit, run_task
 from nisse_dev.bench.tasks import load_tasks
 from nisse_dev.catalog import ModelCatalog
 from nisse_dev.config import AgentConfig, ConfigError
 from nisse_dev.envfile import load_env_file
+from nisse_dev.loop.build import build_host, make_llm_clients
 
 REPO = Path(__file__).resolve().parents[1]
 RESULTS = REPO / "bench" / "results"
@@ -46,7 +46,8 @@ def run(args: argparse.Namespace) -> int:
     catalog = ModelCatalog.discover(config.llm_base_url, config.llm_token)
     catalog.require_limits({m: m for m in models})
     normalized, _passthrough = make_llm_clients(config, catalog)
-    host = build_host(config)  # runs happen on the build VM, as the agent's do
+    # Runs happen on the first project's build VM, as the agent's do.
+    host = build_host(config, config.projects[0])
     commit = agent_commit(REPO)
 
     for model in models:
@@ -54,7 +55,7 @@ def run(args: argparse.Namespace) -> int:
         out = results_file(args.results, model)
         for task in tasks:
             for n in range(args.runs):
-                # The executor always takes the normalized route in the agent,
+                # The driver always takes the normalized route in the agent,
                 # so it does here too.
                 record = run_task(
                     task,

scripts/e2e.sh

@@ -53,7 +53,7 @@ print(i["spend"], i.get("max_budget"))' 2>/dev/null) || echo "? ?"
 }
 read -r before budget < <(spend)
 status=0
-uv run pytest -m e2e -rs tests/integration/test_end_to_end.py tests/integration/test_end_to_end_loop.py "$@" || status=$?
+uv run pytest -m e2e -rs tests/integration/test_end_to_end_loop.py "$@" || status=$?
 read -r after _ < <(spend)
 python3 -c "
 b, a, cap = '$before', '$after', '$budget'

scripts/evaluate_model.sh

@@ -1,5 +1,5 @@
 #!/usr/bin/env bash
-# Evaluate a model for the executor role, from a declared LiteLLM alias to a
+# Evaluate a model for the driver role, from a declared LiteLLM alias to a
 # benchmark report next to every model measured before it.
 #
 #   scripts/evaluate_model.sh ALIAS "how it is served" [RUNS] [--config PATH]