M7.1: extract nisse-core

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-core/tests/test_canonical.py

@@ -0,0 +1,120 @@
+import sqlite3
+from pathlib import Path
+
+import pytest
+
+from nisse_core.canonical import (
+    AssistantText,
+    Compaction,
+    DelegationCompleted,
+    DelegationStarted,
+    GateOpened,
+    PlanSubmitted,
+    ToolCall,
+    ToolResult,
+    UserMessage,
+    from_payload,
+    mint_call_id,
+)
+from nisse_core.state import StateSchemaError, StateStore
+
+ALL_EVENTS = [
+    UserMessage(text="use demo", source="matrix:$abc"),
+    AssistantText(text="starting", model="local/qwen"),
+    ToolCall(call_id="call-1", tool="run_tests", arguments={"path": "tests/"}),
+    ToolResult(call_id="call-1", result="3 passed", success=True),
+    PlanSubmitted(plan="the plan", model="claude-opus-5"),
+    GateOpened(gate="arch", reason="plan submitted (rev-1)"),
+    DelegationStarted(
+        brief="fix the flaky test",
+        role="helper",
+        model="claude-opus-5",
+        caps={"turns": 20, "tokens": 100000},
+    ),
+    DelegationCompleted(
+        handback="test fixed; root cause was a shared tmpdir",
+        artifacts=["nisse-dev/tests/test_x.py"],
+        tokens=51234,
+        duration_s=61.5,
+        exit_status="success",
+    ),
+    Compaction(summary="earlier work summarized", model="local/qwen", replaces=[1, 4]),
+]
+
+
+def test_all_event_types_round_trip(tmp_path: Path) -> None:
+    with StateStore(tmp_path / "state.db") as store:
+        for event in ALL_EVENTS:
+            store.transcript("t", "driver").append(event)
+    with StateStore(tmp_path / "state.db") as store:
+        stored = store.transcript("t", "driver").read()
+    assert [s.event for s in stored] == ALL_EVENTS
+    assert [s.id for s in stored] == sorted(s.id for s in stored)
+    assert {(s.work_item, s.actor) for s in stored} == {("t", "driver")}
+
+
+def test_read_events_after_id(tmp_path: Path) -> None:
+    with StateStore(tmp_path / "state.db") as store:
+        log = store.transcript("t", "operator")
+        first = log.append(UserMessage(text="one", source="m:1"))
+        log.append(UserMessage(text="two", source="m:2"))
+        tail = log.read(after=first)
+    assert [s.event.text for s in tail] == ["two"]
+
+
+def test_mint_call_id_is_unique_and_prefixed() -> None:
+    ids = {mint_call_id() for _ in range(100)}
+    assert len(ids) == 100
+    assert all(i.startswith("call-") for i in ids)
+
+
+def test_unknown_event_type_rejected() -> None:
+    with pytest.raises(ValueError, match="unknown canonical event type"):
+        from_payload("mystery", {})
+
+
+def test_unknown_fields_rejected() -> None:
+    with pytest.raises(ValueError, match="unknown fields"):
+        from_payload("user_message", {"text": "hi", "source": "m", "extra": 1})
+
+
+# The v1 schema as it shipped (PR #2), frozen here to prove a newer image
+# opens and upgrades a database written by the older one (L8).
+V1_SCHEMA = [
+    "CREATE TABLE meta (key TEXT PRIMARY KEY, value TEXT NOT NULL)",
+    """
+    CREATE TABLE agent_state (
+        id INTEGER PRIMARY KEY CHECK (id = 1),
+        state TEXT NOT NULL,
+        project TEXT,
+        data TEXT NOT NULL DEFAULT '{}',
+        updated_at TEXT NOT NULL
+    )
+    """,
+    """
+    CREATE TABLE intent_log (
+        id INTEGER PRIMARY KEY AUTOINCREMENT,
+        kind TEXT NOT NULL,
+        payload TEXT NOT NULL,
+        status TEXT NOT NULL DEFAULT 'pending',
+        created_at TEXT NOT NULL,
+        resolved_at TEXT
+    )
+    """,
+]
+
+
+def test_a_store_from_before_work_items_is_refused_clearly(tmp_path: Path) -> None:
+    # No production data predates work items, so there is no migration from
+    # these versions (Development Plan M1.1): the store must say so rather
+    # than open half-understood.
+    db = tmp_path / "state.db"
+    conn = sqlite3.connect(db)
+    for statement in V1_SCHEMA:
+        conn.execute(statement)
+    conn.execute("INSERT INTO meta (key, value) VALUES ('schema_version', '3')")
+    conn.commit()
+    conn.close()
+
+    with pytest.raises(StateSchemaError, match="predates work items"):
+        StateStore(db)

nisse-core/tests/test_catalog.py

@@ -0,0 +1,202 @@
+"""Runtime model discovery (architecture L9).
+
+The point of these tests is that nothing here reads the model's *name*. A
+semantic alias like "remote-architecture" is free to point anywhere, so the
+route has to come from what LiteLLM says the model is.
+"""
+
+import httpx
+import pytest
+
+from nisse_core.catalog import (
+    CatalogUnavailable,
+    ModelCatalog,
+    UnknownModel,
+    select_client,
+)
+
+
+def entry(alias: str, backing: str, provider: str | None, tools: bool | None = None):
+    """One /model/info row, shaped as the live proxy returns it."""
+    info: dict = {"supports_function_calling": tools}
+    if provider is not None:
+        info["litellm_provider"] = provider
+    return {
+        "model_name": alias,
+        "litellm_params": {"model": backing, "api_base": "http://proxy.test/v1"},
+        "model_info": info,
+    }
+
+
+def catalog(*entries) -> ModelCatalog:
+    return ModelCatalog.from_payload({"data": list(entries)})
+
+
+# --- parsing -------------------------------------------------------------
+
+
+def test_facts_are_read_from_the_payload() -> None:
+    cat = catalog(
+        entry("claude-haiku-4-5", "anthropic/claude-haiku-4-5", "anthropic", True)
+    )
+    facts = cat.facts("claude-haiku-4-5")
+    assert facts.backing == "anthropic/claude-haiku-4-5"
+    assert facts.provider == "anthropic"
+    assert facts.supports_tools is True
+    assert facts.native_anthropic
+
+
+def test_provider_falls_back_to_the_backing_prefix() -> None:
+    # litellm_provider is absent on some deployments; the backing model's
+    # own "vendor/model" prefix says the same thing.
+    cat = catalog(entry("mystery", "anthropic/claude-opus-5", None))
+    assert cat.facts("mystery").provider == "anthropic"
+
+
+def test_unnamed_entries_are_skipped() -> None:
+    cat = ModelCatalog.from_payload({"data": [{"litellm_params": {"model": "x"}}]})
+    assert cat.aliases() == []
+
+
+def test_missing_tool_support_is_unknown_not_false() -> None:
+    # LiteLLM reports null for models it has no capability data on. That is
+    # not the same as declaring the model tool-less, and the startup warning
+    # depends on the difference.
+    cat = catalog(entry("local-fast", "openai/qwen3-8b-awq", "openai", None))
+    assert cat.facts("local-fast").supports_tools is None
+
+
+# --- route selection by provider, never by name --------------------------
+
+
+def test_a_semantic_alias_routes_by_its_provider() -> None:
+    # The whole feature: nothing in "remote-architecture" says Anthropic.
+    cat = catalog(entry("remote-architecture", "anthropic/claude-opus-5", "anthropic"))
+    normalized, passthrough = object(), object()
+    chosen = select_client("remote-architecture", normalized, passthrough, cat)
+    assert chosen is passthrough
+
+
+def test_the_same_alias_reroutes_when_it_is_repointed() -> None:
+    # L9: provisionally accept LiteLLM's normalization for a vendor with no
+    # native adapter, rather than sending it Anthropic-shaped bodies.
+    cat = catalog(entry("remote-architecture", "openai/gpt-6-astra", "openai"))
+    normalized, passthrough = object(), object()
+    assert (
+        select_client("remote-architecture", normalized, passthrough, cat) is normalized
+    )
+
+
+def test_a_name_that_merely_mentions_a_vendor_is_not_that_vendor() -> None:
+    cat = catalog(entry("claude-clone-9", "openai/some-clone", "openai"))
+    normalized, passthrough = object(), object()
+    assert select_client("claude-clone-9", normalized, passthrough, cat) is normalized
+
+
+def test_local_models_take_the_normalized_route() -> None:
+    cat = catalog(
+        entry("local-coder", "openai/qwen3-14b", "openai"),
+        entry("local-fast", "openai/qwen3-8b-awq", "openai"),
+    )
+    normalized, passthrough = object(), object()
+    for alias in ("local-coder", "local-fast"):
+        assert select_client(alias, normalized, passthrough, cat) is normalized
+
+
+# --- misconfiguration is loud --------------------------------------------
+
+
+def test_an_unknown_model_raises_rather_than_guessing_a_route() -> None:
+    cat = catalog(entry("local-fast", "openai/qwen3-8b-awq", "openai"))
+    with pytest.raises(UnknownModel, match="local-fsat"):
+        select_client("local-fsat", object(), object(), cat)
+
+
+def test_the_error_lists_what_is_available() -> None:
+    cat = catalog(
+        entry("local-fast", "openai/qwen3-8b-awq", "openai"),
+        entry("claude-opus-5", "anthropic/claude-opus-5", "anthropic"),
+    )
+    with pytest.raises(UnknownModel, match="claude-opus-5, local-fast"):
+        cat.facts("typo")
+
+
+def test_validate_names_every_bad_role_at_once() -> None:
+    # One boot should surface the whole list, not make the operator
+    # rediscover the next typo on the next start.
+    cat = catalog(entry("local-fast", "openai/qwen3-8b-awq", "openai"))
+    with pytest.raises(UnknownModel) as caught:
+        cat.validate(
+            {
+                "executor": "local-fast",
+                "planner": "clause-opus-5",
+                "helper": "remote-architecture",
+                "expert": "local-fast",
+            }
+        )
+    message = str(caught.value)
+    assert "planner='clause-opus-5'" in message
+    assert "helper='remote-architecture'" in message
+    assert "executor" not in message  # the good ones are not noise
+
+
+def test_validate_passes_a_whole_good_mapping() -> None:
+    cat = catalog(
+        entry("local-fast", "openai/qwen3-8b-awq", "openai"),
+        entry("claude-opus-5", "anthropic/claude-opus-5", "anthropic"),
+    )
+    cat.validate({"executor": "local-fast", "planner": "claude-opus-5"})
+
+
+# --- an outage is not a misconfiguration ---------------------------------
+
+
+def test_an_unreachable_proxy_is_reported_as_an_outage() -> None:
+    def handler(request: httpx.Request) -> httpx.Response:
+        raise httpx.ConnectError("no route to host")
+
+    with pytest.raises(CatalogUnavailable, match="could not reach"):
+        ModelCatalog.discover(
+            "http://litellm.test", "sk-test", transport=httpx.MockTransport(handler)
+        )
+
+
+def test_a_refused_listing_is_reported_as_an_outage() -> None:
+    cat = httpx.MockTransport(lambda request: httpx.Response(401, text="bad key"))
+    with pytest.raises(CatalogUnavailable, match="401"):
+        ModelCatalog.discover("http://litellm.test", "sk-test", transport=cat)
+
+
+def test_discovery_asks_the_proxy_with_the_token() -> None:
+    seen: dict = {}
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        seen["url"] = str(request.url)
+        seen["auth"] = request.headers.get("Authorization")
+        return httpx.Response(
+            200, json={"data": [entry("local-fast", "openai/qwen3-8b-awq", "openai")]}
+        )
+
+    cat = ModelCatalog.discover(
+        "http://litellm.test/", "sk-test", transport=httpx.MockTransport(handler)
+    )
+    assert seen["url"] == "http://litellm.test/model/info"
+    assert seen["auth"] == "Bearer sk-test"
+    assert cat.aliases() == ["local-fast"]
+
+
+# --- what gets pinned ----------------------------------------------------
+
+
+def test_pinning_records_what_the_alias_resolved_to() -> None:
+    # An alias may point somewhere else tomorrow; eval history (L9) keyed on
+    # the alias alone would average across whatever it pointed at.
+    cat = catalog(entry("remote-architecture", "anthropic/claude-opus-5", "anthropic"))
+    pinned = cat.pinned({"planner": "remote-architecture"})
+    assert pinned == {
+        "planner": {
+            "alias": "remote-architecture",
+            "backing": "anthropic/claude-opus-5",
+            "provider": "anthropic",
+        }
+    }

nisse-core/tests/test_context.py

@@ -0,0 +1,300 @@
+"""The context budget (architecture L6)."""
+
+import io
+from pathlib import Path
+from typing import Any
+
+import pytest
+
+from nisse_core.canonical import (
+    AssistantText,
+    CanonicalEvent,
+    StoredEvent,
+    ToolCall,
+    ToolResult,
+    UserMessage,
+)
+from nisse_core.catalog import ModelCatalog, UnknownLimits
+from nisse_core.context import (
+    Budget,
+    BudgetedClient,
+    ContextOverflow,
+    estimate_tokens,
+    fit,
+    head_and_tail,
+)
+from nisse_core.oplog import OpLog
+from nisse_core.state import StateStore
+from nisse_dev.loop.driver import Driver
+
+
+def stored(events: list[CanonicalEvent]) -> list[StoredEvent]:
+    return [StoredEvent(id=i + 1, ts="t", event=e) for i, e in enumerate(events)]
+
+
+def turn(n: int, output: str) -> list[CanonicalEvent]:
+    """One executor turn: a command and its (possibly long) output."""
+    return [
+        ToolCall(call_id=f"c{n}", tool="run_command", arguments={"command": "make"}),
+        ToolResult(call_id=f"c{n}", result=output, success=True),
+    ]
+
+
+def transcript(turns: int, output_chars: int) -> list[StoredEvent]:
+    events: list[CanonicalEvent] = [UserMessage(text="start milestone 1", source="t")]
+    for n in range(turns):
+        events += turn(n, f"output {n} " + "x" * output_chars)
+    return stored(events)
+
+
+def budget(window: int, keep: int = 2) -> Budget:
+    return Budget(
+        window=window, max_output=500, keep_recent_turns=keep, safety_margin=0.1
+    )
+
+
+def results(events: list[StoredEvent]) -> dict[str, str]:
+    return {
+        e.event.call_id: e.event.result
+        for e in events
+        if isinstance(e.event, ToolResult)
+    }
+
+
+# --- keeping the end of long output ---------------------------------------
+
+
+def test_head_and_tail_keeps_the_failure_summary_at_the_end() -> None:
+    output = (
+        "exit code: 1\nstdout:\n"
+        + "collected 400 items\n" * 500
+        + "FAILED tests/test_app.py::test_greet - AssertionError\n"
+        + "==== 1 failed, 399 passed ====\n"
+    )
+    kept = head_and_tail(output, 1000)
+    assert len(kept) < 1100
+    assert kept.startswith("exit code: 1")  # the head keeps the exit code
+    assert "FAILED tests/test_app.py::test_greet" in kept
+    assert "1 failed, 399 passed" in kept
+    assert "characters omitted" in kept
+
+
+def test_short_text_is_untouched() -> None:
+    assert head_and_tail("fine", 1000) == "fine"
+
+
+# --- fitting, step by step -------------------------------------------------
+
+
+def test_a_transcript_that_fits_is_sent_unchanged() -> None:
+    events = transcript(turns=3, output_chars=100)
+    assert fit(events, "system", None, budget(100_000)) == events
+
+
+def test_step_one_cuts_older_tool_results_and_keeps_recent_ones_whole() -> None:
+    events = transcript(turns=6, output_chars=3000)
+    fitted = fit(events, "system", None, budget(4000, keep=2))
+
+    kept = results(fitted)
+    assert len(kept) == 6  # nothing dropped yet
+    for older in ("c0", "c1", "c2", "c3"):
+        assert "characters omitted" in kept[older]
+    for recent in ("c4", "c5"):
+        assert "characters omitted" not in kept[recent]
+
+
+def test_step_two_drops_the_oldest_turns_and_says_so() -> None:
+    events = transcript(turns=20, output_chars=3000)
+    fitted = fit(events, "system", None, budget(4000, keep=2))
+
+    first = fitted[0].event
+    assert isinstance(first, UserMessage)
+    assert "omitted to fit the context window" in first.text
+    kept = results(fitted)
+    assert "c0" not in kept  # the oldest went
+    assert {"c18", "c19"} <= kept.keys()  # the recent turns stayed
+    assert "characters omitted" not in kept["c19"]
+    # Every surviving result still has its call: a provider rejects an
+    # answer to a call it never saw.
+    calls = {e.event.call_id for e in fitted if isinstance(e.event, ToolCall)}
+    assert kept.keys() <= calls
+    assert estimate_tokens(fitted, "system", None) <= budget(4000).for_prompt
+
+
+def test_step_three_refuses_what_cannot_fit() -> None:
+    # Even the two most recent turns are bigger than the whole window.
+    events = transcript(turns=3, output_chars=20_000)
+    with pytest.raises(ContextOverflow, match="last 2 turns"):
+        fit(events, "system", None, budget(2000, keep=2))
+
+
+def test_a_huge_system_prompt_alone_overflows() -> None:
+    events = transcript(turns=1, output_chars=10)
+    with pytest.raises(ContextOverflow):
+        fit(events, "s" * 50_000, None, budget(4096))
+
+
+def test_the_output_ceiling_and_margin_come_out_of_the_window() -> None:
+    b = Budget(window=4096, max_output=1024, keep_recent_turns=6, safety_margin=0.1)
+    assert b.for_prompt == int(4096 * 0.9) - 1024
+
+
+# --- the client wrapper ----------------------------------------------------
+
+
+class RecordingClient:
+    def __init__(self, reported_input: int = 0) -> None:
+        self.calls: list[dict[str, Any]] = []
+        self.last_usage = {"input": reported_input, "output": 5}
+
+    def complete(self, events, tools, model, system=None, *, max_tokens=None, **kwargs):
+        self.calls.append({"events": events, "max_tokens": max_tokens, **kwargs})
+        return [AssistantText(text="ok", model=model)]
+
+
+def test_the_ceiling_is_the_smaller_of_config_and_model() -> None:
+    inner = RecordingClient()
+    client = BudgetedClient(inner, lambda m: (4096, 1024), max_output_tokens=16384)
+    client.complete(transcript(1, 10), None, "llama3.2")
+    assert inner.calls[0]["max_tokens"] == 1024
+
+    client = BudgetedClient(inner, lambda m: (200_000, 64_000), max_output_tokens=16384)
+    client.complete(transcript(1, 10), None, "claude-haiku-4-5")
+    assert inner.calls[1]["max_tokens"] == 16384
+
+
+def test_the_wrapper_sends_the_fitted_events() -> None:
+    inner = RecordingClient()
+    client = BudgetedClient(
+        inner, lambda m: (4000, 500), keep_recent_turns=2, safety_margin=0.1
+    )
+    client.complete(transcript(20, 3000), None, "m", system="system")
+    sent = inner.calls[0]["events"]
+    assert "omitted to fit" in sent[0].event.text
+
+
+def test_reported_usage_only_ever_raises_the_estimate() -> None:
+    events = transcript(1, 10)
+    estimate = estimate_tokens(events, None, None)
+
+    under = RecordingClient(reported_input=estimate * 2)  # we undercounted
+    client = BudgetedClient(under, lambda m: (100_000, None))
+    client.complete(events, None, "m")
+    assert client._factors["m"] == pytest.approx(2.0, rel=0.05)
+
+    over = RecordingClient(reported_input=estimate // 2)  # we overcounted
+    under_then_over = BudgetedClient(over, lambda m: (100_000, None))
+    under_then_over._factors["m"] = 2.0
+    under_then_over.complete(events, None, "m")
+    assert under_then_over._factors["m"] == 2.0  # never loosened
+
+
+def test_last_usage_passes_through() -> None:
+    client = BudgetedClient(RecordingClient(reported_input=7), lambda m: (100, None))
+    assert client.last_usage["input"] == 7
+
+
+# --- the limits come from LiteLLM's catalog --------------------------------
+
+
+CATALOG = ModelCatalog.from_payload(
+    {
+        "data": [
+            {
+                "model_name": "claude-haiku-4-5",
+                "litellm_params": {"model": "anthropic/claude-haiku-4-5"},
+                "model_info": {"max_input_tokens": 200000, "max_output_tokens": 64000},
+            },
+            {
+                "model_name": "llama3.2",
+                "litellm_params": {"model": "ollama/llama3.2"},
+                "model_info": {"max_input_tokens": 4096, "max_output_tokens": 1024},
+            },
+            {
+                "model_name": "undeclared",
+                "litellm_params": {"model": "ollama/qwen3"},
+                "model_info": {"max_input_tokens": None},
+            },
+        ]
+    }
+)
+
+
+def test_the_catalog_reads_each_models_limits() -> None:
+    facts = CATALOG.facts("llama3.2")
+    assert (facts.max_input_tokens, facts.max_output_tokens) == (4096, 1024)
+    assert CATALOG.facts("undeclared").max_input_tokens is None
+
+
+def test_a_role_on_a_model_without_a_window_is_refused() -> None:
+    CATALOG.require_limits({"driver": "llama3.2", "helper": "claude-haiku-4-5"})
+    with pytest.raises(UnknownLimits, match="driver='undeclared'"):
+        CATALOG.require_limits({"driver": "undeclared", "helper": "llama3.2"})
+
+
+def test_a_request_that_cannot_fit_ends_the_run(tmp_path: Path) -> None:
+    # Never sent: the driver's run fails and the item waits on the operator,
+    # like any failed call, rather than a server getting a prompt it would
+    # silently cut.
+    store = StateStore(tmp_path / "state.db", oplog=OpLog("t", io.StringIO()))
+    store.transcript("t", "operator").append(UserMessage(text="go", source="t"))
+    inner = RecordingClient()
+    client = BudgetedClient(inner, lambda m: (4096, 1024))
+    outcome = Driver(client, "m").run(
+        store.transcript("t", "driver"),
+        system="s" * 20_000,  # a system prompt too big to fit
+        tools=[],
+        workspace=None,
+    )
+
+    assert outcome.kind == "failed"
+    assert "fit" in outcome.note
+    assert inner.calls == []
+
+
+# --- tool_choice is only asked of servers that enforce it -------------------
+
+FINISH = [{"type": "function", "function": {"name": "finish", "parameters": {}}}]
+
+
+def test_required_is_asked_only_of_models_declared_to_enforce_it() -> None:
+    # Measured: qwen3-14b under "required" cannot stop and rambles to the
+    # output ceiling, so a model that is not declared is never asked.
+    inner = RecordingClient()
+    client = BudgetedClient(
+        inner, lambda m: (32768, 8192), enforces_tool_choice=lambda m: m == "local-fast"
+    )
+    client.complete(transcript(1, 10), FINISH, "local-fast")
+    client.complete(transcript(1, 10), FINISH, "local-coder")
+    client.complete(transcript(1, 10), None, "local-fast")
+    assert [c["require_tool_call"] for c in inner.calls] == [True, False, False]
+
+
+def test_the_catalog_reads_supports_tool_choice() -> None:
+    catalog = ModelCatalog.from_payload(
+        {
+            "data": [
+                {"model_name": "a", "model_info": {"supports_tool_choice": True}},
+                {"model_name": "b", "model_info": {"supports_tool_choice": False}},
+                {"model_name": "c", "model_info": {}},
+            ]
+        }
+    )
+    assert [catalog.facts(m).supports_tool_choice for m in "abc"] == [True, False, None]
+
+
+def test_a_fallback_answer_is_recognised() -> None:
+    facts = ModelCatalog.from_payload(
+        {
+            "data": [
+                {
+                    "model_name": "claude-haiku-4-5",
+                    "litellm_params": {"model": "anthropic/claude-haiku-4-5"},
+                }
+            ]
+        }
+    ).facts("claude-haiku-4-5")
+    assert facts.answered("claude-haiku-4-5")  # LiteLLM's alias
+    assert facts.answered("claude-haiku-4-5-20251001")  # the provider's own name
+    assert facts.answered(None)  # nothing reported
+    assert not facts.answered("claude-sonnet-5")  # somebody else answered

nisse-core/tests/test_dump.py

@@ -0,0 +1,80 @@
+import io
+from pathlib import Path
+
+from nisse_core.canonical import Fact, PlanSubmitted, ToolCall, ToolResult, UserMessage
+from nisse_core.dump import main, render
+from nisse_core.oplog import OpLog
+from nisse_core.state import StateStore
+
+
+def store_with_a_run(path: Path) -> None:
+    store = StateStore(path, oplog=OpLog("nisse-dev", io.StringIO()))
+    store.transcript(None, "operator").append(
+        UserMessage(text="hello before use", source="matrix")
+    )
+    store.open_work_item(
+        "demo-1",
+        title="Add a truncate helper",
+        scope="demo",
+        position="WAITING",
+        data={"loop": {"wait": "go-ahead", "rev": 1}},
+        cause="test",
+    )
+    store.set_active("demo-1", cause="test")
+    log = store.transcript("demo-1", "operator")
+    log.append(UserMessage(text="Add a truncate helper", source="matrix"))
+    driver = log.as_actor("driver")
+    driver.append(PlanSubmitted(plan="x" * 500, model="opus"))
+    driver.append(
+        ToolCall(call_id="c1", tool="run_command", arguments={"command": "ls"})
+    )
+    driver.append(ToolResult(call_id="c1", result="README.md", success=False))
+    log.begin_intent("open_pr", {"branch": "b"}, cause="test")
+    store.close()
+
+
+def test_the_dump_shows_items_intents_and_each_event_with_its_actor(
+    tmp_path: Path,
+) -> None:
+    db = tmp_path / "state.db"
+    store_with_a_run(db)
+    text = render(db)
+    assert "active work item: demo-1" in text
+    assert (
+        "demo-1  open  position: WAITING  scope: demo  rev-1  waits for: go-ahead"
+        in text
+    )
+    assert "pending intent #1 open_pr [demo-1]" in text
+    assert "[agent/operator] USER [matrix]: hello before use" in text
+    assert "[demo-1/operator] USER [matrix]: Add a truncate helper" in text
+    assert "[demo-1/driver] PLAN (opus, 500 chars): " in text and "[340 more]" in text
+    assert '[demo-1/driver] CALL run_command {"command": "ls"}' in text
+    assert "result (FAILED): README.md" in text
+
+
+def test_full_shows_long_texts_whole(tmp_path: Path) -> None:
+    db = tmp_path / "state.db"
+    store_with_a_run(db)
+    assert "x" * 500 in render(db, full=True)
+
+
+def test_facts_are_shown_as_the_orchestrators(tmp_path: Path) -> None:
+    db = tmp_path / "state.db"
+    store_with_a_run(db)
+    store = StateStore(db, oplog=OpLog("nisse-dev", io.StringIO()))
+    store.transcript("demo-1", "orchestrator").append(Fact(text="PR #3 opened"))
+    store.close()
+    assert "[demo-1/orchestrator] FACT: PR #3 opened" in render(db)
+
+
+def test_the_dump_never_writes_to_the_store(tmp_path: Path) -> None:
+    db = tmp_path / "state.db"
+    store_with_a_run(db)
+    before = db.read_bytes()
+    render(db)
+    assert db.read_bytes() == before
+
+
+def test_a_missing_store_is_reported(tmp_path: Path, capsys) -> None:
+    assert main([str(tmp_path / "nope.db")]) == 1
+    assert "no state store" in capsys.readouterr().err

nisse-core/tests/test_execution.py

@@ -0,0 +1,108 @@
+"""Running commands on a tool VM over SSH (L13), against a throwaway sshd."""
+
+import os
+from pathlib import Path
+
+import pytest
+
+from nisse_core.execution import REDACTED, HostUnavailable, SSHHost, redact
+
+
+def test_exit_codes_and_streams_come_back(ssh_host: SSHHost) -> None:
+    result = ssh_host.execute("echo out; echo err >&2; exit 3")
+    assert (result.exit_code, result.stdout, result.stderr) == (3, "out\n", "err\n")
+    assert not result.ok
+
+
+def test_cwd_and_stdin(ssh_host: SSHHost, tmp_path: Path) -> None:
+    result = ssh_host.execute("pwd; cat", cwd=str(tmp_path), stdin="line one\nline two")
+    assert result.stdout == f"{tmp_path}\nline one\nline two\n"
+
+
+def test_stdin_that_looks_like_a_delimiter_survives(ssh_host: SSHHost) -> None:
+    payload = "NISSE_EOF\nEOF\n'quoted' $notexpanded `nor this`"
+    assert ssh_host.execute("cat", stdin=payload).stdout == payload + "\n"
+
+
+def test_without_stdin_a_command_cannot_read_the_script(ssh_host: SSHHost) -> None:
+    # The script arrives on the same channel; the command must see /dev/null.
+    result = ssh_host.execute("cat; echo done")
+    assert result.stdout == "done\n"
+
+
+def test_a_missing_cwd_fails_before_running(ssh_host: SSHHost, tmp_path: Path) -> None:
+    marker = tmp_path / "ran"
+    result = ssh_host.execute(f"touch {marker}", cwd=str(tmp_path / "absent"))
+    assert result.exit_code == 125
+    assert not marker.exists()
+
+
+def test_a_command_exiting_255_is_not_mistaken_for_ssh(ssh_host: SSHHost) -> None:
+    assert ssh_host.execute("exit 255").exit_code == 254
+
+
+def test_an_unreachable_host_raises(tmp_path: Path) -> None:
+    host = SSHHost("127.0.0.1", key=tmp_path / "k", state_dir=tmp_path / "s", port=1)
+    with pytest.raises(HostUnavailable):
+        host.execute("true")
+
+
+def test_a_timeout_is_reported(ssh_host: SSHHost) -> None:
+    result = ssh_host.execute("sleep 30", timeout_s=1)
+    assert result.timed_out
+
+
+def test_secrets_reach_the_command_and_never_its_output(ssh_host: SSHHost) -> None:
+    value = "s-" + os.urandom(12).hex()
+    result = ssh_host.execute(
+        'echo "$API_KEY"; echo "$API_KEY" >&2', secrets={"API_KEY": value}
+    )
+    assert result.ok
+    assert value not in result.stdout + result.stderr
+    assert result.stdout == f"{REDACTED}\n"
+
+
+def test_secrets_are_never_on_a_command_line(ssh_host: SSHHost, tmp_path: Path) -> None:
+    # Made up here, so no command line in this test run can contain it except
+    # by leaking it.
+    value = "s-" + os.urandom(12).hex()
+    seen = tmp_path / "ps.txt"
+    ssh_host.execute(f"ps -eo args > {seen}", secrets={"API_KEY": value})
+    assert "bash -s" in seen.read_text()  # the listing did include our shell
+    assert value not in seen.read_text()
+
+
+def test_redact_masks_longest_values_first() -> None:
+    assert redact("abc-abcdef", ["abcdef", "abc-abcdef"]) == REDACTED
+    assert redact("short ab stays", ["ab"]) == "short ab stays"  # too short to mask
+
+
+def test_a_changed_host_key_is_its_own_failure_and_only_forgetting_clears_it(
+    ssh_host: SSHHost, tmp_path: Path
+) -> None:
+    import subprocess
+
+    from nisse_core.execution import HostKeyChanged
+
+    host = SSHHost(
+        ssh_host.address,
+        key=ssh_host._key,
+        user=ssh_host.user,
+        port=ssh_host.port,
+        state_dir=tmp_path / "state",
+    )
+    assert host.execute("true").ok  # first contact pins the key
+    host.close()
+    # Someone else's key where the pinned one was: a rebuilt VM, or not it.
+    subprocess.run(
+        ["ssh-keygen", "-q", "-t", "ed25519", "-N", "", "-f", str(tmp_path / "other")],
+        check=True,
+    )
+    other = (tmp_path / "other.pub").read_text().split()
+    known = tmp_path / "state" / "known_hosts"
+    name = host.address if host.port == 22 else f"[{host.address}]:{host.port}"
+    known.write_text(f"{name} {other[0]} {other[1]}\n")
+    with pytest.raises(HostKeyChanged):
+        host.execute("true")
+    host.forget_host_key()
+    assert host.execute("true").ok  # the real key pinned again

nisse-core/tests/test_gitea.py

@@ -0,0 +1,321 @@
+import json
+
+import httpx
+import pytest
+
+from nisse_core.gitea import (
+    PR_PAGE_SIZE,
+    Comment,
+    GiteaClient,
+    GiteaError,
+    pull_request_reconciler,
+)
+from nisse_core.state import Intent
+
+INTENT = Intent(
+    id=1,
+    kind="open_pr",
+    payload={"branch": "milestone-1"},
+    status="pending",
+    created_at="t",
+    resolved_at=None,
+)
+
+
+def client(handler) -> GiteaClient:
+    return GiteaClient(
+        "https://gitea.test/", "tok", transport=httpx.MockTransport(handler)
+    )
+
+
+def pr_json(number: int, head: str, state: str = "open") -> dict:
+    return {
+        "number": number,
+        "state": state,
+        "head": {"ref": head},
+        "html_url": f"https://gitea.test/o/r/pulls/{number}",
+    }
+
+
+def test_open_pull_request_posts_expected_body() -> None:
+    seen: dict = {}
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        seen["url"] = str(request.url)
+        seen["auth"] = request.headers.get("authorization")
+        seen["body"] = json.loads(request.content)
+        return httpx.Response(201, json=pr_json(7, "milestone-1"))
+
+    pr = client(handler).open_pull_request(
+        "o/r", head="milestone-1", base="main", title="T", body="B"
+    )
+    assert seen["url"] == "https://gitea.test/api/v1/repos/o/r/pulls"
+    assert seen["auth"] == "token tok"
+    assert seen["body"] == {
+        "head": "milestone-1",
+        "base": "main",
+        "title": "T",
+        "body": "B",
+    }
+    assert pr.number == 7
+    assert pr.head == "milestone-1"
+
+
+def test_find_pull_request_matches_head_branch() -> None:
+    handler = lambda request: httpx.Response(200, json=[pr_json(3, "other")])
+    assert client(handler).find_pull_request("o/r", "milestone-1") is None
+
+
+def test_find_pull_request_also_finds_closed() -> None:
+    # A crash after opening must not be answered with "nothing open, open
+    # another" — the effect already happened.
+    handler = lambda request: httpx.Response(
+        200, json=[pr_json(4, "milestone-1", "closed")]
+    )
+    found = client(handler).find_pull_request("o/r", "milestone-1")
+    assert found is not None
+    assert found.number == 4
+
+
+def test_find_pull_request_pages_past_the_first_page() -> None:
+    # The match sits on page 2; sampling only the newest page would report
+    # "absent" and licence a duplicate PR.
+    def handler(request: httpx.Request) -> httpx.Response:
+        page = int(request.url.params.get("page", 1))
+        if page == 1:
+            return httpx.Response(
+                200, json=[pr_json(i, f"other-{i}") for i in range(PR_PAGE_SIZE)]
+            )
+        return httpx.Response(200, json=[pr_json(999, "milestone-1", "closed")])
+
+    found = client(handler).find_pull_request("o/r", "milestone-1")
+    assert found is not None
+    assert found.number == 999
+
+
+def test_short_page_ends_the_search_definitively() -> None:
+    calls: list[int] = []
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        calls.append(int(request.url.params.get("page", 1)))
+        return httpx.Response(200, json=[pr_json(1, "other")])
+
+    assert client(handler).find_pull_request("o/r", "milestone-1") is None
+    assert calls == [1]  # one short page proves absence; no further paging
+
+
+def test_unbounded_history_raises_rather_than_guessing() -> None:
+    # Never answer "absent" from an incomplete search: the reconciler turns
+    # this error into "unverifiable" and gates to the human.
+    handler = lambda request: httpx.Response(
+        200, json=[pr_json(i, f"other-{i}") for i in range(PR_PAGE_SIZE)]
+    )
+    with pytest.raises(GiteaError, match="without reaching the end"):
+        client(handler).find_pull_request("o/r", "milestone-1")
+
+
+def test_error_status_raises_gitea_error() -> None:
+    handler = lambda request: httpx.Response(403, text="forbidden")
+    with pytest.raises(GiteaError, match="403"):
+        client(handler).whoami()
+
+
+def test_list_and_post_comments() -> None:
+    def handler(request: httpx.Request) -> httpx.Response:
+        if request.method == "POST":
+            return httpx.Response(
+                201,
+                json={
+                    "id": 9,
+                    "user": {"login": "don"},
+                    "body": "hi",
+                    "created_at": "t",
+                },
+            )
+        return httpx.Response(
+            200,
+            json=[
+                {
+                    "id": 8,
+                    "user": {"login": "don"},
+                    "body": "approve rev-1",
+                    "created_at": "t",
+                }
+            ],
+        )
+
+    c = client(handler)
+    assert c.list_comments("o/r", 1) == [
+        Comment(id=8, author="don", body="approve rev-1", created_at="t")
+    ]
+    assert c.post_comment("o/r", 1, "hi").id == 9
+
+
+def test_list_comments_passes_since() -> None:
+    seen: dict = {}
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        seen["since"] = request.url.params.get("since")
+        return httpx.Response(200, json=[])
+
+    client(handler).list_comments("o/r", 1, since="2026-09-07T00:00:00Z")
+    assert seen["since"] == "2026-09-07T00:00:00Z"
+
+
+# --- reconciler (L8 external truth) --------------------------------------
+
+
+def test_reconciler_reports_existing_pr_as_done() -> None:
+    handler = lambda request: httpx.Response(200, json=[pr_json(5, "milestone-1")])
+    assert pull_request_reconciler(client(handler), "o/r")(INTENT) is True
+
+
+def test_reconciler_reports_missing_pr_as_not_done() -> None:
+    handler = lambda request: httpx.Response(200, json=[])
+    assert pull_request_reconciler(client(handler), "o/r")(INTENT) is False
+
+
+def test_unreachable_gitea_is_unverifiable_not_a_guess() -> None:
+    def handler(request: httpx.Request) -> httpx.Response:
+        raise httpx.ConnectError("refused")
+
+    # None gates to the human: a delayed decision is cheap, a duplicate PR
+    # is not.
+    assert pull_request_reconciler(client(handler), "o/r")(INTENT) is None
+
+
+def test_intent_without_a_branch_is_unverifiable() -> None:
+    handler = lambda request: httpx.Response(200, json=[])
+    blank = Intent(
+        id=2,
+        kind="open_pr",
+        payload={},
+        status="pending",
+        created_at="t",
+        resolved_at=None,
+    )
+    assert pull_request_reconciler(client(handler), "o/r")(blank) is None
+
+
+def test_closing_an_issue_and_a_pr() -> None:
+    seen: list[tuple[str, str, dict]] = []
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        body = json.loads(request.content) if request.content else {}
+        seen.append((request.method, request.url.path, body))
+        if request.method == "POST":
+            return httpx.Response(
+                201, json={"id": 5, "user": {"login": "bot"}, "body": body["body"]}
+            )
+        return httpx.Response(201, json={})
+
+    gitea = client(handler)
+    gitea.post_comment("o/r", 10, "abandoned by don")
+    gitea.close_issue("o/r", 10)
+    gitea.post_comment("o/r", 20, "abandoned by don")
+    gitea.close_pull_request("o/r", 20)
+
+    assert seen == [
+        ("POST", "/api/v1/repos/o/r/issues/10/comments", {"body": "abandoned by don"}),
+        ("PATCH", "/api/v1/repos/o/r/issues/10", {"state": "closed"}),
+        ("POST", "/api/v1/repos/o/r/issues/20/comments", {"body": "abandoned by don"}),
+        ("PATCH", "/api/v1/repos/o/r/pulls/20", {"state": "closed"}),
+    ]
+
+
+def test_assigned_at_is_the_last_assignment_across_pages() -> None:
+    def assigned(login: str, at: str, removed: bool = False) -> dict:
+        return {
+            "type": "assignees",
+            "assignee": {"login": login},
+            "removed_assignee": removed,
+            "created_at": at,
+        }
+
+    filler = {"type": "comment", "created_at": "t0"}
+    pages = {
+        1: [assigned("dev-agent", "t1")] + [filler] * (PR_PAGE_SIZE - 1),
+        2: [
+            assigned("dev-agent", "t2", removed=True),
+            assigned("someone", "t3"),
+            assigned("dev-agent", "t4"),
+        ],
+    }
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        assert request.url.path == "/api/v1/repos/o/r/issues/6/timeline"
+        return httpx.Response(200, json=pages[int(request.url.params["page"])])
+
+    assert client(handler).assigned_at("o/r", 6, "dev-agent") == "t4"
+    assert client(handler).assigned_at("o/r", 6, "nobody") is None
+
+
+def test_approvals_are_the_standing_approving_reviews() -> None:
+    reviews = [
+        {"state": "APPROVED", "user": {"login": "don"}, "commit_id": "abc"},
+        {
+            "state": "APPROVED",
+            "user": {"login": "don"},
+            "commit_id": "old",
+            "stale": True,
+        },
+        {
+            "state": "APPROVED",
+            "user": {"login": "x"},
+            "commit_id": "abc",
+            "dismissed": True,
+        },
+        {"state": "REQUEST_CHANGES", "user": {"login": "don"}, "commit_id": "abc"},
+        {"state": "COMMENT", "user": {"login": "don"}, "commit_id": "abc"},
+    ]
+    handler = lambda request: httpx.Response(200, json=reviews)
+    approvals = client(handler).approvals("o/r", 3)
+    assert [(a.author, a.commit, a.stale) for a in approvals] == [
+        ("don", "abc", False),
+        ("don", "old", True),
+    ]
+
+
+def test_checks_are_the_combined_status_or_none() -> None:
+    def handler(request: httpx.Request) -> httpx.Response:
+        assert request.url.path == "/api/v1/repos/o/r/commits/abc/status"
+        return httpx.Response(200, json=answer)
+
+    answer: dict = {"state": "", "total_count": 0, "statuses": []}
+    assert client(handler).checks("o/r", "abc") == "none"
+    answer = {"state": "pending", "total_count": 1, "statuses": [{}]}
+    assert client(handler).checks("o/r", "abc") == "pending"
+    answer = {"state": "success", "total_count": 2, "statuses": [{}, {}]}
+    assert client(handler).checks("o/r", "abc") == "success"
+
+
+def test_a_pull_request_carries_its_head_commit_and_mergeability() -> None:
+    data = {
+        **pr_json(5, "m-1"),
+        "head": {"ref": "m-1", "sha": "f00"},
+        "mergeable": True,
+    }
+    pr = client(lambda request: httpx.Response(200, json=data)).pull_request("o/r", 5)
+    assert (pr.head_sha, pr.mergeable) == ("f00", True)
+
+
+def test_a_check_branch_is_made_through_the_api() -> None:
+    seen: list = []
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        seen.append((request.method, request.url.path, request.content))
+        return httpx.Response(201, json={})
+
+    gitea = client(handler)
+    gitea.branch_with_file(
+        "o/r", base="main", branch="chk", path="f", text="x", message="m"
+    )
+    gitea.delete_branch("o/r", "chk")
+    body = json.loads(seen[0][2])
+    assert (seen[0][0], seen[0][1]) == ("POST", "/api/v1/repos/o/r/contents/f")
+    assert (body["branch"], body["new_branch"], body["content"]) == (
+        "main",
+        "chk",
+        "eA==",
+    )
+    assert seen[1][:2] == ("DELETE", "/api/v1/repos/o/r/branches/chk")

nisse-core/tests/test_independence.py

@@ -0,0 +1,25 @@
+"""nisse-core is what every agent is built on (Part L): it never imports an
+agent."""
+
+import ast
+from pathlib import Path
+
+import nisse_core
+
+AGENTS = ("nisse_dev", "nisse_ops")
+
+
+def test_core_imports_no_agent() -> None:
+    root = Path(nisse_core.__file__).parent
+    found = []
+    for path in sorted(root.rglob("*.py")):
+        for node in ast.walk(ast.parse(path.read_text())):
+            if isinstance(node, ast.Import):
+                names = [a.name for a in node.names]
+            elif isinstance(node, ast.ImportFrom):
+                names = [node.module or ""]
+            else:
+                continue
+            found += [f"{path.name}: {n}" for n in names if n.split(".")[0] in AGENTS]
+    assert len(list(root.rglob("*.py"))) > 10  # the modules really are here
+    assert found == []

nisse-core/tests/test_llm.py

@@ -0,0 +1,516 @@
+import json
+from pathlib import Path
+from typing import Any
+
+import httpx
+import pytest
+
+from nisse_core.canonical import (
+    AssistantText,
+    StoredEvent,
+    ToolCall,
+    UserMessage,
+)
+from nisse_core.envfile import load_env_file
+from nisse_core.llm import (
+    DEFAULT_MAX_OUTPUT_TOKENS,
+    OUTAGE_BACKOFF_S,
+    RATE_LIMIT_RETRY_S,
+    AnthropicPassthroughClient,
+    BudgetExhausted,
+    LiteLLMClient,
+    LLMError,
+    RoleMap,
+    TruncatedResponse,
+    UpstreamUnavailable,
+)
+
+EVENTS = [StoredEvent(id=1, ts="t", event=UserMessage(text="hi", source="m:1"))]
+
+
+def make_client(handler, sleep=lambda _s: None) -> LiteLLMClient:
+    return LiteLLMClient(
+        "http://litellm.test/v1",
+        "sk-test",
+        transport=httpx.MockTransport(handler),
+        sleep=sleep,  # never wait in unit tests
+    )
+
+
+def completion(message: dict) -> httpx.Response:
+    return httpx.Response(
+        200, json={"model": "gpt-test-1", "choices": [{"message": message}]}
+    )
+
+
+def test_request_shape_and_auth() -> None:
+    seen: dict = {}
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        seen["url"] = str(request.url)
+        seen["auth"] = request.headers.get("authorization")
+        seen["body"] = json.loads(request.content)
+        return completion({"content": "hello"})
+
+    make_client(handler).complete(EVENTS, tools=None, model="local/qwen", system="s")
+    assert seen["url"] == "http://litellm.test/v1/chat/completions"
+    assert seen["auth"] == "Bearer sk-test"
+    assert seen["body"]["model"] == "local/qwen"
+    assert seen["body"]["messages"][0] == {"role": "system", "content": "s"}
+    assert seen["body"]["messages"][1] == {"role": "user", "content": "hi"}
+
+
+def test_assistant_text_parsed_with_producing_model() -> None:
+    client = make_client(lambda request: completion({"content": "hello"}))
+    [event] = client.complete(EVENTS, tools=None, model="local/qwen")
+    assert event == AssistantText(text="hello", model="gpt-test-1")
+
+
+def test_tool_calls_get_locally_minted_canonical_ids() -> None:
+    message = {
+        "content": None,
+        "tool_calls": [
+            {
+                "id": "call_provider_abc",
+                "type": "function",
+                "function": {
+                    "name": "run_tests",
+                    "arguments": '{"path": "tests/"}',
+                },
+            }
+        ],
+    }
+    client = make_client(lambda request: completion(message))
+    [event] = client.complete(EVENTS, tools=None, model="local/qwen")
+    assert isinstance(event, ToolCall)
+    assert event.tool == "run_tests"
+    assert event.arguments == {"path": "tests/"}
+    assert event.call_id != "call_provider_abc"  # provider ID never a join key
+    assert event.call_id.startswith("call-")
+
+
+def test_truncated_tool_arguments_raise_llm_error() -> None:
+    message = {
+        "content": None,
+        "tool_calls": [
+            {
+                "id": "c",
+                "type": "function",
+                "function": {"name": "submit_plan", "arguments": '{"content": "'},
+            }
+        ],
+    }
+    client = make_client(lambda request: completion(message))
+    with pytest.raises(LLMError, match="non-JSON arguments"):
+        client.complete(EVENTS, tools=None, model="local/qwen")
+
+
+def test_non_object_tool_arguments_raise_llm_error() -> None:
+    message = {
+        "content": None,
+        "tool_calls": [
+            {
+                "id": "c",
+                "type": "function",
+                "function": {"name": "submit_plan", "arguments": '"just a string"'},
+            }
+        ],
+    }
+    client = make_client(lambda request: completion(message))
+    with pytest.raises(LLMError, match="expected object"):
+        client.complete(EVENTS, tools=None, model="local/qwen")
+
+
+def test_http_error_status_raises_llm_error() -> None:
+    client = make_client(lambda request: httpx.Response(500, text="boom"))
+    with pytest.raises(LLMError, match="500"):
+        client.complete(EVENTS, tools=None, model="local/qwen")
+
+
+def test_rate_limit_is_retried_once_then_surfaced() -> None:
+    # A 429 is transient. Losing a long delegation to one costs far more
+    # than waiting out the window, but retrying forever would hide a real
+    # ceiling, so it is retried exactly once.
+    attempts: list[int] = []
+    waited: list[float] = []
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        attempts.append(1)
+        return httpx.Response(429, text="rate limited")
+
+    client = make_client(handler, sleep=waited.append)
+    with pytest.raises(LLMError, match="429"):
+        client.complete(EVENTS, tools=None, model="local/qwen")
+    assert len(attempts) == 2
+    assert waited == [RATE_LIMIT_RETRY_S]
+    # Reported even though the call failed, so a caller can leave the wait
+    # out of how long the model took.
+    assert client.last_rate_limit_wait_s == RATE_LIMIT_RETRY_S
+
+
+def test_rate_limit_that_clears_succeeds() -> None:
+    responses = [httpx.Response(429, text="slow down"), completion({"content": "hi"})]
+    client = make_client(lambda request: responses.pop(0), sleep=lambda _s: None)
+    [event] = client.complete(EVENTS, tools=None, model="local/qwen")
+    assert isinstance(event, AssistantText)
+    assert client.last_rate_limit_wait_s == RATE_LIMIT_RETRY_S
+
+
+def test_the_rate_limit_wait_is_per_call() -> None:
+    responses = [
+        httpx.Response(429, text="slow down"),
+        completion({"content": "one"}),
+        completion({"content": "two"}),
+    ]
+    client = make_client(lambda request: responses.pop(0), sleep=lambda _s: None)
+    client.complete(EVENTS, tools=None, model="local/qwen")
+    client.complete(EVENTS, tools=None, model="local/qwen")
+    assert client.last_rate_limit_wait_s == 0.0
+
+
+def test_connection_failure_raises_llm_error() -> None:
+    def handler(request: httpx.Request) -> httpx.Response:
+        raise httpx.ConnectError("refused")
+
+    with pytest.raises(LLMError, match="request failed"):
+        make_client(handler).complete(EVENTS, tools=None, model="local/qwen")
+
+
+def test_malformed_response_raises_llm_error() -> None:
+    client = make_client(lambda request: httpx.Response(200, json={"oops": True}))
+    with pytest.raises(LLMError, match="malformed"):
+        client.complete(EVENTS, tools=None, model="local/qwen")
+
+
+ROLES = RoleMap(
+    driver="local/qwen",
+    helper="anthropic/claude-sonnet-5",
+    expert="anthropic/claude-opus-5",
+)
+
+
+def test_role_map_resolves_and_pins() -> None:
+    roles = ROLES
+    assert roles.resolve("driver") == "local/qwen"
+    assert roles.pinned() == {
+        "driver": "local/qwen",
+        "helper": "anthropic/claude-sonnet-5",
+        "expert": "anthropic/claude-opus-5",
+        # Unset, the classifier falls back to the expert's model (L9).
+        "classifier": "anthropic/claude-opus-5",
+    }
+
+
+def test_a_pinned_map_from_before_the_driver_reads_its_own_roles() -> None:
+    # Pinned before M2.8: the old loop's roles are there, and are ignored.
+    old = {"planner": "p", "executor": "e", "driver": "d", "helper": "h"}
+    roles = RoleMap.from_pinned(old, ROLES)
+    assert (roles.driver, roles.helper, roles.expert) == ("d", "h", ROLES.expert)
+
+
+def test_unknown_role_raises() -> None:
+    with pytest.raises(LLMError, match="unknown role"):
+        ROLES.resolve("boss")
+
+
+def test_load_env_file(tmp_path: Path) -> None:
+    envfile = tmp_path / ".env"
+    envfile.write_text(
+        "# comment\n\nNISSE_LLM_TOKEN=sk-abc\nNISSE_LLM_BASE_URL = http://x \n"
+    )
+    env: dict[str, str] = {"NISSE_LLM_TOKEN": "sk-already-set"}
+    load_env_file(envfile, env)
+    assert env["NISSE_LLM_TOKEN"] == "sk-already-set"  # real env wins
+    assert env["NISSE_LLM_BASE_URL"] == "http://x"
+
+
+def test_load_env_file_missing_is_noop(tmp_path: Path) -> None:
+    env: dict[str, str] = {}
+    load_env_file(tmp_path / "nope.env", env)
+    assert env == {}
+
+
+# --- truncation (silent until measured against Opus) ---------------------
+
+
+def test_openai_route_raises_on_truncation() -> None:
+    # finish_reason "length" means the answer was cut off; parsing it as
+    # complete yields a tool call with required arguments simply absent.
+    truncated = httpx.Response(
+        200,
+        json={
+            "model": "gpt-test-1",
+            "choices": [
+                {
+                    "finish_reason": "length",
+                    "message": {
+                        "content": None,
+                        "tool_calls": [
+                            {
+                                "id": "c",
+                                "type": "function",
+                                "function": {
+                                    "name": "submit_plan",
+                                    "arguments": '{"type": "arch"}',
+                                },
+                            }
+                        ],
+                    },
+                }
+            ],
+        },
+    )
+    client = make_client(lambda request: truncated)
+    with pytest.raises(TruncatedResponse, match="output limit"):
+        client.complete(EVENTS, tools=None, model="gpt-6-astra")
+
+
+def test_anthropic_route_raises_on_truncation() -> None:
+    # Observed for real: Opus spent 4,064 output tokens against a 4,096
+    # ceiling, and a longer plan came back as submit_plan with no content.
+    def handler(request: httpx.Request) -> httpx.Response:
+        return httpx.Response(
+            200,
+            json={
+                "model": "claude-opus-5",
+                "stop_reason": "max_tokens",
+                "content": [
+                    {
+                        "type": "tool_use",
+                        "id": "t",
+                        "name": "submit_plan",
+                        "input": {"type": "arch"},
+                    }
+                ],
+            },
+        )
+
+    client = AnthropicPassthroughClient(
+        "https://litellm.test",
+        "tok",
+        transport=httpx.MockTransport(handler),
+        sleep=lambda _s: None,
+    )
+    with pytest.raises(TruncatedResponse, match="output limit"):
+        client.complete(EVENTS, tools=None, model="claude-opus-5")
+
+
+def test_a_complete_answer_is_not_mistaken_for_truncation() -> None:
+    def handler(request: httpx.Request) -> httpx.Response:
+        return httpx.Response(
+            200,
+            json={
+                "model": "claude-opus-5",
+                "stop_reason": "tool_use",
+                "content": [
+                    {
+                        "type": "tool_use",
+                        "id": "t",
+                        "name": "submit_plan",
+                        "input": {"type": "arch", "content": "a real plan"},
+                    }
+                ],
+            },
+        )
+
+    client = AnthropicPassthroughClient(
+        "https://litellm.test", "tok", transport=httpx.MockTransport(handler)
+    )
+    [call] = client.complete(EVENTS, tools=None, model="claude-opus-5")
+    assert call.arguments["content"] == "a real plan"
+
+
+def test_output_ceiling_is_generous_by_default() -> None:
+    # Output is billed by what is produced, so a low ceiling buys nothing and
+    # costs silent truncation. The configured value is tested in test_config.
+    assert DEFAULT_MAX_OUTPUT_TOKENS >= 8192
+    assert LiteLLMClient("http://x", "t")._max_tokens == DEFAULT_MAX_OUTPUT_TOKENS
+
+
+def test_truncation_is_an_llm_error_so_callers_already_handle_it() -> None:
+    # The planner retries, the executor and delegation gate. None of them
+    # need to learn a new exception type.
+    assert issubclass(TruncatedResponse, LLMError)
+
+
+def test_a_5xx_is_upstream_trouble() -> None:
+    client = make_client(lambda request: httpx.Response(503, text="bad gateway"))
+    with pytest.raises(UpstreamUnavailable, match="503"):
+        client.complete(EVENTS, tools=None, model="local-fast")
+
+
+def test_a_connection_failure_is_upstream_trouble() -> None:
+    def handler(request: httpx.Request) -> httpx.Response:
+        raise httpx.ConnectError("refused")
+
+    with pytest.raises(UpstreamUnavailable, match="request failed"):
+        make_client(handler).complete(EVENTS, tools=None, model="local-fast")
+
+
+def test_a_persistent_rate_limit_is_upstream_trouble() -> None:
+    # It cleared for neither attempt, so the caller should keep its work
+    # rather than discard it.
+    client = make_client(lambda request: httpx.Response(429, text="slow down"))
+    with pytest.raises(UpstreamUnavailable):
+        client.complete(EVENTS, tools=None, model="local-fast")
+
+
+def test_a_bad_request_is_not_upstream_trouble() -> None:
+    client = make_client(lambda request: httpx.Response(400, text="bad schema"))
+    with pytest.raises(LLMError) as excinfo:
+        client.complete(EVENTS, tools=None, model="local-fast")
+    assert not isinstance(excinfo.value, UpstreamUnavailable)
+
+
+def test_budget_refusal_is_a_kind_of_upstream_trouble() -> None:
+    # So one catch covers the family, but it can still be told apart where
+    # the difference matters — it will not clear within the hour.
+    assert issubclass(BudgetExhausted, UpstreamUnavailable)
+
+
+# --- tool_choice (L9: constrained decoding) --------------------------------
+
+TOOLS = [
+    {
+        "type": "function",
+        "function": {
+            "name": "finish",
+            "description": "done",
+            "parameters": {"type": "object", "properties": {}},
+        },
+    }
+]
+
+
+def capture_bodies() -> tuple[list[dict], Any]:
+    bodies: list[dict] = []
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        bodies.append(json.loads(request.content))
+        return completion({"role": "assistant", "content": "ok"})
+
+    return bodies, handler
+
+
+def test_required_is_sent_only_when_asked_and_tools_are_offered() -> None:
+    bodies, handler = capture_bodies()
+    client = make_client(handler)
+    client.complete(EVENTS, tools=TOOLS, model="m", require_tool_call=True)
+    client.complete(EVENTS, tools=TOOLS, model="m")
+    client.complete(EVENTS, tools=None, model="m", require_tool_call=True)
+    assert bodies[0]["tool_choice"] == "required"
+    assert "tool_choice" not in bodies[1]  # the server's default, auto
+    assert "tool_choice" not in bodies[2]  # nothing to choose from
+
+
+def test_the_passthrough_asks_in_anthropics_terms() -> None:
+    bodies: list[dict] = []
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        bodies.append(json.loads(request.content))
+        return httpx.Response(
+            200,
+            json={
+                "model": "claude-haiku-4-5",
+                "stop_reason": "tool_use",
+                "content": [
+                    {"type": "tool_use", "id": "t", "name": "finish", "input": {}}
+                ],
+            },
+        )
+
+    client = AnthropicPassthroughClient(
+        "https://litellm.test", "tok", transport=httpx.MockTransport(handler)
+    )
+    client.complete(
+        EVENTS, tools=TOOLS, model="claude-haiku-4-5", require_tool_call=True
+    )
+    assert bodies[0]["tool_choice"] == {"type": "any"}
+
+
+def test_the_answering_model_is_recorded() -> None:
+    # A proxy fallback answers under another model's name; this is how a
+    # caller can tell.
+    client = make_client(
+        lambda request: completion({"role": "assistant", "content": "hi"})
+    )
+    client.complete(EVENTS, tools=None, model="local-fast")
+    assert client.last_model == "gpt-test-1"
+
+
+# --- outages delay, never fail (L5) --------------------------------------------
+
+
+def test_a_gateway_error_is_waited_out_with_backoff() -> None:
+    responses = [
+        httpx.Response(502, text="bad gateway"),
+        httpx.Response(503, text="unavailable"),
+        completion({"content": "back"}),
+    ]
+    waited: list[float] = []
+    client = make_client(lambda request: responses.pop(0), sleep=waited.append)
+    [event] = client.complete(EVENTS, tools=None, model="local/qwen")
+    assert isinstance(event, AssistantText) and event.text == "back"
+    assert waited == list(OUTAGE_BACKOFF_S[:2])
+    # Waiting isn't the model's time.
+    assert client.last_rate_limit_wait_s == sum(OUTAGE_BACKOFF_S[:2])
+
+
+def test_an_outage_that_outlasts_the_backoff_is_surfaced() -> None:
+    attempts: list[int] = []
+    waited: list[float] = []
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        attempts.append(1)
+        return httpx.Response(502, text="bad gateway")
+
+    client = make_client(handler, sleep=waited.append)
+    with pytest.raises(UpstreamUnavailable, match="502"):
+        client.complete(EVENTS, tools=None, model="local/qwen")
+    assert len(attempts) == len(OUTAGE_BACKOFF_S) + 1
+    assert waited == list(OUTAGE_BACKOFF_S)
+
+
+def test_a_dropped_connection_is_retried() -> None:
+    calls: list[int] = []
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        calls.append(1)
+        if len(calls) == 1:
+            raise httpx.ConnectError("refused")
+        return completion({"content": "ok"})
+
+    [event] = make_client(handler).complete(EVENTS, tools=None, model="local/qwen")
+    assert isinstance(event, AssistantText) and len(calls) == 2
+
+
+def test_a_500_or_a_timeout_is_not_retried() -> None:
+    # A 500 may be the request's fault; a timeout already had its full
+    # request_timeout_s.
+    for failure in (
+        lambda request: httpx.Response(500, text="boom"),
+        lambda request: (_ for _ in ()).throw(httpx.ReadTimeout("slow")),
+    ):
+        calls: list[int] = []
+
+        def handler(
+            request: httpx.Request, failure=failure, calls=calls
+        ) -> httpx.Response:
+            calls.append(1)
+            return failure(request)
+
+        with pytest.raises(LLMError):
+            make_client(handler).complete(EVENTS, tools=None, model="local/qwen")
+        assert len(calls) == 1
+
+
+def test_a_budget_refusal_is_not_waited_on() -> None:
+    calls: list[int] = []
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        calls.append(1)
+        return httpx.Response(503, text="Budget has been exceeded! Current cost: 5.1")
+
+    with pytest.raises(BudgetExhausted):
+        make_client(handler).complete(EVENTS, tools=None, model="local/qwen")
+    assert len(calls) == 1

nisse-core/tests/test_matrix.py

@@ -0,0 +1,361 @@
+import json
+import threading
+
+import httpx
+import pytest
+
+from nisse_core.matrix import MatrixConfig, MatrixTransport, TransportError
+from nisse_core.transport import InboundMessage
+
+CONFIG = MatrixConfig(
+    homeserver="https://matrix.test",
+    user_id="@nisse-dev:test",
+    token="syt_secret",
+    room_id="!room:test",
+)
+
+
+def sync_body(events: list[dict], next_batch: str) -> dict:
+    return {
+        "next_batch": next_batch,
+        "rooms": {"join": {"!room:test": {"timeline": {"events": events}}}},
+    }
+
+
+def message_event(sender: str, body: str) -> dict:
+    return {
+        "type": "m.room.message",
+        "sender": sender,
+        "content": {"msgtype": "m.text", "body": body},
+    }
+
+
+def test_send_message_puts_event_with_auth_and_unique_txn() -> None:
+    seen: list[httpx.Request] = []
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        seen.append(request)
+        return httpx.Response(200, json={"event_id": "$e"})
+
+    transport = MatrixTransport(CONFIG, http_transport=httpx.MockTransport(handler))
+    transport.send_message("hello")
+    transport.send_approval_request("plan rev-1")
+
+    first, second = seen
+    assert first.headers["authorization"] == "Bearer syt_secret"
+    assert "/rooms/!room:test/send/m.room.message/" in str(first.url)
+    assert json.loads(first.content) == {"msgtype": "m.text", "body": "hello"}
+    assert str(first.url) != str(second.url)  # txn ids must be unique
+
+
+def test_send_failure_raises_transport_error() -> None:
+    def handler(request: httpx.Request) -> httpx.Response:
+        return httpx.Response(403, text="forbidden")
+
+    transport = MatrixTransport(CONFIG, http_transport=httpx.MockTransport(handler))
+    with pytest.raises(TransportError, match="403"):
+        transport.send_message("hello")
+
+
+class SyncScript:
+    """Serves scripted /sync responses, then endless empty ones."""
+
+    def __init__(self, responses: list[dict]) -> None:
+        self._responses = responses
+        self._i = 0
+        self.batches_seen: list[str | None] = []
+
+    def __call__(self, request: httpx.Request) -> httpx.Response:
+        assert request.url.path.endswith("/sync")
+        self.batches_seen.append(request.url.params.get("since"))
+        if self._i < len(self._responses):
+            body = self._responses[self._i]
+            self._i += 1
+            return httpx.Response(200, json=body)
+        return httpx.Response(200, json=sync_body([], f"b{self._i}"))
+
+
+def collect_messages(
+    responses: list[dict], expected: int, timeout_s: float = 5.0
+) -> list[InboundMessage]:
+    received: list[InboundMessage] = []
+    done = threading.Event()
+
+    def on_message(msg: InboundMessage) -> None:
+        received.append(msg)
+        if len(received) >= expected:
+            done.set()
+
+    transport = MatrixTransport(
+        CONFIG,
+        poll_timeout_ms=10,
+        http_transport=httpx.MockTransport(SyncScript(responses)),
+    )
+    transport.on_message(on_message)
+    transport.start()
+    done.wait(timeout_s)
+    transport.close()
+    return received
+
+
+def test_listener_delivers_operator_messages() -> None:
+    responses = [
+        sync_body([message_event("@don:test", "backlog: never delivered")], "b1"),
+        sync_body([message_event("@don:test", "approve")], "b2"),
+    ]
+    received = collect_messages(responses, expected=1)
+    assert received == [InboundMessage(sender="@don:test", text="approve")]
+
+
+def test_own_and_non_message_events_are_filtered() -> None:
+    responses = [
+        sync_body([], "b1"),
+        sync_body(
+            [
+                message_event("@nisse-dev:test", "my own echo"),
+                {"type": "m.room.member", "sender": "@don:test", "content": {}},
+                message_event("@don:test", "status"),
+            ],
+            "b2",
+        ),
+    ]
+    received = collect_messages(responses, expected=1)
+    assert received == [InboundMessage(sender="@don:test", text="status")]
+
+
+def test_first_sync_learns_position_without_delivering() -> None:
+    script = SyncScript([sync_body([message_event("@don:test", "old")], "b1")])
+    transport = MatrixTransport(
+        CONFIG, poll_timeout_ms=10, http_transport=httpx.MockTransport(script)
+    )
+    got = threading.Event()
+    transport.on_message(lambda msg: got.set())
+    transport.start()
+    assert not got.wait(0.3)
+    transport.close()
+    assert script.batches_seen[0] is None  # initial sync carries no since
+    assert "b1" in script.batches_seen  # later polls resume from its batch
+
+
+def test_start_without_handler_raises() -> None:
+    transport = MatrixTransport(
+        CONFIG, http_transport=httpx.MockTransport(lambda r: httpx.Response(200))
+    )
+    with pytest.raises(TransportError, match="no message handler"):
+        transport.start()
+
+
+# --- leading mentions --------------------------------------------------------
+
+
+def own_member_state(name: str) -> dict:
+    """The first sync's room state, which carries the agent's display name."""
+    body = sync_body([], "b1")
+    body["rooms"]["join"]["!room:test"]["state"] = {
+        "events": [
+            {
+                "type": "m.room.member",
+                "state_key": "@nisse-dev:test",
+                "sender": "@nisse-dev:test",
+                "content": {"membership": "join", "displayname": name},
+            }
+        ]
+    }
+    return body
+
+
+def pill_event(sender: str, user_id: str, label: str, rest: str) -> dict:
+    """A message as Element sends one that starts by mentioning user_id."""
+    return {
+        "type": "m.room.message",
+        "sender": sender,
+        "content": {
+            "msgtype": "m.text",
+            "body": f"{label}: {rest}",
+            "format": "org.matrix.custom.html",
+            "formatted_body": (
+                f'<a href="https://matrix.to/#/{user_id}">{label}</a>: {rest}'
+            ),
+        },
+    }
+
+
+def test_a_leading_pill_for_the_agent_is_removed_before_matching() -> None:
+    responses = [
+        sync_body([], "b1"),
+        sync_body(
+            [pill_event("@don:test", "@nisse-dev:test", "Nisse Dev", "approve")],
+            "b2",
+        ),
+    ]
+    [msg] = collect_messages(responses, expected=1)
+    assert msg.sender == "@don:test"  # from the event, never from the text
+    assert msg.text == "approve"
+    assert msg.original == "Nisse Dev: approve"
+
+
+def test_the_agents_display_name_typed_in_front_is_removed() -> None:
+    responses = [
+        own_member_state("Nisse Dev"),
+        sync_body([message_event("@don:test", "Nisse Dev: re-plan")], "b2"),
+    ]
+    [msg] = collect_messages(responses, expected=1)
+    assert msg.text == "re-plan"
+
+
+def test_the_agents_user_id_typed_in_front_is_removed() -> None:
+    responses = [
+        sync_body([], "b1"),
+        sync_body([message_event("@don:test", "@nisse-dev:test: status")], "b2"),
+    ]
+    [msg] = collect_messages(responses, expected=1)
+    assert msg.text == "status"
+
+
+def test_messages_addressed_to_someone_else_are_skipped() -> None:
+    responses = [
+        sync_body([], "b1"),
+        sync_body(
+            [
+                pill_event("@don:test", "@nisse-ops:test", "Nisse Ops", "approve"),
+                message_event("@don:test", "@alice:test: looks good"),
+                message_event("@don:test", "status"),
+            ],
+            "b2",
+        ),
+    ]
+    received = collect_messages(responses, expected=1)
+    assert received == [InboundMessage(sender="@don:test", text="status")]
+
+
+def test_a_mention_later_in_the_message_is_left_alone() -> None:
+    responses = [
+        sync_body([], "b1"),
+        sync_body([message_event("@don:test", "approve, cc @alice:test")], "b2"),
+    ]
+    received = collect_messages(responses, expected=1)
+    assert received == [
+        InboundMessage(sender="@don:test", text="approve, cc @alice:test")
+    ]
+
+
+# --- threads (L5): a work item's conversation is a thread --------------------
+
+
+def threaded_transport() -> tuple[MatrixTransport, list[dict], list[InboundMessage]]:
+    sent: list[dict] = []
+    ids = iter(f"$sent{n}" for n in range(1, 100))
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        sent.append(json.loads(request.content))
+        return httpx.Response(200, json={"event_id": next(ids)})
+
+    transport = MatrixTransport(CONFIG, http_transport=httpx.MockTransport(handler))
+    received: list[InboundMessage] = []
+    transport.on_message(received.append)
+    return transport, sent, received
+
+
+def event(event_id: str, body: str, relates: dict | None = None) -> dict:
+    content: dict = {"msgtype": "m.text", "body": body}
+    if relates is not None:
+        content["m.relates_to"] = relates
+    return {
+        "type": "m.room.message",
+        "sender": "@don:test",
+        "event_id": event_id,
+        "content": content,
+    }
+
+
+def deliver(transport: MatrixTransport, *events: dict) -> None:
+    transport._deliver(sync_body(list(events), "b"))
+
+
+def test_post_returns_the_event_id_and_threads_a_reply() -> None:
+    transport, sent, _ = threaded_transport()
+    root = transport.post("queued sandbox#14")
+    transport.post("As I understand it: ...", thread=root)
+
+    assert root == "$sent1"
+    assert "m.relates_to" not in sent[0]
+    assert sent[1]["m.relates_to"] == {
+        "rel_type": "m.thread",
+        "event_id": "$sent1",
+        # Clients without threads show it as a reply to the root.
+        "is_falling_back": True,
+        "m.in_reply_to": {"event_id": "$sent1"},
+    }
+
+
+def test_a_message_in_a_thread_carries_its_root() -> None:
+    transport, _, received = threaded_transport()
+    deliver(
+        transport,
+        event("$top", "status"),
+        event("$in", "go", {"rel_type": "m.thread", "event_id": "$root"}),
+    )
+    assert [(m.text, m.event_id, m.thread) for m in received] == [
+        ("status", "$top", None),
+        ("go", "$in", "$root"),
+    ]
+
+
+def test_a_plain_reply_to_a_threaded_message_joins_its_thread() -> None:
+    # Clients without thread support send an ordinary reply, quoting.
+    transport, _, received = threaded_transport()
+    root = transport.post("queued sandbox#14")
+    in_thread = transport.post("As I understand it: ...", thread=root)
+    deliver(
+        transport,
+        event(
+            "$r1",
+            "> <@nisse-dev:test> As I understand it: ...\n\nyes, go",
+            {"m.in_reply_to": {"event_id": in_thread}},
+        ),
+        event(
+            "$r2",
+            "> <@nisse-dev:test> queued sandbox#14\n\nthanks",
+            {"m.in_reply_to": {"event_id": root}},
+        ),
+    )
+    assert [(m.text, m.thread) for m in received] == [
+        ("yes, go", root),  # the quote is not what was said
+        ("thanks", root),  # a reply to the root joins its thread
+    ]
+
+
+def test_a_reply_to_an_unknown_message_stays_at_top_level() -> None:
+    transport, _, received = threaded_transport()
+    deliver(
+        transport,
+        event("$r", "> <@x:test> old\n\nhm", {"m.in_reply_to": {"event_id": "$old"}}),
+    )
+    assert [(m.text, m.thread) for m in received] == [("hm", None)]
+
+
+def test_a_reply_to_the_operators_own_threaded_message_joins_the_thread() -> None:
+    transport, _, received = threaded_transport()
+    deliver(
+        transport,
+        event("$t", "first", {"rel_type": "m.thread", "event_id": "$root"}),
+        event(
+            "$r",
+            "> <@don:test> first\n\nand also",
+            {"m.in_reply_to": {"event_id": "$t"}},
+        ),
+    )
+    assert [m.thread for m in received] == ["$root", "$root"]
+
+
+def test_in_memory_transport_plays_threads() -> None:
+    from nisse_core.transport import InMemoryTransport
+
+    transport = InMemoryTransport()
+    seen: list[InboundMessage] = []
+    transport.on_message(seen.append)
+    root = transport.post("queued #14")
+    transport.post("restated", thread=root)
+    reply = transport.receive("@don:test", "go", thread=root)
+    assert transport.outbound[-1].thread == root
+    assert (seen[0].thread, seen[0].event_id) == (root, reply)

nisse-core/tests/test_memory.py

@@ -0,0 +1,97 @@
+"""Agent memory (L6.4, M5.2): one note per fact in the agent's folder, an
+index the harness keeps, and a snapshot of it for the transcript."""
+
+from pathlib import Path
+
+import pytest
+
+from nisse_core.memory import Memory
+from nisse_core.vault import Vault, VaultError, split_frontmatter
+
+FOLDER = "memory/nisse-dev"
+
+
+@pytest.fixture
+def memory(tmp_path: Path) -> Memory:
+    return Memory(
+        Vault(tmp_path / "v", {FOLDER: "rw", "notes": "ro"}, agent="nisse-dev"), FOLDER
+    )
+
+
+def note(memory: Memory, name: str) -> tuple[dict, str]:
+    path = memory._vault.root / FOLDER / f"{name}.md"
+    return split_frontmatter(path.read_text())
+
+
+def test_a_fact_is_a_note_with_its_description_and_provenance(memory: Memory) -> None:
+    replaced = memory.remember(
+        "prefers-small-prs",
+        "The operator prefers  small PRs",
+        "Don said PRs over 400 lines are hard to review.\n",
+        provenance={"source": "work item x"},
+    )
+    assert not replaced
+    front, body = note(memory, "prefers-small-prs")
+    assert front["description"] == "The operator prefers small PRs"
+    assert front["provenance"]["source"] == "work item x"
+    assert body == "Don said PRs over 400 lines are hard to review.\n"
+
+
+def test_the_same_name_replaces_and_the_index_follows(memory: Memory) -> None:
+    memory.remember("b-fact", "Bee", "b", provenance={})
+    memory.remember("a-fact", "Ay", "a", provenance={})
+    assert memory.remember("a-fact", "Ay, again", "a2", provenance={})
+    index = memory.snapshot().text
+    assert index.endswith(
+        "- [[memory/nisse-dev/a-fact]] — Ay, again\n- [[memory/nisse-dev/b-fact]] — Bee"
+    )
+    memory.forget("a-fact")
+    assert "a-fact" not in memory.snapshot().text
+    with pytest.raises(VaultError, match="no such note"):
+        memory.forget("a-fact")
+
+
+@pytest.mark.parametrize("name", ["index", "Caps", "../x", "a/b", "", "-x"])
+def test_names_are_plain(memory: Memory, name: str) -> None:
+    with pytest.raises(VaultError, match="a note's name"):
+        memory.remember(name, "d", "f", provenance={})
+
+
+def test_a_note_needs_a_description_and_a_fact(memory: Memory) -> None:
+    with pytest.raises(VaultError, match="description and the fact"):
+        memory.remember("x", "  ", "f", provenance={})
+
+
+def test_the_snapshot_is_hashed_and_made_if_missing(memory: Memory) -> None:
+    first = memory.snapshot()
+    assert first.paths == (f"{FOLDER}/index.md",)
+    assert "Nothing yet." in first.text
+    assert memory.snapshot().sha == first.sha  # same index, same hash
+    memory.remember("x", "Ex", "f", provenance={})
+    assert memory.snapshot().sha != first.sha
+
+
+def test_a_projects_notes_have_their_own_folder_and_index(memory: Memory) -> None:
+    memory.remember("general-one", "General", "g", provenance={})
+    memory.remember("deploy", "Deploys by tag", "d", provenance={}, project="sandbox")
+    memory.remember("other", "Other's", "o", provenance={}, project="other")
+    root = memory._vault.root / FOLDER
+    assert (root / "projects" / "sandbox" / "deploy.md").exists()
+    top = memory.snapshot().text
+    assert "- [[memory/nisse-dev/general-one]] — General" in top
+    assert "deploy" not in top.split("Projects:")[0]
+    assert "- [[memory/nisse-dev/projects/sandbox/index]] — notes about sandbox" in top
+    seen = memory.snapshot("sandbox")
+    assert seen.paths == (
+        f"{FOLDER}/index.md",
+        f"{FOLDER}/projects/sandbox/index.md",
+    )
+    assert (
+        "- [[memory/nisse-dev/projects/sandbox/deploy]] — Deploys by tag" in seen.text
+    )
+    assert "projects/other/other]]" not in seen.text
+    memory.forget("deploy", "sandbox")
+    assert memory.snapshot("sandbox").paths == (f"{FOLDER}/index.md",)
+    assert "sandbox/index" not in memory.snapshot().text
+    with pytest.raises(VaultError, match="not a project's name"):
+        memory.remember("x", "d", "f", provenance={}, project="../x")

nisse-core/tests/test_oplog.py

@@ -0,0 +1,183 @@
+"""State-change log conventions (architecture L12): every durable mutation emits one
+JSON-lines event with the fixed envelope, from the store's own write path."""
+
+import io
+import json
+import sqlite3
+from pathlib import Path
+
+import pytest
+
+from nisse_core.oplog import OpLog
+from nisse_core.reconcile import reconcile_on_boot
+from nisse_core.state import StateStore
+
+ENVELOPE_FIELDS = {"ts", "agent", "event", "task_id", "cause"}
+
+
+@pytest.fixture
+def stream() -> io.StringIO:
+    return io.StringIO()
+
+
+@pytest.fixture
+def store(tmp_path: Path, stream: io.StringIO) -> StateStore:
+    return StateStore(tmp_path / "state.db", oplog=OpLog("nisse-dev", stream))
+
+
+def events(stream: io.StringIO) -> list[dict]:
+    return [json.loads(line) for line in stream.getvalue().splitlines()]
+
+
+def start_item(
+    store: StateStore, item_id: str, *, scope: str | None = None, cause: str
+) -> None:
+    """Open a work item, make it the active one and set it running."""
+    store.open_work_item(
+        item_id, title=item_id, scope=scope, position="IDLE", cause=cause
+    )
+    store.set_active(item_id, cause=cause)
+    store.update_work_item(item_id, position="RUNNING", data={}, cause=cause)
+
+
+def test_every_event_carries_the_envelope(
+    store: StateStore, stream: io.StringIO
+) -> None:
+    start_item(store, "demo-1", cause="issue assigned")
+    intent_id = store.begin_intent("push_branch", {}, cause="milestone execution")
+    store.commit_intent(intent_id, cause="push verified")
+    for event in events(stream):
+        assert ENVELOPE_FIELDS <= event.keys()
+        assert event["agent"] == "nisse-dev"
+
+
+def test_a_position_change_logs_from_to_and_cause(
+    store: StateStore, stream: io.StringIO
+) -> None:
+    start_item(store, "d-1", scope="d", cause="issue assigned")
+    store.update_work_item("d-1", position="WAITING", data={}, cause="restated")
+    # Opening the work item logs its own events first (work_item.open, .switch).
+    first, second = [e for e in events(stream) if e["event"] == "state.transition"]
+    assert (first["from"], first["to"]) == ("IDLE", "RUNNING")
+    assert (second["from"], second["to"], second["project"]) == (
+        "RUNNING",
+        "WAITING",
+        "d",
+    )
+    assert second["cause"] == "restated"
+
+
+def test_intent_lifecycle_events(store: StateStore, stream: io.StringIO) -> None:
+    intent_id = store.begin_intent("push_branch", {"b": "m2"}, cause="milestone")
+    store.abort_intent(intent_id, cause="discarded")
+    begin, abort = events(stream)
+    assert begin["event"] == "intent.begin"
+    assert abort["event"] == "intent.abort"
+    assert begin["intent_id"] == abort["intent_id"] == intent_id
+    assert begin["kind"] == abort["kind"] == "push_branch"
+
+
+def test_task_id_threads_through_events(store: StateStore, stream: io.StringIO) -> None:
+    store.open_work_item("general", title="general", position="IDLE", cause="boot")
+    store.oplog.task_id = "task-42"
+    start_item(store, "task-42", cause="issue assigned")
+    before, *during = events(stream)
+    assert before["task_id"] is None
+    assert {e["task_id"] for e in during} == {"task-42"}
+
+
+def test_migration_emits_event(tmp_path: Path, stream: io.StringIO) -> None:
+    db = tmp_path / "state.db"
+    with StateStore(db, schema_version=1):
+        pass
+
+    def add_notes(conn: sqlite3.Connection) -> None:
+        conn.execute("CREATE TABLE notes (id INTEGER PRIMARY KEY)")
+
+    with StateStore(
+        db,
+        schema_version=2,
+        migrations={1: add_notes},
+        oplog=OpLog("nisse-dev", stream),
+    ):
+        pass
+    [event] = events(stream)
+    assert event["event"] == "schema.migrate"
+    assert (event["from_version"], event["to_version"]) == (1, 2)
+
+
+def test_boot_reconcile_emits_action_and_verdicts(
+    store: StateStore, stream: io.StringIO
+) -> None:
+    store.begin_intent("push_branch", {}, cause="milestone")
+    reconcile_on_boot(store, {"push_branch": lambda intent: True})
+    boot = [e for e in events(stream) if e["event"] == "boot.reconcile"]
+    assert len(boot) == 1
+    assert boot[0]["action"] == "resume"
+    assert "effect verified" in boot[0]["verdicts"][0]
+    commit = [e for e in events(stream) if e["event"] == "intent.commit"]
+    assert commit[0]["cause"] == "reconciliation verdict: effect verified"
+
+
+def test_literal_log_lines_for_human_readability(
+    store: StateStore, stream: io.StringIO, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    """Asserts the exact bytes written per line, so the on-disk format is
+    visible here rather than only in a running system."""
+
+    class _FrozenDatetime:
+        @staticmethod
+        def now(tz: object) -> object:
+            from datetime import UTC, datetime
+
+            return datetime(2026, 9, 5, 12, 0, 0, tzinfo=UTC)
+
+    monkeypatch.setattr("nisse_core.oplog.datetime", _FrozenDatetime)
+
+    start_item(store, "demo-m2", scope="demo", cause="issue assigned: demo#2")
+    store.oplog.task_id = "demo-m2"
+    intent_id = store.begin_intent(
+        "push_branch", {"branch": "milestone-2"}, cause="milestone 2 execution"
+    )
+    store.commit_intent(intent_id, cause="push verified by Gitea query")
+
+    assert stream.getvalue().splitlines() == [
+        (
+            '{"ts": "2026-09-05T12:00:00+00:00", "agent": "nisse-dev",'
+            ' "event": "work_item.open", "task_id": null,'
+            ' "cause": "issue assigned: demo#2",'
+            ' "work_item": "demo-m2", "scope": "demo"}'
+        ),
+        (
+            '{"ts": "2026-09-05T12:00:00+00:00", "agent": "nisse-dev",'
+            ' "event": "work_item.switch", "task_id": null,'
+            ' "cause": "issue assigned: demo#2",'
+            ' "from": null, "to": "demo-m2"}'
+        ),
+        (
+            '{"ts": "2026-09-05T12:00:00+00:00", "agent": "nisse-dev",'
+            ' "event": "state.transition", "task_id": null,'
+            ' "cause": "issue assigned: demo#2",'
+            ' "from": "IDLE", "to": "RUNNING", "project": "demo"}'
+        ),
+        (
+            '{"ts": "2026-09-05T12:00:00+00:00", "agent": "nisse-dev",'
+            ' "event": "intent.begin", "task_id": "demo-m2",'
+            ' "cause": "milestone 2 execution", "intent_id": 1,'
+            ' "kind": "push_branch"}'
+        ),
+        (
+            '{"ts": "2026-09-05T12:00:00+00:00", "agent": "nisse-dev",'
+            ' "event": "intent.commit", "task_id": "demo-m2",'
+            ' "cause": "push verified by Gitea query",'
+            ' "intent_id": 1, "kind": "push_branch"}'
+        ),
+    ]
+
+
+def test_failed_resolution_emits_nothing(
+    store: StateStore, stream: io.StringIO
+) -> None:
+    with pytest.raises(KeyError):
+        store.commit_intent(9999, cause="test")
+    assert events(stream) == []

nisse-core/tests/test_probe.py

@@ -0,0 +1,89 @@
+import json
+
+import httpx
+
+from nisse_core.probe import interpret, probe
+
+
+def test_a_cut_at_the_declared_window_matches() -> None:
+    result = interpret(declared=4096, sent=6144, reported=4096, error=None)
+    assert result.verdict == "matches"
+    assert result.server_window == 4096
+
+
+def test_a_cut_below_the_declaration_says_what_to_set() -> None:
+    # The dangerous case: prompts between the two would be silently cut.
+    result = interpret(declared=8192, sent=12288, reported=4096, error=None)
+    assert result.verdict == "declared_too_high"
+    assert "set max_input_tokens to 4096" in result.message
+
+
+def test_a_cut_above_the_declaration_is_conservative() -> None:
+    result = interpret(declared=4096, sent=6144, reported=5000, error=None)
+    assert result.verdict == "declared_conservative"
+
+
+def test_an_uncut_prompt_means_the_server_allows_more() -> None:
+    result = interpret(declared=4096, sent=6144, reported=6100, error=None)
+    assert result.verdict == "declared_conservative"
+    assert result.server_window is None
+
+
+def test_a_refusal_is_the_safe_behaviour() -> None:
+    result = interpret(declared=4096, sent=6144, reported=None, error="400: too long")
+    assert result.verdict == "refuses"
+
+
+def test_probe_calibrates_then_sends_half_again_the_window() -> None:
+    """Against a fake Ollama: 2 tokens a word, silently cut at 4096."""
+    sizes: list[int] = []
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        words = len(json.loads(request.content)["messages"][0]["content"].split())
+        sizes.append(words * 2)
+        reported = min(words * 2, 4096)
+        return httpx.Response(
+            200, json={"model": "llama3.2", "usage": {"prompt_tokens": reported}}
+        )
+
+    with httpx.Client(
+        base_url="http://litellm", transport=httpx.MockTransport(handler)
+    ) as http:
+        result = probe(http, "llama3.2", declared=4096)
+
+    assert result.verdict == "matches"
+    assert sizes[1] >= 4096 * 1.4  # the probe really exceeded the window
+
+
+def test_a_fallback_answer_is_never_credited_to_the_model() -> None:
+    """Measured: LiteLLM answered an oversized prompt to local-coder with
+    claude-sonnet-5, and the probe called the refusing server roomy."""
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        words = len(json.loads(request.content)["messages"][0]["content"].split())
+        model = "local-coder" if words < 1000 else "claude-sonnet-5"
+        return httpx.Response(
+            200, json={"model": model, "usage": {"prompt_tokens": words * 2}}
+        )
+
+    with httpx.Client(
+        base_url="http://litellm", transport=httpx.MockTransport(handler)
+    ) as http:
+        result = probe(http, "local-coder", declared=32768, backing="openai/qwen3-14b")
+
+    assert result.verdict == "fallback"
+    assert "claude-sonnet-5 answered instead of local-coder" in result.message
+
+
+def test_a_provider_reporting_its_own_model_name_is_not_a_fallback() -> None:
+    def handler(request: httpx.Request) -> httpx.Response:
+        return httpx.Response(
+            200, json={"model": "qwen3-14b", "usage": {"prompt_tokens": 400}}
+        )
+
+    with httpx.Client(
+        base_url="http://litellm", transport=httpx.MockTransport(handler)
+    ) as http:
+        result = probe(http, "local-coder", declared=32768, backing="openai/qwen3-14b")
+
+    assert result.verdict != "fallback"

nisse-core/tests/test_reconcile.py

@@ -0,0 +1,68 @@
+from pathlib import Path
+
+import pytest
+
+from nisse_core.reconcile import reconcile_on_boot
+from nisse_core.state import Intent, StateStore
+
+
+@pytest.fixture
+def store(tmp_path: Path) -> StateStore:
+    return StateStore(tmp_path / "state.db")
+
+
+def test_no_pending_intents_resumes(store: StateStore) -> None:
+    report = reconcile_on_boot(store, {})
+    assert report.action == "resume"
+    assert report.notes == []
+
+
+def test_verified_effect_commits_intent_and_resumes(store: StateStore) -> None:
+    store.begin_intent("push_branch", {"branch": "m2"}, cause="test")
+    report = reconcile_on_boot(store, {"push_branch": lambda intent: True})
+    assert report.action == "resume"
+    assert store.pending_intents() == []
+
+
+def test_verified_no_effect_aborts_intent_and_resumes(store: StateStore) -> None:
+    store.begin_intent("push_branch", {"branch": "m2"}, cause="test")
+    report = reconcile_on_boot(store, {"push_branch": lambda intent: False})
+    assert report.action == "resume"
+    assert store.pending_intents() == []
+
+
+def test_unverifiable_intent_gates_to_await_input(store: StateStore) -> None:
+    store.begin_intent("push_branch", {"branch": "m2"}, cause="test")
+    report = reconcile_on_boot(store, {"push_branch": lambda intent: None})
+    assert report.action == "await_input"
+    assert [i.kind for i in store.pending_intents()] == ["push_branch"]
+
+
+def test_unknown_intent_kind_gates_to_await_input(store: StateStore) -> None:
+    store.begin_intent("mystery", {}, cause="test")
+    report = reconcile_on_boot(store, {})
+    assert report.action == "await_input"
+    assert "no reconciler registered" in report.notes[0]
+
+
+def test_reconciler_receives_the_intent(store: StateStore) -> None:
+    store.begin_intent("push_branch", {"branch": "m2"}, cause="test")
+    seen: list[Intent] = []
+
+    def check(intent: Intent) -> bool:
+        seen.append(intent)
+        return True
+
+    reconcile_on_boot(store, {"push_branch": check})
+    assert seen[0].payload == {"branch": "m2"}
+
+
+def test_mixed_verdicts_still_gate(store: StateStore) -> None:
+    store.begin_intent("push_branch", {}, cause="test")
+    store.begin_intent("open_pr", {}, cause="test")
+    report = reconcile_on_boot(
+        store,
+        {"push_branch": lambda intent: True, "open_pr": lambda intent: None},
+    )
+    assert report.action == "await_input"
+    assert [i.kind for i in store.pending_intents()] == ["open_pr"]

nisse-core/tests/test_render.py

@@ -0,0 +1,151 @@
+import json
+from itertools import pairwise
+
+from nisse_core.canonical import (
+    AssistantText,
+    Compaction,
+    GateOpened,
+    StoredEvent,
+    ToolCall,
+    ToolResult,
+    UserMessage,
+)
+from nisse_core.render import OpenAIChatRenderer, effective_events
+
+
+def stored(events: list) -> list[StoredEvent]:
+    return [StoredEvent(id=i + 1, ts="t", event=e) for i, e in enumerate(events)]
+
+
+def test_renders_mixed_transcript_in_order() -> None:
+    events = stored(
+        [
+            UserMessage(text="run the tests", source="m:1"),
+            ToolCall(call_id="call-9", tool="run_tests", arguments={"path": "x"}),
+            ToolResult(call_id="call-9", result="1 passed", success=True),
+            AssistantText(text="all green", model="local/qwen"),
+        ]
+    )
+    body = OpenAIChatRenderer().render(events, tools=None, model="local/qwen")
+    assert body["model"] == "local/qwen"
+    roles = [m["role"] for m in body["messages"]]
+    assert roles == ["user", "assistant", "tool", "assistant"]
+    call = body["messages"][1]["tool_calls"][0]
+    assert call["id"] == "call-9"
+    assert call["function"]["name"] == "run_tests"
+    assert json.loads(call["function"]["arguments"]) == {"path": "x"}
+    assert body["messages"][2]["tool_call_id"] == "call-9"
+    assert "tools" not in body
+
+
+def test_system_prompt_and_tools_pass_through() -> None:
+    tools = [{"type": "function", "function": {"name": "run_tests"}}]
+    body = OpenAIChatRenderer().render(
+        stored([UserMessage(text="hi", source="m:1")]),
+        tools=tools,
+        model="local/qwen",
+        system="you are nisse-dev",
+    )
+    assert body["messages"][0] == {"role": "system", "content": "you are nisse-dev"}
+    assert body["tools"] == tools
+
+
+def test_compaction_replaces_its_range_only() -> None:
+    events = stored(
+        [
+            UserMessage(text="old-1", source="m:1"),
+            AssistantText(text="old-2", model="q"),
+            UserMessage(text="kept", source="m:3"),
+        ]
+    )
+    events.append(
+        StoredEvent(
+            id=4,
+            ts="t",
+            event=Compaction(summary="the old stuff", model="q", replaces=[1, 2]),
+        )
+    )
+    remaining = effective_events(events)
+    assert [e.id for e in remaining] == [3, 4]
+    body = OpenAIChatRenderer().render(events, tools=None, model="q")
+    contents = [m["content"] for m in body["messages"]]
+    # Both render as user messages, so they share one: roles must alternate.
+    assert contents == ["kept\n\n[summary of earlier conversation]\nthe old stuff"]
+
+
+def test_tool_call_without_a_result_gets_a_placeholder() -> None:
+    # Providers reject a dangling tool_use; this happens after a crash (L8)
+    # or with turn-ending gate tools like submit_plan.
+    events = stored(
+        [
+            UserMessage(text="go", source="m:1"),
+            ToolCall(call_id="call-x", tool="submit_plan", arguments={}),
+        ]
+    )
+    body = OpenAIChatRenderer().render(events, tools=None, model="q")
+    assert [m["role"] for m in body["messages"]] == ["user", "assistant", "tool"]
+    assert body["messages"][2]["tool_call_id"] == "call-x"
+    assert body["messages"][2]["content"] == "[no result recorded]"
+
+
+def test_answered_tool_call_gets_no_placeholder() -> None:
+    events = stored(
+        [
+            ToolCall(call_id="call-x", tool="run_command", arguments={}),
+            ToolResult(call_id="call-x", result="ok", success=True),
+        ]
+    )
+    body = OpenAIChatRenderer().render(events, tools=None, model="q")
+    assert [m["role"] for m in body["messages"]] == ["assistant", "tool"]
+    assert body["messages"][1]["content"] == "ok"
+
+
+def test_gate_event_renders_as_context() -> None:
+    events = stored([GateOpened(gate="arch", reason="plan submitted (rev-1)")])
+    body = OpenAIChatRenderer().render(events, tools=None, model="q")
+    assert body["messages"][0]["role"] == "user"
+    assert "arch" in body["messages"][0]["content"]
+
+
+def _roles(body: dict) -> list[str]:
+    return [m["role"] for m in body["messages"]]
+
+
+def test_one_turn_is_one_assistant_message() -> None:
+    # Text and two parallel tool calls from one model turn: the OpenAI format
+    # carries them as a single assistant message with a tool_calls list.
+    events = stored(
+        [
+            UserMessage(text="go", source="m:1"),
+            AssistantText(text="Looking around first.", model="q"),
+            ToolCall(call_id="a", tool="list_dir", arguments={"path": "."}),
+            ToolCall(call_id="b", tool="read_file", arguments={"path": "README.md"}),
+            ToolResult(call_id="a", result="README.md", success=True),
+            ToolResult(call_id="b", result="hello", success=True),
+        ]
+    )
+    body = OpenAIChatRenderer().render(events, tools=None, model="q")
+    assert _roles(body) == ["user", "assistant", "tool", "tool"]
+    turn = body["messages"][1]
+    assert turn["content"] == "Looking around first."
+    assert [c["id"] for c in turn["tool_calls"]] == ["a", "b"]
+
+
+def test_roles_never_repeat_for_strict_templates() -> None:
+    """Mistral's chat template raises on two user or two assistant messages
+    in a row; every mix of events must render without one."""
+    events = stored(
+        [
+            UserMessage(text="one", source="m:1"),
+            UserMessage(text="two", source="m:2"),
+            AssistantText(text="thinking", model="q"),
+            ToolCall(call_id="a", tool="list_dir", arguments={}),
+            ToolResult(call_id="a", result="x", success=True),
+            AssistantText(text="more", model="q"),
+            AssistantText(text="and more", model="q"),
+            ToolCall(call_id="b", tool="finish", arguments={}),
+        ]
+    )
+    roles = _roles(OpenAIChatRenderer().render(events, tools=None, model="q"))
+    for first, second in pairwise(roles):
+        assert not (first == second and first in ("user", "assistant")), roles

nisse-core/tests/test_render_anthropic.py

@@ -0,0 +1,196 @@
+"""The Anthropic Messages shape differs structurally from the OpenAI one;
+these tests pin the differences that a normalising middlebox would blur."""
+
+import httpx
+import pytest
+
+from nisse_core.canonical import (
+    AssistantText,
+    Compaction,
+    StoredEvent,
+    ToolCall,
+    ToolResult,
+    UserMessage,
+)
+from nisse_core.llm import AnthropicPassthroughClient, LLMError
+from nisse_core.render_anthropic import (
+    AnthropicMessagesRenderer,
+    to_anthropic_tools,
+)
+
+TOOLS = [
+    {
+        "type": "function",
+        "function": {
+            "name": "run_command",
+            "description": "Run a shell command.",
+            "parameters": {
+                "type": "object",
+                "properties": {"command": {"type": "string"}},
+                "required": ["command"],
+            },
+        },
+    }
+]
+
+
+def stored(events: list) -> list[StoredEvent]:
+    return [StoredEvent(id=i + 1, ts="t", event=e) for i, e in enumerate(events)]
+
+
+def render(events: list, **kw) -> dict:
+    return AnthropicMessagesRenderer().render(
+        stored(events), kw.pop("tools", None), "claude-haiku-4-5", **kw
+    )
+
+
+def test_system_is_a_top_level_field_not_a_message() -> None:
+    body = render([UserMessage(text="hi", source="m")], system="you are nisse")
+    assert body["system"] == "you are nisse"
+    assert all(m["role"] != "system" for m in body["messages"])
+
+
+def test_max_tokens_is_always_present() -> None:
+    # The Messages API rejects a request without it.
+    assert render([UserMessage(text="hi", source="m")])["max_tokens"] > 0
+
+
+def test_tool_call_is_a_content_block_not_an_array() -> None:
+    body = render(
+        [ToolCall(call_id="call-1", tool="run_command", arguments={"command": "ls"})]
+    )
+    block = body["messages"][0]["content"][0]
+    assert block["type"] == "tool_use"
+    assert block["id"] == "call-1"
+    assert block["input"] == {"command": "ls"}
+    assert "tool_calls" not in body["messages"][0]
+
+
+def test_tool_result_is_a_user_block_directly_after_the_call() -> None:
+    body = render(
+        [
+            ToolCall(call_id="call-1", tool="run_command", arguments={}),
+            ToolResult(call_id="call-1", result="exit code: 0", success=True),
+        ]
+    )
+    assistant, user = body["messages"]
+    assert assistant["role"] == "assistant"
+    assert assistant["content"][0]["type"] == "tool_use"
+    assert user["role"] == "user"  # not a "tool" role, as OpenAI would have
+    assert user["content"][0] == {
+        "type": "tool_result",
+        "tool_use_id": "call-1",
+        "content": "exit code: 0",
+    }
+
+
+def test_adjacent_same_role_events_are_merged() -> None:
+    body = render(
+        [
+            UserMessage(text="one", source="m"),
+            UserMessage(text="two", source="m"),
+            AssistantText(text="reply", model="m"),
+        ]
+    )
+    assert [m["role"] for m in body["messages"]] == ["user", "assistant"]
+    assert [b["text"] for b in body["messages"][0]["content"]] == ["one", "two"]
+
+
+def test_dangling_tool_call_gets_a_placeholder_result() -> None:
+    body = render([ToolCall(call_id="call-x", tool="submit_plan", arguments={})])
+    assert [m["role"] for m in body["messages"]] == ["assistant", "user"]
+    assert body["messages"][1]["content"][0]["tool_use_id"] == "call-x"
+
+
+def test_tools_are_converted_to_input_schema() -> None:
+    [tool] = to_anthropic_tools(TOOLS)
+    assert tool["name"] == "run_command"
+    assert tool["input_schema"]["properties"] == {"command": {"type": "string"}}
+    assert "parameters" not in tool
+    assert "function" not in tool
+
+
+def test_compaction_still_replaces_its_range() -> None:
+    events = stored(
+        [
+            UserMessage(text="old", source="m"),
+            UserMessage(text="kept", source="m"),
+        ]
+    )
+    events.append(
+        StoredEvent(
+            id=3,
+            ts="t",
+            event=Compaction(summary="summary", model="m", replaces=[1, 1]),
+        )
+    )
+    body = AnthropicMessagesRenderer().render(events, None, "claude-haiku-4-5")
+    texts = [b["text"] for m in body["messages"] for b in m["content"]]
+    assert "old" not in texts
+    assert any("summary" in t for t in texts)
+
+
+# --- passthrough client --------------------------------------------------
+
+
+def client(handler) -> AnthropicPassthroughClient:
+    return AnthropicPassthroughClient(
+        "https://litellm.test", "tok", transport=httpx.MockTransport(handler)
+    )
+
+
+def test_client_posts_to_the_messages_route_with_version_header() -> None:
+    seen: dict = {}
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        seen["url"] = str(request.url)
+        seen["version"] = request.headers.get("anthropic-version")
+        return httpx.Response(
+            200,
+            json={
+                "model": "claude-haiku-4-5",
+                "content": [{"type": "text", "text": "hi"}],
+            },
+        )
+
+    events = stored([UserMessage(text="hello", source="m")])
+    [event] = client(handler).complete(events, None, "claude-haiku-4-5")
+    assert seen["url"] == "https://litellm.test/v1/messages"
+    assert seen["version"] == "2023-06-01"
+    assert event == AssistantText(text="hi", model="claude-haiku-4-5")
+
+
+def test_client_maps_tool_use_to_local_call_ids() -> None:
+    response = {
+        "model": "claude-haiku-4-5",
+        "content": [
+            {"type": "text", "text": "running it"},
+            {
+                "type": "tool_use",
+                "id": "toolu_vendor_xyz",
+                "name": "run_command",
+                "input": {"command": "pytest"},
+            },
+        ],
+    }
+    events = stored([UserMessage(text="go", source="m")])
+    text, call = client(lambda r: httpx.Response(200, json=response)).complete(
+        events, TOOLS, "claude-haiku-4-5"
+    )
+    assert isinstance(text, AssistantText)
+    assert isinstance(call, ToolCall)
+    assert call.arguments == {"command": "pytest"}
+    assert call.call_id != "toolu_vendor_xyz"  # provider id never a join key
+    assert call.call_id.startswith("call-")
+
+
+def test_client_errors_are_llm_errors() -> None:
+    events = stored([UserMessage(text="go", source="m")])
+    with pytest.raises(LLMError, match="400"):
+        client(lambda r: httpx.Response(400, text="bad")).complete(
+            events, None, "claude-haiku-4-5"
+        )
+    with pytest.raises(LLMError, match="malformed"):
+        client(lambda r: httpx.Response(200, json={"oops": 1})).complete(
+            events, None, "claude-haiku-4-5"
+        )

nisse-core/tests/test_state.py

@@ -0,0 +1,123 @@
+import sqlite3
+from pathlib import Path
+
+import pytest
+
+from nisse_core.state import StateSchemaError, StateStore
+
+
+@pytest.fixture
+def db(tmp_path: Path) -> Path:
+    return tmp_path / "state.db"
+
+
+def test_wal_mode_and_full_sync(db: Path) -> None:
+    with StateStore(db) as store:
+        journal = store._conn.execute("PRAGMA journal_mode").fetchone()[0]
+        sync = store._conn.execute("PRAGMA synchronous").fetchone()[0]
+    assert journal == "wal"
+    assert sync == 2  # FULL
+
+
+def test_fresh_store_stamps_schema_version(db: Path) -> None:
+    with StateStore(db, schema_version=1) as store:
+        assert store.schema_version == 1
+
+
+def test_a_work_item_persists_across_reopen(db: Path) -> None:
+    with StateStore(db) as store:
+        store.open_work_item(
+            "demo-1", title="t", scope="demo", position="WAITING", cause="test"
+        )
+        store.set_active("demo-1", cause="test")
+        store.update_work_item(
+            "demo-1", position="RUNNING", data={"milestone": 2}, cause="test"
+        )
+    with StateStore(db) as store:
+        item = store.active_item()
+    assert item is not None
+    assert (item.id, item.scope, item.position) == ("demo-1", "demo", "RUNNING")
+    assert item.data == {"milestone": 2}
+
+
+def test_a_fresh_store_has_no_active_item(db: Path) -> None:
+    with StateStore(db) as store:
+        assert store.active_item() is None
+
+
+def test_intent_lifecycle(db: Path) -> None:
+    with StateStore(db) as store:
+        intent_id = store.begin_intent("push_branch", {"branch": "m2"}, cause="test")
+        [pending] = store.pending_intents()
+        assert pending.id == intent_id
+        assert pending.kind == "push_branch"
+        assert pending.payload == {"branch": "m2"}
+        store.commit_intent(intent_id, cause="test")
+        assert store.pending_intents() == []
+
+
+def test_pending_intent_survives_reopen(db: Path) -> None:
+    with StateStore(db) as store:
+        store.begin_intent("open_pr", {"title": "hello"}, cause="test")
+    with StateStore(db) as store:
+        [pending] = store.pending_intents()
+    assert pending.kind == "open_pr"
+
+
+def test_aborted_intent_is_not_pending(db: Path) -> None:
+    with StateStore(db) as store:
+        intent_id = store.begin_intent("push_branch", {}, cause="test")
+        store.abort_intent(intent_id, cause="test")
+        assert store.pending_intents() == []
+
+
+def test_resolving_nonpending_intent_raises(db: Path) -> None:
+    with StateStore(db) as store:
+        intent_id = store.begin_intent("push_branch", {}, cause="test")
+        store.commit_intent(intent_id, cause="test")
+        with pytest.raises(KeyError):
+            store.commit_intent(intent_id, cause="test")
+        with pytest.raises(KeyError):
+            store.abort_intent(9999, cause="test")
+
+
+def test_migration_runs_and_bumps_version(db: Path) -> None:
+    with StateStore(db, schema_version=1):
+        pass
+
+    def add_notes_table(conn: sqlite3.Connection) -> None:
+        conn.execute("CREATE TABLE notes (id INTEGER PRIMARY KEY, body TEXT)")
+
+    with StateStore(db, schema_version=2, migrations={1: add_notes_table}) as store:
+        assert store.schema_version == 2
+        store._conn.execute("SELECT * FROM notes")
+
+
+def test_missing_migration_raises(db: Path) -> None:
+    with StateStore(db, schema_version=1):
+        pass
+    with pytest.raises(StateSchemaError, match="no migration"):
+        StateStore(db, schema_version=2, migrations={})
+
+
+def test_failed_migration_leaves_version_unchanged(db: Path) -> None:
+    with StateStore(db, schema_version=1):
+        pass
+
+    def broken(conn: sqlite3.Connection) -> None:
+        conn.execute("CREATE TABLE notes (id INTEGER PRIMARY KEY)")
+        raise RuntimeError("boom")
+
+    with pytest.raises(RuntimeError, match="boom"):
+        StateStore(db, schema_version=2, migrations={1: broken})
+    with StateStore(db, schema_version=1) as store:
+        assert store.schema_version == 1
+        with pytest.raises(sqlite3.OperationalError):
+            store._conn.execute("SELECT * FROM notes")
+
+
+def test_newer_schema_is_refused(db: Path) -> None:
+    with StateStore(db, schema_version=3):
+        pass
+    with pytest.raises(StateSchemaError, match="newer"):
+        StateStore(db, schema_version=1)

nisse-core/tests/test_vault.py

@@ -0,0 +1,145 @@
+"""The knowledge vault (L10, M5.1): only configured folders exist, paths
+can't escape them, read-only folders refuse writes, writes are atomic and
+record provenance, and the editor's own files are never notes."""
+
+import os
+from datetime import UTC, datetime
+from pathlib import Path
+
+import pytest
+import yaml
+
+from nisse_core import vault as vault_module
+from nisse_core.vault import Vault, VaultError, split_frontmatter
+
+
+@pytest.fixture
+def vault(tmp_path: Path) -> Vault:
+    root = tmp_path / "vault"
+    (root / "notes" / "deep").mkdir(parents=True)
+    (root / "notes" / "Procedures.md").write_text("# Procedures\n\nRestart the hub.\n")
+    (root / "notes" / "deep" / "Hub.md").write_text(
+        "The hub is in the shed.\nhub hub\n"
+    )
+    (root / "notes" / ".silverbullet").mkdir()
+    (root / "notes" / ".silverbullet" / "state.md").write_text("hub")
+    (root / "notes" / "_plug").mkdir()
+    (root / "notes" / "_plug" / "x.md").write_text("hub")
+    (root / "private").mkdir()
+    (root / "private" / "Secret.md").write_text("the hub password")
+    (root / "memory" / "nisse-ops").mkdir(parents=True)
+    (root / "memory" / "nisse-ops" / "Ops.md").write_text("ops hub")
+    return Vault(
+        root,
+        {"memory/nisse-dev": "rw", "notes": "ro"},
+        agent="nisse-dev",
+        clock=lambda: datetime(2026, 10, 8, 12, 0, tzinfo=UTC),
+    )
+
+
+def test_only_configured_folders_exist(vault: Vault) -> None:
+    assert vault.read("notes/Procedures.md").startswith("# Procedures")
+    for path in ("private/Secret.md", "memory/nisse-ops/Ops.md", "Top.md"):
+        with pytest.raises(VaultError, match="not in a folder you can see"):
+            vault.read(path)
+    assert vault.notes() == ["notes/Procedures.md", "notes/deep/Hub.md"]
+    assert vault.notes("notes/deep") == ["notes/deep/Hub.md"]
+
+
+@pytest.mark.parametrize(
+    "path",
+    ["/etc/passwd", "notes/../private/Secret.md", "", "notes\\x.md", "../x.md"],
+)
+def test_paths_cant_escape(vault: Vault, path: str) -> None:
+    with pytest.raises(VaultError):
+        vault.read(path)
+
+
+def test_a_link_out_of_its_folder_is_refused(vault: Vault) -> None:
+    os.symlink(vault.root / "private" / "Secret.md", vault.root / "notes" / "Ln.md")
+    with pytest.raises(VaultError, match="leads outside its folder"):
+        vault.read("notes/Ln.md")
+    assert "notes/Ln.md" not in vault.notes()
+    assert all(f.path != "notes/Ln.md" for f in vault.search("password"))
+
+
+def test_the_editors_files_are_not_notes(vault: Vault) -> None:
+    with pytest.raises(VaultError, match="editor's own files"):
+        vault.read("notes/.silverbullet/state.md")
+    with pytest.raises(VaultError, match="editor's own files"):
+        vault.read("notes/_plug/x.md")
+    with pytest.raises(VaultError, match="Markdown"):
+        vault.read("notes/deep")
+
+
+def test_search_ranks_matches_and_sees_only_what_it_may(vault: Vault) -> None:
+    found = vault.search("HUB")
+    assert [f.path for f in found] == ["notes/deep/Hub.md", "notes/Procedures.md"]
+    assert found[0].lines == ("The hub is in the shed.", "hub hub")
+    with pytest.raises(VaultError):
+        vault.search("  ")
+
+
+def test_a_long_note_is_cut_and_says_so(vault: Vault) -> None:
+    (vault.root / "notes" / "Long.md").write_text("x" * 30_000)
+    text = vault.read("notes/Long.md")
+    assert text.endswith("[10000 more characters not shown]")
+
+
+def test_read_only_folders_refuse_writes(vault: Vault) -> None:
+    with pytest.raises(VaultError, match="read-only"):
+        vault.write("notes/New.md", "hi", provenance={"source": "t"})
+    assert not (vault.root / "notes" / "New.md").exists()
+
+
+def test_a_write_records_provenance_and_keeps_the_rest(vault: Vault) -> None:
+    text = "---\ntags: [hub]\naliases: [h]\n---\n# Hub\n\nIn the shed.\n"
+    vault.write(
+        "memory/nisse-dev/sub/Hub.md",
+        text,
+        provenance={"source": "matrix", "work_item": "sandbox-14"},
+    )
+    written = (vault.root / "memory" / "nisse-dev" / "sub" / "Hub.md").read_text()
+    front, body = split_frontmatter(written)
+    assert front["tags"] == ["hub"] and front["aliases"] == ["h"]
+    assert front["provenance"] == {
+        "by": "nisse-dev",
+        "source": "matrix",
+        "work_item": "sandbox-14",
+        "written": "2026-10-08T12:00:00+00:00",
+    }
+    assert body == "# Hub\n\nIn the shed.\n"
+    assert vault.notes("memory/nisse-dev") == ["memory/nisse-dev/sub/Hub.md"]
+
+
+def test_an_interrupted_write_leaves_the_old_note_and_no_litter(
+    vault: Vault, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    vault.write("memory/nisse-dev/N.md", "old\n", provenance={"source": "t"})
+    folder = vault.root / "memory" / "nisse-dev"
+    before = (folder / "N.md").read_text()
+
+    def crash(src: str, dst: str) -> None:
+        raise OSError("disk gone")
+
+    monkeypatch.setattr(vault_module.os, "replace", crash)
+    with pytest.raises(VaultError, match="couldn't write it"):
+        vault.write("memory/nisse-dev/N.md", "new\n", provenance={"source": "t"})
+    assert (folder / "N.md").read_text() == before
+    assert sorted(p.name for p in folder.iterdir()) == ["N.md"]
+
+
+def test_bad_frontmatter_is_refused(vault: Vault) -> None:
+    with pytest.raises(VaultError, match="valid YAML"):
+        vault.write(
+            "memory/nisse-dev/B.md", "---\n: [\n---\nx\n", provenance={"source": "t"}
+        )
+    with pytest.raises(VaultError, match="mapping"):
+        vault.write("memory/nisse-dev/B.md", "---\n- a\n---\nx\n", provenance={})
+
+
+def test_read_write_folders_are_made(tmp_path: Path) -> None:
+    Vault(tmp_path / "v", {"memory/nisse-dev": "rw", "notes": "ro"}, agent="a")
+    assert (tmp_path / "v" / "memory" / "nisse-dev").is_dir()
+    assert not (tmp_path / "v" / "notes").exists()  # not ours to make
+    assert yaml  # frontmatter is YAML

nisse-core/tests/test_views.py

@@ -0,0 +1,153 @@
+"""Per-role transcript views (architecture L6; Development Plan M1.2)."""
+
+import io
+from collections.abc import Mapping
+from pathlib import Path
+
+from nisse_core.canonical import (
+    AssistantText,
+    DelegationCompleted,
+    DelegationStarted,
+    Fact,
+    PlanSubmitted,
+    StoredEvent,
+    ToolCall,
+    ToolResult,
+    UserMessage,
+)
+from nisse_core.oplog import OpLog
+from nisse_core.render import OpenAIChatRenderer
+from nisse_core.render_anthropic import AnthropicMessagesRenderer
+from nisse_core.state import StateStore
+
+
+class Store(StateStore):
+    """An agent's store: the helper is shown the driver's work as narration,
+    the driver only the helper's report."""
+
+    SHOWN: Mapping[str, Mapping[str, str]] = {
+        "driver": {"helper": "report"},
+        "helper": {"driver": "narrate"},
+    }
+
+
+def new_store(tmp_path: Path) -> StateStore:
+    return Store(tmp_path / "state.db", oplog=OpLog("nisse-dev", io.StringIO()))
+
+
+def texts(events: list[StoredEvent]) -> list[str]:
+    return [getattr(e.event, "text", "") for e in events]
+
+
+# --- the projection, per role ----------------------------------------------------
+
+
+def delegated_milestone(store: StateStore) -> None:
+    """A milestone in which the driver failed and handed off to a helper."""
+    store.transcript("a", "driver").append(PlanSubmitted(plan="1. fix it", model="y"))
+    driver = store.transcript("a", "driver")
+    driver.append(
+        ToolCall(call_id="e1", tool="run_command", arguments={"command": "make"})
+    )
+    driver.append(ToolResult(call_id="e1", result="error: missing ;", success=False))
+    helper = driver.as_actor("helper")
+    helper.append(
+        DelegationStarted(brief="make fails", role="helper", model="y", caps={})
+    )
+    helper.append(AssistantText(text="looking at main.c", model="y"))
+    helper.append(
+        ToolCall(call_id="h1", tool="read_file", arguments={"path": "main.c"})
+    )
+    helper.append(ToolResult(call_id="h1", result="int main()", success=True))
+    helper.append(
+        DelegationCompleted(
+            handback="fixed the missing semicolon",
+            artifacts=["main.c"],
+            tokens=0,
+            duration_s=1.0,
+            exit_status="completed",
+        )
+    )
+
+
+def test_the_driver_sees_the_helpers_report_not_its_turns(tmp_path: Path) -> None:
+    store = new_store(tmp_path)
+    delegated_milestone(store)
+    view = store.transcript("a", "driver").view()
+    assert [type(e.event).__name__ for e in view] == [
+        "ToolCall",
+        "ToolResult",
+        "UserMessage",
+    ]
+    assert view[-1].event.text == (
+        "[helper report, completed] fixed the missing semicolon"
+    )
+
+
+def test_the_helper_sees_the_plan_and_the_drivers_work_as_narration(
+    tmp_path: Path,
+) -> None:
+    store = new_store(tmp_path)
+    delegated_milestone(store)
+    view = store.transcript("a", "helper").view()
+    said = texts(view)
+    assert said[0] == "[approved plan] 1. fix it"
+    assert said[1] == "[driver] run_command(make) → failed: error: missing ;"
+    # Its own events are unchanged, tool calls and all.
+    assert any(isinstance(e.event, ToolCall) and e.event.call_id == "h1" for e in view)
+    assert not any(
+        isinstance(e.event, ToolCall) and e.event.call_id == "e1" for e in view
+    )
+
+
+def test_a_view_keeps_log_ids_so_cursors_still_work(tmp_path: Path) -> None:
+    store = new_store(tmp_path)
+    delegated_milestone(store)
+    raw = store.transcript("a", "driver").read()
+    view = store.transcript("a", "driver").view(after=raw[2].id)
+    assert [e.id for e in view] == [raw[-1].id]  # the helper's report
+
+
+# --- rendering ---------------------------------------------------------------------
+
+
+def test_both_renderers_show_a_fact_as_a_labelled_user_message() -> None:
+    events = [StoredEvent(id=1, ts="", event=Fact(text="PR #3 opened"))]
+    openai = OpenAIChatRenderer().render(events, tools=None, model="m")
+    assert openai["messages"] == [
+        {"role": "user", "content": "[orchestrator] PR #3 opened"}
+    ]
+    anthropic = AnthropicMessagesRenderer().render(events, tools=None, model="m")
+    assert anthropic["messages"] == [
+        {
+            "role": "user",
+            "content": [{"type": "text", "text": "[orchestrator] PR #3 opened"}],
+        }
+    ]
+
+
+def test_operator_messages_carry_their_channel(tmp_path: Path) -> None:
+    store = new_store(tmp_path)
+    operator = store.transcript("a", "operator")
+    for source in ("transport:@don", "gitea:issue:9:comment:4", "gitea:comment:5"):
+        operator.append(UserMessage(text="hi", source=source))
+    assert texts(store.transcript("a", "driver").view()) == [
+        "[operator, Matrix] hi",
+        "[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"

nisse-core/tests/test_worker.py

@@ -0,0 +1,83 @@
+import io
+import threading
+from pathlib import Path
+
+import pytest
+
+from nisse_core.oplog import OpLog
+from nisse_core.state import StateStore
+from nisse_core.worker import InlineWorker, SerialWorker
+
+
+@pytest.fixture
+def worker() -> SerialWorker:
+    w = SerialWorker()
+    w.start()
+    yield w
+    w.stop(timeout=5)
+
+
+def test_jobs_run_in_order_on_the_worker_thread(worker: SerialWorker) -> None:
+    seen: list[tuple[int, str]] = []
+    for n in range(3):
+        worker.submit(lambda n=n: seen.append((n, threading.current_thread().name)))
+    worker.call(lambda: None)  # drains everything queued before it
+
+    assert [n for n, _ in seen] == [0, 1, 2]
+    assert {name for _, name in seen} == {"nisse-worker"}
+
+
+def test_call_returns_the_result_and_raises_the_error(worker: SerialWorker) -> None:
+    assert worker.call(lambda: 42) == 42
+
+    def boom() -> None:
+        raise ValueError("bad job")
+
+    with pytest.raises(ValueError, match="bad job"):
+        worker.call(boom)
+
+
+def test_a_failed_job_is_reported_and_the_worker_carries_on() -> None:
+    errors: list[BaseException] = []
+    worker = SerialWorker(on_error=errors.append)
+    worker.start()
+
+    def boom() -> None:
+        raise RuntimeError("lost job")
+
+    worker.submit(boom)
+    assert worker.call(lambda: "still here") == "still here"
+    assert [str(e) for e in errors] == ["lost job"]
+    worker.stop(timeout=5)
+
+
+def test_call_from_the_worker_itself_runs_inline(worker: SerialWorker) -> None:
+    # Queueing behind itself would deadlock.
+    assert worker.call(lambda: worker.call(lambda: "nested")) == "nested"
+
+
+def test_the_store_is_usable_only_from_the_thread_that_owns_it(
+    worker: SerialWorker, tmp_path: Path
+) -> None:
+    store = worker.call(
+        lambda: StateStore(tmp_path / "state.db", oplog=OpLog("t", io.StringIO()))
+    )
+    assert worker.call(store.active_item) is None
+    # The tripwire the worker design keeps on: any other thread fails loudly.
+    with pytest.raises(Exception, match="same thread"):
+        store.active_item()
+    worker.call(store.close)
+
+
+def test_inline_worker_defers_a_job_submitted_from_inside_a_job() -> None:
+    worker = InlineWorker()
+    order: list[str] = []
+
+    def outer() -> None:
+        order.append("outer start")
+        worker.submit(lambda: order.append("inner"))
+        order.append("outer end")
+
+    worker.submit(outer)
+    assert order == ["outer start", "outer end", "inner"]
+    assert not worker.busy