nisse-core: group modules into packages by architecture section

A pure move; behaviour unchanged.

- models (L9, L7): llm, catalog, probe, render, render_anthropic, context
- log (L7, L6): canonical, views
- state (L8): store (was state), reconcile, worker, cancel, dump
- knowledge (L10, L6.4): vault, memory
- conversation (L5): transport, matrix
- execution (L13): ssh (was execution), testing
- authority (P3, P4, L4): actions
- gitea, oplog, envfile stay at the top until each gains a second module.

Every import, monkeypatch target, script and doc reference rewritten.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01J5Utyfzw7gF7UhjRDEtjTk

nisse-core/src/nisse_core/authority/__init__.py

@@ -0,0 +1 @@
+"""Actions by their effects (architecture P3, P4, L3, L4)."""

nisse-core/src/nisse_core/conversation/__init__.py

@@ -0,0 +1 @@
+"""Conversation with the operator (architecture L5)."""

nisse-core/src/nisse_core/conversation/matrix.py

@@ -27,7 +27,7 @@ from urllib.parse import unquote
 
 import httpx
 
-from nisse_core.transport import InboundMessage
+from nisse_core.conversation.transport import InboundMessage
 
 # A mention pill at the start of an HTML body: the link holds the user ID,
 # the link text is what the plain body starts with.

nisse-core/src/nisse_core/execution/__init__.py

@@ -0,0 +1 @@
+"""Running commands on tool VMs over SSH (architecture L13)."""

nisse-core/src/nisse_core/execution/testing.py

@@ -14,7 +14,7 @@ from collections.abc import Iterator
 from contextlib import contextmanager
 from pathlib import Path
 
-from nisse_core.execution import SSHHost
+from nisse_core.execution.ssh import SSHHost
 
 SSHD = "/usr/sbin/sshd"
 

nisse-core/src/nisse_core/gitea.py

@@ -16,8 +16,8 @@ from typing import Any
 
 import httpx
 
-from nisse_core.reconcile import Reconciler
-from nisse_core.state import Intent
+from nisse_core.state.reconcile import Reconciler
+from nisse_core.state.store import Intent
 
 PR_PAGE_SIZE = 50
 MAX_PR_PAGES = 40  # 2000 PRs; beyond that, say so rather than guess

nisse-core/src/nisse_core/knowledge/__init__.py

@@ -0,0 +1 @@
+"""The knowledge vault (architecture L10) and agent memory (L6.4)."""

nisse-core/src/nisse_core/knowledge/memory.py

@@ -14,7 +14,7 @@ from dataclasses import dataclass
 
 import yaml
 
-from nisse_core.vault import Vault, VaultError, split_frontmatter
+from nisse_core.knowledge.vault import Vault, VaultError, split_frontmatter
 
 INDEX = "index"  # the note the harness keeps; never the model's
 _NAME = re.compile(r"[a-z0-9][a-z0-9-]{0,63}")

nisse-core/src/nisse_core/log/__init__.py

@@ -0,0 +1,2 @@
+"""The canonical log (architecture L7) and what each role is shown of it
+(L6)."""

nisse-core/src/nisse_core/log/views.py

@@ -20,7 +20,7 @@ ranges still line up with the log. Everything not the viewer's own becomes a
 from collections.abc import Mapping
 from dataclasses import replace
 
-from nisse_core.canonical import (
+from nisse_core.log.canonical import (
     ActionVerdict,
     AssistantText,
     DelegationCompleted,

nisse-core/src/nisse_core/models/__init__.py

@@ -0,0 +1,2 @@
+"""Talking to models (architecture L9): the client, what the models are, and
+the request bodies built from the log (L7)."""

nisse-core/src/nisse_core/models/catalog.py

@@ -19,7 +19,7 @@ from typing import Any
 
 import httpx
 
-from nisse_core.llm import LLMClient, LLMError
+from nisse_core.models.llm import LLMClient, LLMError
 
 # LiteLLM's provider name for models we have a vendor-native adapter for.
 ANTHROPIC_PROVIDER = "anthropic"

nisse-core/src/nisse_core/models/context.py

@@ -28,7 +28,7 @@ from collections.abc import Callable, Sequence
 from dataclasses import dataclass, replace
 from typing import Any
 
-from nisse_core.canonical import (
+from nisse_core.log.canonical import (
     AssistantText,
     CanonicalEvent,
     DelegationStarted,
@@ -38,7 +38,7 @@ from nisse_core.canonical import (
     ToolResult,
     UserMessage,
 )
-from nisse_core.llm import DEFAULT_MAX_OUTPUT_TOKENS, LLMError
+from nisse_core.models.llm import DEFAULT_MAX_OUTPUT_TOKENS, LLMError
 
 # English and code run at about four characters a token. Three is deliberately
 # pessimistic, before any correction from what the server reports.

nisse-core/src/nisse_core/models/llm.py

@@ -17,15 +17,15 @@ from typing import Any, ClassVar, Protocol
 
 import httpx
 
-from nisse_core.canonical import (
+from nisse_core.log.canonical import (
     AssistantText,
     CanonicalEvent,
     StoredEvent,
     ToolCall,
     mint_call_id,
 )
-from nisse_core.render import OpenAIChatRenderer
-from nisse_core.render_anthropic import AnthropicMessagesRenderer
+from nisse_core.models.render import OpenAIChatRenderer
+from nisse_core.models.render_anthropic import AnthropicMessagesRenderer
 
 
 class LLMError(RuntimeError):

nisse-core/src/nisse_core/models/probe.py

@@ -24,7 +24,7 @@ from dataclasses import dataclass
 
 import httpx
 
-from nisse_core.catalog import ModelFacts
+from nisse_core.models.catalog import ModelFacts
 
 # Of a probe's size, how much must come back for it to count as uncut. A
 # chat template adds a little, and a tokenizer counts a little differently.

nisse-core/src/nisse_core/models/render.py

@@ -11,7 +11,7 @@ this skeleton.
 import json
 from typing import Any
 
-from nisse_core.canonical import (
+from nisse_core.log.canonical import (
     AssistantText,
     Compaction,
     DelegationCompleted,

nisse-core/src/nisse_core/models/render_anthropic.py

@@ -18,7 +18,7 @@ route and this renderer builds the vendor-native body.
 
 from typing import Any
 
-from nisse_core.canonical import (
+from nisse_core.log.canonical import (
     AssistantText,
     Compaction,
     DelegationCompleted,
@@ -31,7 +31,7 @@ from nisse_core.canonical import (
     ToolResult,
     UserMessage,
 )
-from nisse_core.render import effective_events
+from nisse_core.models.render import effective_events
 
 DEFAULT_MAX_TOKENS = 4096
 NO_RESULT_PLACEHOLDER = "[no result recorded]"

nisse-core/src/nisse_core/state/__init__.py

@@ -0,0 +1,2 @@
+"""Durability (architecture L8): the state store, reconciliation, the worker
+thread and the kill switch."""

nisse-core/src/nisse_core/state/cancel.py

@@ -8,8 +8,8 @@ record of a stop is written by the worker when the stopped work returns.
 
 import threading
 
-from nisse_core.canonical import ToolCall, ToolResult
-from nisse_core.state import Transcript
+from nisse_core.log.canonical import ToolCall, ToolResult
+from nisse_core.state.store import Transcript
 
 STOPPED_RESULT = "not run: the operator stopped the agent"
 

nisse-core/src/nisse_core/state/dump.py

@@ -1,6 +1,6 @@
 """A readable dump of a state store, for bring-up runs and debugging.
 
-    python -m nisse_core.dump STATE_DB [--full]
+    python -m nisse_core.state.dump STATE_DB [--full]
 
 Opens the store read-only (SQLite `mode=ro`), so it is safe to run against a
 live agent's database: it never migrates, writes or locks it. Prints the work

nisse-core/src/nisse_core/state/reconcile.py

@@ -11,7 +11,7 @@ from collections.abc import Callable, Mapping
 from dataclasses import dataclass
 from typing import Literal
 
-from nisse_core.state import Intent, StateStore
+from nisse_core.state.store import Intent, StateStore
 
 # Returns True if the intent's external effect verifiably happened (do not
 # repeat it), False if it verifiably never started (safe to discard), or

nisse-core/src/nisse_core/state/store.py

@@ -6,7 +6,7 @@ intent-log write is a single immediately-committed statement — in-memory state
 that hasn't hit disk is assumed lost.
 
 The intent log records a unit of work *before* it starts, so a restart knows
-what was in flight (see ``nisse_core.reconcile`` for how pending intents are
+what was in flight (see ``nisse_core.state.reconcile`` for how pending intents are
 resolved against external truth on boot).
 """
 
@@ -18,7 +18,7 @@ from datetime import UTC, datetime
 from pathlib import Path
 from typing import Any, Self
 
-from nisse_core.canonical import (
+from nisse_core.log.canonical import (
     CanonicalEvent,
     StoredEvent,
     UserMessage,
@@ -26,8 +26,8 @@ from nisse_core.canonical import (
     from_payload,
     to_payload,
 )
+from nisse_core.log.views import project
 from nisse_core.oplog import OpLog
-from nisse_core.views import project
 
 # The tables every agent has. An agent's own tables come on top: a subclass
 # names them, the version they bring, and the migrations to it.