Fix the run-breaking tool-result order; e2e test runs the slice (M2.1, PR D)

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

nisse-dev/src/nisse_dev/loop/agent.py

@@ -70,8 +70,8 @@ After the go-ahead, work in a milestone: call start_milestone before
 changing anything (never create or switch branches yourself), make the
 change, run the project's tests, then call finish_milestone with a summary.
 The harness commits, pushes and opens a PR, and you wait for review. Review
-comments arrive as messages: make the changes, run the tests, and call
-finish_milestone again to push the fix. When the PR has merged, call
+comments arrive as messages: make the changes within the same milestone (don't
+start another), run the tests, and call finish_milestone again to push the fix. When the PR has merged, call
 close_item with what was done, what was verified, and what is left open. If
 the PR is closed without merging, ask the operator what next.
 

nisse-dev/src/nisse_dev/views.py

@@ -90,8 +90,9 @@ def project(events: list[StoredEvent], viewer: str) -> list[StoredEvent]:
     shown = _SHOWN.get(viewer)
     if shown is None:
         # A role with no view defined (the expert is stateless, the
-        # classifier sees only an action) reads the log as it is.
-        return list(events)
+        # classifier sees only an action) reads the log as it is, with
+        # results kept after their calls.
+        return _results_follow_calls(list(events))
     results = {
         e.event.call_id: e.event for e in events if isinstance(e.event, ToolResult)
     }
@@ -128,4 +129,30 @@ def project(events: list[StoredEvent], viewer: str) -> list[StoredEvent]:
                 out.append(_said(stored, actor, line))
             elif isinstance(event, AssistantText) and event.text.strip():
                 out.append(_said(stored, actor, event.text.strip()))
-    return out
+    return _results_follow_calls(out)
+
+
+def _results_follow_calls(events: list[StoredEvent]) -> list[StoredEvent]:
+    """Keep each tool result right after its call, as both API formats
+    require: anything recorded while a call was in flight (a fact the
+    harness wrote while handling it, say) moves to just after the results.
+    A call with no result yet keeps its place; the renderer answers it."""
+    out: list[StoredEvent] = []
+    pending: set[str] = set()
+    held: list[StoredEvent] = []
+    for stored in events:
+        event = stored.event
+        if isinstance(event, ToolCall):
+            pending.add(event.call_id)
+            out.append(stored)
+        elif isinstance(event, ToolResult) and event.call_id in pending:
+            pending.discard(event.call_id)
+            out.append(stored)
+            if not pending:
+                out.extend(held)
+                held = []
+        elif pending:
+            held.append(stored)
+        else:
+            out.append(stored)
+    return out + held

nisse-dev/tests/model_contract.py

@@ -0,0 +1,58 @@
+"""What a real model API insists on, for scripted test models.
+
+Both formats the agent sends require a tool call's result to come right
+after it: in Anthropic's, the next message is the user's and starts with a
+tool_result for every tool_use, before any text; in OpenAI's, the assistant
+message with tool_calls is followed directly by a tool message for each. A
+scripted model that doesn't check this lets a broken transcript pass the
+unit tests and be rejected live; this makes it fail here.
+"""
+
+from typing import Any
+
+from nisse_dev.render import OpenAIChatRenderer
+from nisse_dev.render_anthropic import AnthropicMessagesRenderer
+
+
+def assert_renderable(events: list[Any]) -> None:
+    _anthropic(AnthropicMessagesRenderer().render(list(events), None, "m")["messages"])
+    _openai(OpenAIChatRenderer().render(list(events), None, "m")["messages"])
+
+
+def _anthropic(messages: list[dict[str, Any]]) -> None:
+    for index, message in enumerate(messages):
+        if message["role"] != "assistant":
+            continue
+        used = [b["id"] for b in message["content"] if b.get("type") == "tool_use"]
+        if not used:
+            continue
+        following = messages[index + 1] if index + 1 < len(messages) else None
+        assert following is not None and following["role"] == "user", (
+            f"anthropic messages.{index}: tool_use with no user message after it"
+        )
+        blocks = following["content"]
+        leading = []
+        for block in blocks:
+            if block.get("type") != "tool_result":
+                break
+            leading.append(block["tool_use_id"])
+        assert set(used) <= set(leading), (
+            f"anthropic messages.{index}: tool_use ids {sorted(set(used) - set(leading))} "
+            "lack a tool_result at the start of the next message"
+        )
+
+
+def _openai(messages: list[dict[str, Any]]) -> None:
+    for index, message in enumerate(messages):
+        calls = [c["id"] for c in message.get("tool_calls") or []]
+        if not calls:
+            continue
+        answered = []
+        for after in messages[index + 1 :]:
+            if after["role"] != "tool":
+                break
+            answered.append(after["tool_call_id"])
+        assert set(calls) <= set(answered), (
+            f"openai messages.{index}: tool_calls {sorted(set(calls) - set(answered))} "
+            "are not answered by the tool messages right after them"
+        )

nisse-dev/tests/test_loop.py

@@ -6,6 +6,7 @@ from pathlib import Path
 from typing import Any
 
 import pytest
+from model_contract import assert_renderable
 
 from nisse_dev.cancel import Cancel
 from nisse_dev.canonical import AssistantText, CanonicalEvent, Fact, ToolCall
@@ -34,6 +35,7 @@ class ScriptedModel:
         self.seen: list[list[Any]] = []
 
     def complete(self, events, tools, model, system=None) -> list[CanonicalEvent]:
+        assert_renderable(events)  # as the real API would
         self.offered.append({t["function"]["name"] for t in tools or []})
         self.seen.append(events)
         assert self.turns, "the model was called more often than scripted"

nisse-dev/tests/test_loop_milestones.py

@@ -8,6 +8,7 @@ from pathlib import Path
 from typing import Any
 
 import pytest
+from model_contract import assert_renderable
 
 from nisse_dev.cancel import Cancel
 from nisse_dev.canonical import ToolCall
@@ -42,6 +43,7 @@ class ScriptedModel:
         self.turns = list(turns)
 
     def complete(self, events, tools, model, system=None) -> list[Any]:
+        assert_renderable(events)  # as the real API would
         assert self.turns, "the model was called more often than scripted"
         return self.turns.pop(0)
 

nisse-dev/tests/test_loop_properties.py

@@ -13,6 +13,7 @@ from pathlib import Path
 from typing import Any
 
 import pytest
+from model_contract import assert_renderable
 from test_loop_milestones import DON, Forge, call
 from test_loop_milestones import world as world  # noqa: PLC0414 -- the fixture
 
@@ -41,6 +42,7 @@ class PolicyModel:
         self.n = 0
 
     def complete(self, events, tools, model, system=None) -> list[Any]:
+        assert_renderable(events)  # as the real API would
         self.n += 1
         offered = {t["function"]["name"] for t in tools}
         facts = [getattr(e.event, "text", "") for e in events]

nisse-dev/tests/test_views.py

@@ -282,3 +282,18 @@ def test_operator_messages_carry_their_channel(tmp_path: Path) -> None:
         "[operator, comment on plan issue #9] hi",
         "[operator, Gitea comment] hi",
     ]
+
+
+def test_something_written_mid_call_moves_after_the_result(tmp_path: Path) -> None:
+    # Live, M2.1: start_milestone's fact landed between the call and its
+    # result, and the model API rejected every later request in the run.
+    store = new_store(tmp_path)
+    driver = store.transcript("a", "driver")
+    driver.append(ToolCall(call_id="c1", tool="start_milestone", arguments={}))
+    store.transcript("a", "orchestrator").append(Fact(text="milestone 1 started"))
+    driver.append(ToolResult(call_id="c1", result="started", success=True))
+    kinds = [type(e.event).__name__ for e in driver.view()]
+    assert kinds == ["ToolCall", "ToolResult", "UserMessage"]
+    body = AnthropicMessagesRenderer().render(driver.view(), tools=None, model="m")
+    result_first = body["messages"][1]["content"][0]
+    assert result_first["type"] == "tool_result"