feat(#20): rewire the web turn surface onto the wt adapter (slice-2, part 2b-ii)

The Starlette endpoints (create / stream / cancel / tools / messages) now go through
ratatoskr.wt over the worldtree-sdk; the browser contract is preserved. This is the
last consumer of the hand-rolled turn-stream family — after this, stream_turn* /
cancel_turn are orphaned and get deleted in part 2b-iii (with the live smoke).

- _wt_client wraps a client_factory transport as the adapter's WorldtreeClient
  (INV-CUT-1), reading base_url + bearer off the transport (a no-auth test transport
  falls back to a placeholder key). The hand-rolled endpoints (persona / agents /
  admin / bifrost) keep using the raw transport until their slices.
- _event_to_browser_payload derives the browser payload from the SDK's `raw` (the
  wire body) minus the redundant `type`, plus the composite `sse_id` string — the
  SAME shape the old dataclasses produced, so the presentation fixture + browser JS
  are unchanged; the browser event_type is the wire `type`, not the SDK class name.
- The stream endpoint captures the upstream cancel target from the composite sse_id
  (the SDK's top-level turn_id is body-derived, absent on text frames); create reads
  the SDK's open create dict; cancel reads CancelResult.cancelled and surfaces a
  generic 502 for CancelFailed (the SDK abstracts the upstream cancel HTTP status).
- test_web_presentation_contract builds SDK events via build_event; two cancel tests
  adopt the SDK's (status, error_code) race pairs + the 502.

Suite 570 green; web/server.py + presentation test ruff-clean, mypy unchanged
(same pre-existing errors). Patch (internal; browser contract preserved).
This commit is contained in:
2026-07-19 06:22:13 -07:00
parent e3a10ad80e
commit 5c595b862d
5 changed files with 126 additions and 110 deletions
+1 -1
View File
@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
[project]
name = "ratatoskr"
version = "0.21.6"
version = "0.21.7"
description = "Worldtree Conversation API debug console (web + headless CLI) — multi-pane observability"
readme = "README.md"
requires-python = ">=3.12"
+83 -55
View File
@@ -26,8 +26,10 @@ from starlette.responses import (
)
from starlette.routing import Mount, Route
from starlette.staticfiles import StaticFiles
from worldtree_sdk import CancelledEvent, DoneEvent, ErrorEvent, WorldtreeClient
from ratatoskr import local_agents as _local_agents
from ratatoskr import wt
from ratatoskr.first_message import seed_preset_first_message
from ratatoskr.sessions import (
AgentNotAvailable,
@@ -38,33 +40,44 @@ from ratatoskr.sessions import (
BifrostHandshakeFailed,
PersonaNotConfigured,
SessionApiFailed,
create_session,
endpoint_for_plane,
get_persona_state,
get_session_bifrost,
get_session_messages,
get_session_tools,
list_agents,
)
# The turn path (create / stream / cancel / tools / messages) is served by the
# worldtree-sdk adapter (`wt.*`), which raises ratatoskr's caller-semantic
# exceptions (DEC-2). The hand-rolled endpoints (persona / agents / admin /
# bifrost) stay on the `sessions` / `sse_client` wrappers until their own slices.
from ratatoskr.sse_client import (
AdminEvent,
CancelAlreadyCompleted,
CancelFailed,
Cancelled,
CancelTurnNotFound,
Done,
Error,
MalformedSseData,
MalformedSseId,
SseConnectFailed,
SseConnectionDropped,
TurnIdFlip,
cancel_turn,
stream_admin_events,
stream_turn_resilient,
)
def _wt_client(client: httpx.AsyncClient, *, max_reconnects: int = 5) -> WorldtreeClient:
"""Wrap a client_factory transport as the adapter's WorldtreeClient (INV-CUT-1:
the SDK never closes it). base_url + bearer are read off the transport (the
factory bakes them in); the SDK re-applies auth per request, so the extracted
key just mirrors the transport's default. A no-auth test transport falls back to
a placeholder key (respx ignores auth)."""
base_url = str(client.base_url) or "http://localhost"
header = client.headers.get("Authorization", "")
api_key = header[len("Bearer "):].strip() if header.startswith("Bearer ") else ""
return wt.build_client(
base_url, api_key=api_key or "ratatoskr", transport=client, max_reconnects=max_reconnects
)
def _static_dir() -> str:
"""Locate the bundled static/ directory inside the installed package.
@@ -163,16 +176,17 @@ async def _create_session_endpoint(request: Request) -> JSONResponse:
try:
async with client_factory() as client:
info = await create_session(
client,
info = await wt.create_session(
_wt_client(client),
agent_id,
end_user_id=end_user_id,
bifrost=bifrost,
consumer_key=consumer_key if bifrost else None,
)
# #347 authored first-message: seed the agent's preset opening
# (best-effort; never blocks create — see first_message INV-001).
await seed_preset_first_message(client, info.session_id, agent_id)
# #347 authored first-message: seed the agent's preset opening (best-effort;
# never blocks create). first_message is a slice-3 hand-rolled path — it
# reuses the raw transport (its default bearer), not the adapter client.
await seed_preset_first_message(client, info["session_id"], agent_id)
except AgentNotFound:
return JSONResponse({"error_code": "agent_not_found"}, status_code=404)
except BifrostConsumerKeyMissing:
@@ -188,12 +202,13 @@ async def _create_session_endpoint(request: Request) -> JSONResponse:
},
status_code=502,
)
except SessionApiFailed as exc:
except wt.SessionApiFailed as exc:
return JSONResponse(
{"error_code": "session_api_failed", "status": exc.status},
status_code=exc.status,
)
payload = _as_dict(info)
# The adapter returns the SDK's open-world create dict; the browser reads it as-is.
payload = dict(info)
if bifrost is not None:
# Bound-state for the UI indicator — plane + endpoint only, never the key.
payload["bifrost"] = {
@@ -251,25 +266,19 @@ async def _submit_turn_endpoint(request: Request) -> JSONResponse:
def _event_to_browser_payload(event: object) -> tuple[str, dict]:
"""Serialize an upstream Event dataclass to (browser_event_type, json_dict).
"""Serialize an SDK `TurnEvent` to (browser_event_type, json_dict).
Per INV-008 + FN stream_turn_endpoint STEP 3. The dict shape is
locked by tests/fixtures/presentation_contract.json — one entry per
Event type. Implementation: snake_case class name as event_type;
asdict(event) with sse_id flattened to "T:S" string.
Per INV-008 + FN stream_turn_endpoint STEP 3. The browser contract
(tests/fixtures/presentation_contract.json) is preserved: the SDK's `raw` is
the wire body — the same per-type field set the old dataclasses carried — so the
payload is `raw` minus the redundant `type`, plus the composite `sse_id` string
(already "T:S"). The browser event_type is the wire `type` ("text" / "done" /
…), NOT the SDK class name. Open-world: additive server fields pass through.
"""
type_name = type(event).__name__
# CamelCase → snake_case
browser_type = "".join(
("_" + c.lower() if c.isupper() and i else c.lower())
for i, c in enumerate(type_name)
)
data = asdict(event) # type: ignore[arg-type]
sse_id = data.get("sse_id")
if isinstance(sse_id, (list, tuple)) and len(sse_id) == 2:
data["sse_id"] = f"{sse_id[0]}:{sse_id[1]}"
elif isinstance(sse_id, dict) and "turn_id" in sse_id and "seq" in sse_id:
data["sse_id"] = f"{sse_id['turn_id']}:{sse_id['seq']}"
browser_type = getattr(event, "type", "") or ""
raw = getattr(event, "raw", None) or {}
data = {k: v for k, v in dict(raw).items() if k != "type"}
data["sse_id"] = getattr(event, "sse_id", None)
return browser_type, data
@@ -281,6 +290,20 @@ def _format_sse(event_type: str, data: dict) -> bytes:
return f"event: {event_type}\ndata: {json.dumps(data)}\n\n".encode()
def _turn_id_from_sse_id(sse_id: object) -> int | None:
"""The turn component of the SDK's composite sse_id (`"{turn}:{seq}"`) — the
upstream cancel target, present on every frame (the SDK's top-level `turn_id` is
the body field, absent on text/thinking events)."""
if not isinstance(sse_id, str):
return None
head, _, _ = sse_id.partition(":")
try:
turn = int(head)
except ValueError:
return None
return turn if turn > 0 else None
async def _stream_turn_endpoint(request: Request) -> StreamingResponse:
"""GET /api/turns/{session_id}/stream?turn_id=N → proxy upstream SSE.
@@ -302,21 +325,22 @@ async def _stream_turn_endpoint(request: Request) -> StreamingResponse:
async def gen() -> AsyncIterator[bytes]:
client = client_factory()
wt_client = _wt_client(client)
try:
handle.status = "streaming"
try:
async for event in stream_turn_resilient(client, session_id, handle.content):
# v0.16.0: capture the upstream (Worldtree-assigned)
# turn_id from the first event so cancel paths target
# the real upstream turn, not our local counter.
async for event in wt.stream_turn(wt_client, session_id, handle.content):
# v0.16.0: capture the upstream (Worldtree-assigned) turn_id from
# the first event so cancel paths target the real upstream turn,
# not our local counter — parsed from the composite sse_id.
if handle.upstream_turn_id is None:
sse_id = getattr(event, "sse_id", None)
if sse_id is not None:
handle.upstream_turn_id = sse_id.turn_id
handle.upstream_turn_id = _turn_id_from_sse_id(
getattr(event, "sse_id", None)
)
event_type, data = _event_to_browser_payload(event)
yield _format_sse(event_type, data)
if isinstance(event, (Done, Error, Cancelled)):
handle.status = type(event).__name__.lower()
if isinstance(event, (DoneEvent, ErrorEvent, CancelledEvent)):
handle.status = event.type or "done"
break
except (SseConnectFailed, SseConnectionDropped, MalformedSseId,
MalformedSseData, TurnIdFlip) as exc:
@@ -330,7 +354,7 @@ async def _stream_turn_endpoint(request: Request) -> StreamingResponse:
# turn (if it started) — never the local turn_id.
if handle.status == "streaming" and handle.upstream_turn_id is not None:
try:
await cancel_turn(client, session_id, handle.upstream_turn_id)
await wt.cancel_turn(wt_client, session_id, handle.upstream_turn_id)
except (CancelAlreadyCompleted, CancelTurnNotFound):
pass # cooperative race — turn already terminal upstream
except Exception as exc:
@@ -377,15 +401,18 @@ async def _cancel_turn_endpoint(request: Request) -> JSONResponse:
client_factory = request.app.state.client_factory
try:
async with client_factory() as client:
await cancel_turn(client, session_id, handle.upstream_turn_id)
body = {"cancelled": True}
result = await wt.cancel_turn(
_wt_client(client), session_id, handle.upstream_turn_id
)
body = {"cancelled": bool(result.cancelled)}
except (CancelAlreadyCompleted, CancelTurnNotFound):
body = {"cancelled": False, "reason": "race_or_completed"}
except CancelFailed as exc:
except CancelFailed:
# The SDK abstracts the upstream cancel HTTP status; surface a generic 502.
registry.pop((session_id, turn_id), None)
return JSONResponse(
{"error_code": "cancel_failed", "status": exc.status},
status_code=exc.status,
{"error_code": "cancel_failed"},
status_code=502,
)
registry.pop((session_id, turn_id), None)
return JSONResponse(body, status_code=200)
@@ -465,13 +492,13 @@ async def _session_tools_endpoint(request: Request) -> JSONResponse:
client_factory = request.app.state.client_factory
try:
async with client_factory() as client:
info = await get_session_tools(client, session_id)
except SessionApiFailed as exc:
info = await wt.get_session_tools(_wt_client(client), session_id)
except wt.SessionApiFailed as exc:
return JSONResponse(
{"error_code": "session_tools_unavailable", "status": exc.status},
status_code=exc.status,
)
return JSONResponse(info, status_code=200)
return JSONResponse(dict(info), status_code=200)
async def _session_messages_endpoint(request: Request) -> JSONResponse:
@@ -485,13 +512,13 @@ async def _session_messages_endpoint(request: Request) -> JSONResponse:
client_factory = request.app.state.client_factory
try:
async with client_factory() as client:
data = await get_session_messages(client, session_id)
except SessionApiFailed as exc:
data = await wt.get_session_messages(_wt_client(client), session_id)
except wt.SessionApiFailed as exc:
return JSONResponse(
{"error_code": "session_messages_unavailable", "status": exc.status},
status_code=exc.status,
)
return JSONResponse(data, status_code=200)
return JSONResponse(dict(data), status_code=200)
async def _session_bifrost_endpoint(request: Request) -> JSONResponse:
@@ -609,14 +636,15 @@ def create_app(
]
if in_flight:
client = client_factory()
wt_client = _wt_client(client)
try:
task_to_handle = {
asyncio.create_task(
cancel_turn(client, h.session_id, h.upstream_turn_id)
wt.cancel_turn(wt_client, h.session_id, h.upstream_turn_id)
): h
for h in in_flight
}
done, pending = await asyncio.wait(task_to_handle, timeout=5.0)
_done, pending = await asyncio.wait(task_to_handle, timeout=5.0)
# Per-pending session/turn detail (INV-006 logging fidelity).
for task in pending:
h = task_to_handle[task]
+32 -49
View File
@@ -7,6 +7,12 @@ type. Server-side serialization (`_event_to_browser_payload`) is
unit-tested against the fixture. JS-side rendering in
`src/ratatoskr/web/static/index.html` consumes the same shape — if
this fixture changes, both sides update in lockstep.
Post worldtree-sdk cutover (#20): the presenter consumes SDK `TurnEvent`s.
`_event_to_browser_payload` derives the browser payload from the SDK's `raw`
(the wire body) plus the composite `sse_id` string — the SAME shape the old
dataclasses produced, so the fixture is unchanged. These events are built via
the SDK's own `build_event` from the wire body.
"""
from __future__ import annotations
@@ -14,20 +20,8 @@ from __future__ import annotations
import json
from pathlib import Path
from ratatoskr.sse_client import (
AffectUpdate,
AwaitingLlmFirstToken,
Cancelled,
Done,
Error,
SseId,
Text,
TextBoundary,
Thinking,
ToolResult,
ToolStart,
WorkerPhase,
)
from worldtree_sdk.events import build_event
from ratatoskr.web.server import _event_to_browser_payload
@@ -36,6 +30,13 @@ def _load_fixture() -> dict:
return json.loads(path.read_text())
def _ev(ev_type: str, sse_id: str, **fields: object) -> object:
"""Build an SDK TurnEvent from its wire body (raw includes `type`); turn_id is
the turn component of the composite sse_id."""
turn = int(sse_id.split(":", 1)[0])
return build_event(ev_type, sse_id, turn, {"type": ev_type, **fields})
def _check(name: str, event: object) -> None:
"""Assert (event_type, data) for `event` matches the fixture entry."""
fixture = _load_fixture()
@@ -51,58 +52,40 @@ def _check(name: str, event: object) -> None:
def test_worker_phase_matches_fixture() -> None:
_check(
"worker_phase",
WorkerPhase(sse_id=SseId(42, 3), phase="BuildingPrompt", turn_id=42),
)
_check("worker_phase", _ev("worker_phase", "42:3", phase="BuildingPrompt", turn_id=42))
def test_thinking_matches_fixture() -> None:
_check(
"thinking",
Thinking(sse_id=SseId(42, 5), content="Let me think..."),
)
_check("thinking", _ev("thinking", "42:5", content="Let me think..."))
def test_text_matches_fixture() -> None:
_check(
"text",
Text(sse_id=SseId(42, 7), content="Hello there"),
)
_check("text", _ev("text", "42:7", content="Hello there"))
def test_text_boundary_matches_fixture() -> None:
_check(
"text_boundary",
TextBoundary(
sse_id=SseId(42, 8), kind="sentence",
char_offset=11, ts="2026-05-28T00:00:00Z",
),
_ev("text_boundary", "42:8", kind="sentence", char_offset=11, ts="2026-05-28T00:00:00Z"),
)
def test_tool_start_matches_fixture() -> None:
_check(
"tool_start",
ToolStart(sse_id=SseId(42, 9), name="search", arguments={"q": "ratatoskr"}),
)
_check("tool_start", _ev("tool_start", "42:9", name="search", arguments={"q": "ratatoskr"}))
def test_tool_result_matches_fixture() -> None:
_check(
"tool_result",
ToolResult(
sse_id=SseId(42, 10), name="search",
result={"n": 1}, duration_ms=12,
),
_ev("tool_result", "42:10", name="search", result={"n": 1}, duration_ms=12),
)
def test_done_matches_fixture() -> None:
_check(
"done",
Done(
sse_id=SseId(42, 11), phase="succeeded", response="Hello there",
_ev(
"done", "42:11", phase="succeeded", response="Hello there",
model="qwen3.6-35-a3b", duration_ms=1234,
usage={
"prompt_tokens": 100, "completion_tokens": 50,
@@ -115,8 +98,8 @@ def test_done_matches_fixture() -> None:
def test_error_matches_fixture() -> None:
_check(
"error",
Error(
sse_id=SseId(42, 11), phase="failed",
_ev(
"error", "42:11", phase="failed",
message="llm output invalid", error_code="llm_output_invalid",
),
)
@@ -125,8 +108,8 @@ def test_error_matches_fixture() -> None:
def test_cancelled_matches_fixture() -> None:
_check(
"cancelled",
Cancelled(
sse_id=SseId(42, 11), phase="cancelled", turn_id=42,
_ev(
"cancelled", "42:11", phase="cancelled", turn_id=42,
reason="user_cancel", partial_message_id=None,
),
)
@@ -135,8 +118,8 @@ def test_cancelled_matches_fixture() -> None:
def test_affect_update_matches_fixture() -> None:
_check(
"affect_update",
AffectUpdate(
sse_id=SseId(42, 1), status="current", turn_id=42,
_ev(
"affect_update", "42:1", status="current", turn_id=42,
snapshot={
"agent_id": "mimir",
"pad": {"pleasure": 0.52, "arousal": 0.47, "dominance": 0.50},
@@ -155,8 +138,8 @@ def test_affect_update_matches_fixture() -> None:
def test_awaiting_llm_first_token_matches_fixture() -> None:
_check(
"awaiting_llm_first_token",
AwaitingLlmFirstToken(
sse_id=SseId(42, 2), turn_id=42,
elapsed_ms_since_building_prompt=5012.3,
_ev(
"awaiting_llm_first_token", "42:2",
turn_id=42, elapsed_ms_since_building_prompt=5012.3,
),
)
+9 -4
View File
@@ -456,15 +456,16 @@ class TestCancelTurnEndpoint:
@respx.mock
def test_already_completed_race(self) -> None:
"""already_completed [race]: upstream 409 → 200 reason=race_or_completed."""
"""already_completed [race]: upstream 409 turn_finished → 200 reason=race_or_completed."""
from ratatoskr.web.server import create_app
app = create_app(_mock_client_factory())
c = TestClient(app)
turn_id = c.post("/api/turns/s-1", json={"content": "hi"}).json()["turn_id"]
app.state.turn_registry[("s-1", turn_id)].status = "streaming"
app.state.turn_registry[("s-1", turn_id)].upstream_turn_id = 42
# SDK gates the race on the (status, error_code) pair (B-CAN-3).
respx.post("https://w.example/sessions/s-1/turns/42/cancel").mock(
return_value=httpx.Response(409)
return_value=httpx.Response(409, json={"error_code": "turn_finished"})
)
resp = c.post(f"/api/turns/s-1/cancel?turn_id={turn_id}")
assert resp.status_code == 200
@@ -473,7 +474,11 @@ class TestCancelTurnEndpoint:
@respx.mock
def test_cancel_failed_500(self) -> None:
"""cancel_failed [error]: upstream 500 → 500 with cancel_failed envelope."""
"""cancel_failed [error]: upstream 500 → 502 cancel_failed envelope.
Post-cutover: the SDK abstracts the upstream cancel HTTP status behind a
typed CancelFailed, so the endpoint surfaces a generic 502 (bad gateway)
rather than echoing the upstream 500."""
from ratatoskr.web.server import create_app
app = create_app(_mock_client_factory())
c = TestClient(app)
@@ -484,7 +489,7 @@ class TestCancelTurnEndpoint:
return_value=httpx.Response(500, content=b"boom")
)
resp = c.post(f"/api/turns/s-1/cancel?turn_id={turn_id}")
assert resp.status_code == 500
assert resp.status_code == 502
assert resp.json()["error_code"] == "cancel_failed"
assert ("s-1", turn_id) not in app.state.turn_registry
Generated
+1 -1
View File
@@ -472,7 +472,7 @@ wheels = [
[[package]]
name = "ratatoskr"
version = "0.21.6"
version = "0.21.7"
source = { editable = "." }
dependencies = [
{ name = "httpx" },