1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
|
"""Tests for llm capability detection and feature fallbacks."""
from __future__ import annotations
import base64
from types import SimpleNamespace
from typing import Any
import pytest
from ollama import Message
from ollama._types import Image as OllamaImage
from ollama_web.llm import (
_embed_images_in_content,
_is_image_message,
_token_counts,
stream_chat_with_tools,
)
from ollama_web.tools.registry import ToolRegistry
def _image_message() -> Message:
return Message(
role="user",
content="hello",
images=[OllamaImage(value=b"fake-image-bytes")],
)
def test_is_image_message_detects_images() -> None:
assert _is_image_message(_image_message()) is True
def test_is_image_message_without_images() -> None:
assert _is_image_message(Message(role="user", content="hello")) is False
def test_embed_images_in_content_adds_base64_markdown() -> None:
embedded = _embed_images_in_content(_image_message())
encoded = base64.b64encode(b"fake-image-bytes").decode("ascii")
assert f"" in embedded.content
assert not getattr(embedded, "images", None)
def test_stream_chat_with_tools_disables_unsupported_capabilities(
monkeypatch: pytest.MonkeyPatch,
) -> None:
"""Unsupported tools and thinking options are not sent to Ollama."""
from ollama_web import llm
registry = ToolRegistry()
def lookup(query: str) -> str:
return query
registry.register(lookup)
class FakeClient:
def __init__(self) -> None:
self.calls: list[dict[str, Any]] = []
def chat(self, **kwargs: Any) -> list[SimpleNamespace]:
self.calls.append(kwargs)
return [SimpleNamespace(message={"content": "done"})]
client = FakeClient()
monkeypatch.setattr(llm, "get_client", lambda _host=None: client)
events = list(
stream_chat_with_tools(
[Message(role="user", content="hi")],
model="test-model",
think=True,
capabilities=set(),
registry=registry,
)
)
assert client.calls[0]["tools"] == []
assert client.calls[0]["think"] is None
assert client.calls[0]["stream"] is True
assert events[-1] == {"type": "done"}
def test_token_counts_extracts_from_dict() -> None:
assert _token_counts({"prompt_eval_count": 100, "eval_count": 50}) == {
"prompt_eval_count": 100,
"eval_count": 50,
}
def test_token_counts_extracts_from_object() -> None:
class FakeResp:
prompt_eval_count = 200
eval_count = 30
assert _token_counts(FakeResp()) == {
"prompt_eval_count": 200,
"eval_count": 30,
}
def test_token_counts_returns_none_when_missing() -> None:
assert _token_counts({}) is None
assert _token_counts(None) is None
def test_stream_chat_with_tools_disables_tools_for_final_answer_after_limit(
monkeypatch: pytest.MonkeyPatch,
) -> None:
from ollama_web import llm
executed: list[str] = []
def lookup(query: str) -> str:
"""Return a test lookup result."""
executed.append(query)
return "lookup result"
registry = ToolRegistry()
registry.register(lookup)
class FakeClient:
def __init__(self) -> None:
self.calls: list[dict[str, Any]] = []
def chat(self, **kwargs: Any) -> list[SimpleNamespace]:
self.calls.append(kwargs)
if len(self.calls) == 1:
return [
SimpleNamespace(
message={
"role": "assistant",
"content": "",
"tool_calls": [
{
"function": {
"name": "lookup",
"arguments": {"query": "test"},
}
}
],
}
)
]
return [SimpleNamespace(message={"content": "final answer"})]
client = FakeClient()
monkeypatch.setattr(llm, "get_client", lambda _host=None: client)
monkeypatch.setattr(llm, "MAX_TOOL_ROUNDS", 1)
events = list(
llm.stream_chat_with_tools(
[Message(role="user", content="research this")],
model="test-model",
registry=registry,
capabilities={"tools"},
language="en",
)
)
assert executed == ["test"]
assert len(client.calls) == 2
assert client.calls[0]["tools"]
assert client.calls[1]["tools"] == []
assert client.calls[0]["stream"] is True
assert client.calls[1]["stream"] is True
assert any(event.get("type") == "status" for event in events)
assert any(event == {"type": "delta", "content": "final answer"} for event in events)
assert events[-1] == {"type": "done"}
def test_stream_chat_with_tools_disables_mcp_tools_for_final_answer_after_limit(
monkeypatch: pytest.MonkeyPatch,
) -> None:
from ollama_web import llm
tool_name = "mcp__test__lookup"
executed: list[tuple[str, dict[str, object] | str]] = []
registry = ToolRegistry()
def executor(full_name: str, arguments: dict[str, object] | str = "") -> str:
executed.append((full_name, arguments))
return "mcp lookup result"
registry.register_mcp_tool(
tool_name,
{"type": "function", "function": {"name": tool_name}},
executor,
)
class FakeClient:
def __init__(self) -> None:
self.calls: list[dict[str, Any]] = []
def chat(self, **kwargs: Any) -> list[SimpleNamespace]:
self.calls.append(kwargs)
if len(self.calls) == 1:
return [
SimpleNamespace(
message={
"content": "",
"tool_calls": [
{
"function": {
"name": tool_name,
"arguments": {"query": "test"},
}
}
],
}
)
]
return [SimpleNamespace(message={"content": "final answer"})]
client = FakeClient()
monkeypatch.setattr(llm, "get_client", lambda _host=None: client)
monkeypatch.setattr(llm, "MAX_TOOL_ROUNDS", 1)
events = list(
llm.stream_chat_with_tools(
[Message(role="user", content="research this")],
model="test-model",
registry=registry,
capabilities={"tools"},
language="en",
)
)
assert executed == [(tool_name, {"query": "test"})]
assert len(client.calls) == 2
assert client.calls[0]["tools"]
assert client.calls[1]["tools"] == []
assert client.calls[0]["stream"] is True
assert client.calls[1]["stream"] is True
assert any(event.get("type") == "status" for event in events)
assert any(event == {"type": "delta", "content": "final answer"} for event in events)
assert events[-1] == {"type": "done"}
def test_paged_source_limit_emits_incomplete_warning(
monkeypatch: pytest.MonkeyPatch,
) -> None:
from ollama_web import llm
registry = ToolRegistry()
def scrape_url(url: str) -> str:
"""Return a paged result."""
return "body\n---page-info---\nhas_more: true\nnext_token: token"
registry.register(scrape_url)
class FakeClient:
def __init__(self) -> None:
self.calls: list[dict[str, Any]] = []
def chat(self, **kwargs: Any) -> list[SimpleNamespace]:
self.calls.append(kwargs)
if len(self.calls) == 1:
return [
SimpleNamespace(
message={
"content": "",
"tool_calls": [
{
"function": {
"name": "scrape_url",
"arguments": {"url": "https://example.com"},
}
}
],
}
)
]
return [SimpleNamespace(message={"content": "partial answer"})]
client = FakeClient()
monkeypatch.setattr(llm, "get_client", lambda _host=None: client)
monkeypatch.setattr(llm, "MAX_TOOL_ROUNDS", 1)
events = list(
llm.stream_chat_with_tools(
[Message(role="user", content="read all")],
model="test-model",
registry=registry,
capabilities={"tools"},
language="en",
)
)
assert any(event.get("type") == "warning" for event in events)
final_instruction = str(client.calls[-1]["messages"][-1].content)
assert "unread content" in final_instruction
def test_paged_tool_is_not_executed_when_context_budget_is_exhausted(
monkeypatch: pytest.MonkeyPatch,
) -> None:
from ollama_web import llm
executed: list[str] = []
registry = ToolRegistry()
def pdf_to_text(url_or_path: str) -> str:
"""Read a PDF."""
executed.append(url_or_path)
return "should not execute"
registry.register(pdf_to_text)
class FakeClient:
def __init__(self) -> None:
self.calls = 0
def chat(self, **kwargs: Any) -> list[SimpleNamespace]:
self.calls += 1
if self.calls == 1:
return [
SimpleNamespace(
message={
"content": "",
"tool_calls": [
{
"function": {
"name": "pdf_to_text",
"arguments": {"url_or_path": "https://example.com/a.pdf"},
}
}
],
}
)
]
return [SimpleNamespace(message={"content": "partial"})]
monkeypatch.setattr(llm, "get_client", lambda _host=None: FakeClient())
monkeypatch.setattr(llm, "MAX_TOOL_CONTEXT_CHARS", 1000)
monkeypatch.setattr(llm, "MAX_TOOL_ROUNDS", 1)
events = list(
llm.stream_chat_with_tools(
[Message(role="user", content="read all")],
model="test-model",
registry=registry,
capabilities={"tools"},
language="en",
)
)
assert executed == []
assert any("budget exhausted" in event.get("result", "") for event in events)
|
