Raven-1/tests/test_agent_loop_part_durations.py
Handsome-wzw 7b45527b5b fix(tools): an array sent as its own JSON is still an array
## Summary

Reported as "deliver_files is broken". It is not: **the tool never ran.**

Every `deliver_files` call on one install was refused before dispatch -- five in a row in a single session -- and nothing was ever delivered:

```
Error: Invalid parameters for tool 'deliver_files': files should be array
```

What the model sent was the array, spelled as a string:

```json
{"files": "[{\"path\": \"...poster.png\", \"title\": \"...\"}]", "message": "..."}
```

`cast_params` already forgives that spelling for every scalar type -- `"3"` for an integer, `"true"` for a boolean -- and did not for arrays or objects, which are exactly the types a model is most likely to stringify.

## Type

- [x] Fix
- [ ] Feature

## The edge of the forgiving is the design

When the value is a string, try `json.loads`, and take the result **only if it is the type the schema asked for**. A string that is not JSON, or is JSON of the wrong shape, is returned unchanged so the validator still names the real problem -- the point is to widen what is accepted, not to swallow a type error.

**Nothing here can raise.** Turning "your argument has the wrong type" into a traceback would be worse than the refusal it replaces, so the parse is wrapped and failure falls through.

**A `type: "string"` field is never parsed.** JSON is ordinary text in a message body or a file's content, and replacing it would hand the tool something it never asked for. Two independent things prevent it today (the string branch returns before the parse is reached, and the parse only wins on a type match), which is why the case pinning it is a tripwire rather than a load-bearing assertion -- said so in its own docstring rather than left to look like coverage.

## Not just deliver_files

Six built-in tools declare an array or object parameter and were all exposed:

| tool | parameter |
|---|---|
| `deliver_files` | `files` |
| `ask_user` | `questions` |
| `image_generate` | `images` |
| `message` | `media` |
| `create_playbook` | `skills` |
| `video_generate` | `params` |

The same failure was already seen from the other side: `acp_client/acp_agent.py` records `update_task_state` and `operations should be array` from a real dispatch.

## Verification

```
make lint-python    ruff check + ruff format --check, 1850 files
make lint-imports   9 contracts kept, 0 broken
pytest tests/       19,699 passed, 7 failed (all pre-existing, see below)
```

- [x] Relevant tests pass locally
- [x] Relevant lint / type checks pass locally
- [ ] User-facing docs or screenshots -- no user-facing string

**End to end, through the real tool**, with the exact payload that failed five times in production:

```
files sent as:            str -> [{"path": ".../poster.png", "title": "..."}]
returned:                 Delivered 1 file(s): poster.png (72 B).
still "should be array":  False
registered in the store:  True
```

Twelve cases, five mutations, each killing at least one:

| mutation | red |
|---|---|
| drop the whole branch (the old behaviour) | 3 |
| raise instead of falling through on a parse failure | 5 |
| accept the parse whatever type it yielded | 1 |
| do not re-cast the parsed value, so items are not cast | 1 |
| widen the tuple to include `string` | **survived -- unreachable** |

The last is worth stating rather than hiding: that mutation cannot change behaviour, because the `type == "string"` branch returns earlier and the type check would refuse the result anyway. I also tried moving the new block above that branch, alone and combined with widening the tuple, and neither reddens anything either. The invariant holds by construction; the case that documents it says so.

**Pre-existing failures, checked against the base.** 7 failed in the full run, all in `tests/test_security_network.py` (`test_a_redirect_to_a_dangerous_address_is_refused` and siblings). Running that file alone on this same tree gives 4 -- the count is order-dependent -- and the change touches only `raven/agent/tools/params.py` plus a new test file, neither of which `raven/security/network.py` imports. The failures are the box's resolver: it returns poisoned answers roughly half the time, which I measured separately (`en.wikipedia.org` came back as `2001::1` on 3 of 6 lookups). Not introduced here, and worth someone's attention on its own.

## Risk

- [x] Security impact considered -- widens what reaches a tool's own validation, not what bypasses it. A parsed value is still validated against the same schema, and a parse that yields the wrong type is refused exactly as before
- [x] Backward compatibility considered -- a call that worked before takes the identical path; the new branch is reached only by a `str` where an array or object is declared, which previously always failed
- [x] Rollback path is clear -- single-commit revert

## Related Issues

N/A

## Swept

`mr-review-patterns` pre-submit sweep, over `git diff origin/main...HEAD`.

- **1 (a write location moved)** -- nothing moved; a branch was added to an existing dispatch on `target_type`. Grepped every caller of `cast_params`: the registry is the only one, and it runs `validate_params` right after, which is the reader this affects.
- **2 (a value with several entry points)** -- the cast is the single funnel every tool argument passes through, which is why the fix belongs here and not in `deliver_files`.
- **3 (a guard)** -- checked both directions: what it wrongly lets through (JSON of the wrong type, non-JSON text -- both refused, with cases) and what it wrongly refuses (a real list, unaffected, with a case).
- **4 (tests)** -- five mutations above, one honestly reported as unreachable. Failing-ID sets diffed against the base rather than counts.
- **6 (stale text)** -- the module docstring enumerated what is coerced ("a string that should be an integer"); it now names arrays and objects too. Grepped the tracked tree for other statements of the rule: none.
- **9 (error handling)** -- the new path adds one way to end and it is caught: `json.loads` raises `ValueError` on bad text and `TypeError` on a non-string, and both fall through to the untouched value.
- **11** -- `git ls-files | grep -c node_modules` is 0; the commit-message gate was run, not eyeballed.

Not applicable, with reasons: **3b, 3c, 3d, 3e** (no tool schema produced, no state moved, one lane, no assertion lost its symbol); **4b, 4c, 4d** (no platform branch, no shell, no rendered number); **5, 5b** (no dependency; `json` is stdlib and already imported across this package); **6b, 6c** (no backup, no generated artifact); **7, 7b** (nothing on disk); **8** (no asset, no secret); **10** (no prompt text).

## Deliberately not fixed

The error text still ends with `[Analyze the error above and try a different approach.]`, and the model plainly did not -- it sent the same shape five times. A message naming the expected spelling (`files must be a JSON array, not a string containing one`) would give it something to act on. That is a change to the registry's error surface, affects every tool, and deserves its own argument.

## Moved from !594

Opened on a compliant source branch. AGENTS.md 2.1 asks for a 3-5 word snake_case description and the previous branch had seven; GitLab cannot change an existing merge request's source branch, so the change moves rather than the name.

Same commits, rebased onto `c2a9768f`. Both code findings from the review on !594 are already in them -- `RecursionError` caught with its regression case, and the implementation comment reduced to the two non-obvious reasons. Nothing else changed, so that review carries.
2026-09-09 15:42:36 +00:00

187 lines
7.1 KiB
Python

"""A turn's parts carry how long they took, from the stream to a restored view.
How long the model thought and how long each tool ran were measured only in the
browser that watched the stream, so switching sessions or reloading the page
threw both numbers away -- the restored transcript had to either show nothing or
invent a clock that started at page load. These tests pin the two spans as
stored properties of the entries they belong to, and pin that an entry written
before they existed reads as *unknown* rather than as zero.
"""
from __future__ import annotations
import asyncio
import json
import tempfile
from pathlib import Path
from typing import Any
import pytest
from raven.agent.loop import AgentLoop
from raven.agent.loop.bundles import ToolWiring, TurnPolicy
from raven.providers.base import ChatDelta, LLMProvider, LLMResponse, ToolCallRequest
from raven.rpc.methods.session import _map_to_wire
from raven.rpc.models import TranscriptMessage
from raven.spine.message import ChatType, Source
from raven.spine.turn import Origin, TurnRequest
# Long enough that a real clock separates the deltas, short enough to stay a
# unit test. Assertions use half of it so a loaded machine cannot flake them.
GAP_S = 0.12
FLOOR_MS = int(GAP_S * 1000 / 2)
class ThinkingStreamProvider(LLMProvider):
"""Streams a thought, waits, then emits the scripted call or answer.
One script entry per model call. Each is ``(thought, content, tool_calls)``;
the wait between the thought and what follows is what the reasoning clock
has to measure.
"""
def __init__(self, script: list[tuple[str, str, list[ToolCallRequest]]]) -> None:
super().__init__(api_key="test")
self._script = list(script)
async def chat_stream(self, **kwargs: Any):
thought, content, tool_calls = self._script.pop(0)
for piece in thought:
yield ChatDelta(content=None, reasoning_content=piece)
await asyncio.sleep(GAP_S)
for call in tool_calls:
yield ChatDelta(
content=None,
tool_call_delta={
"tool_calls": [
{
"index": 0,
"id": call.id,
"function": {"name": call.name, "arguments": json.dumps(call.arguments)},
}
]
},
)
if content:
yield ChatDelta(content=content)
yield ChatDelta(content=None, finish_reason="tool_calls" if tool_calls else "stop")
async def chat(self, messages, tools=None, model=None, **kwargs: Any):
return LLMResponse(content="unused", finish_reason="stop")
def get_default_model(self) -> str:
return "stub"
@pytest.fixture
def workspace():
with tempfile.TemporaryDirectory() as td:
yield Path(td)
def _make_msg(text: str = "write it") -> TurnRequest:
return TurnRequest(
origin=Origin.USER,
source=Source(channel="tui", chat_id="chat1", sender_id="user", chat_type=ChatType.DM),
text=text,
)
def _persisted(workspace: Path) -> list[dict[str, Any]]:
path = workspace / "sessions" / "tui" / "chat1.jsonl"
assert path.exists(), "session file was not persisted"
records = [json.loads(line) for line in path.read_text(encoding="utf-8").splitlines() if line.strip()]
return [r for r in records if r.get("_type") != "metadata"]
async def _run_a_thinking_turn_with_a_slow_tool(workspace: Path) -> list[dict[str, Any]]:
provider = ThinkingStreamProvider(
[
(
"let me write the file",
"",
[ToolCallRequest(id="c1", name="write_file", arguments={"path": "a.txt", "content": "one\n"})],
),
("now to report", "done", []),
]
)
agent = AgentLoop(
provider=provider,
workspace=workspace,
model="stub",
policy=TurnPolicy(max_iterations=3),
tools=ToolWiring(restrict_to_workspace=True),
)
real_execute = agent.tools.execute
async def slow_execute(name, arguments, **kwargs):
await asyncio.sleep(GAP_S)
return await real_execute(name, arguments, **kwargs)
agent.tools.execute = slow_execute
async def sink(_text: str) -> None:
return None
out = await agent._process_message(_make_msg(), on_token_delta=sink, on_reasoning_delta=sink)
assert out is not None
return _persisted(workspace)
@pytest.mark.asyncio
async def test_a_thought_is_stored_with_how_long_it_took(workspace):
msgs = await _run_a_thinking_turn_with_a_slow_tool(workspace)
thinking = [m for m in msgs if m.get("role") == "assistant" and m.get("reasoning_content")]
assert thinking, f"no assistant entry carried a thought: {msgs}"
for entry in thinking:
assert "_reasoning_ms" not in entry, f"the in-flight private key was stored: {entry}"
assert entry.get("reasoning_ms", 0) >= FLOOR_MS, f"thought span missing or too small: {entry}"
@pytest.mark.asyncio
async def test_a_tool_result_is_stored_with_how_long_the_call_ran(workspace):
msgs = await _run_a_thinking_turn_with_a_slow_tool(workspace)
tool_entry = next(m for m in msgs if m.get("role") == "tool")
assert "_duration_ms" not in tool_entry, f"the in-flight private key was stored: {tool_entry}"
assert tool_entry.get("duration_ms", 0) >= FLOOR_MS, f"call span missing or too small: {tool_entry}"
@pytest.mark.asyncio
async def test_both_spans_survive_a_restore(workspace):
"""The whole point: a resumed transcript shows the durations the live one did."""
msgs = await _run_a_thinking_turn_with_a_slow_tool(workspace)
wire = [TranscriptMessage.model_validate(e) for e in _map_to_wire(msgs, "tui:chat1")]
thought = next(m for m in wire if m.role == "assistant" and m.reasoning_content)
assert thought.reasoning_ms is not None and thought.reasoning_ms >= FLOOR_MS
tool = next(m for m in wire if m.role == "tool")
assert tool.duration_ms is not None and tool.duration_ms >= FLOOR_MS
def test_a_session_written_before_the_spans_existed_reads_as_unknown():
"""Absent must mean unknown, never zero.
Zero would tell the reader the model thought for no time and the tool
returned instantly -- a fabricated measurement, which is worse than the
bare header a client draws when the field is missing.
"""
stored = [
{"role": "user", "content": "hi", "timestamp": "2026-01-01T00:00:00"},
{
"role": "assistant",
"content": "",
"reasoning_content": "thinking hard",
"tool_calls": [{"id": "c1", "function": {"name": "read_file", "arguments": "{}"}}],
"timestamp": "2026-01-01T00:00:01",
},
{"role": "tool", "tool_call_id": "c1", "name": "read_file", "content": "ok"},
]
wire = _map_to_wire(stored, "tui:old")
for entry in wire:
assert "reasoning_ms" not in entry, f"invented a thought span: {entry}"
assert "duration_ms" not in entry, f"invented a call span: {entry}"
parsed = [TranscriptMessage.model_validate(e) for e in wire]
assert all(m.reasoning_ms is None and m.duration_ms is None for m in parsed)