witnis 1.11.0__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- witnis/__init__.py +3 -0
- witnis/__main__.py +3 -0
- witnis/bench.py +367 -0
- witnis/cli.py +1228 -0
- witnis/compare_baselines.py +275 -0
- witnis/core/__init__.py +1 -0
- witnis/core/ab_experiment.py +171 -0
- witnis/core/bench_report.py +142 -0
- witnis/core/calls_python.py +204 -0
- witnis/core/discover.py +62 -0
- witnis/core/envelope.py +34 -0
- witnis/core/gitinfo.py +359 -0
- witnis/core/ground_truth.py +284 -0
- witnis/core/gt_filters.py +40 -0
- witnis/core/index.py +252 -0
- witnis/core/loop_init.py +206 -0
- witnis/core/lsp_python.py +302 -0
- witnis/core/models.py +23 -0
- witnis/core/native.py +79 -0
- witnis/core/parse_go.py +39 -0
- witnis/core/parse_java.py +48 -0
- witnis/core/parse_js.py +108 -0
- witnis/core/parse_python.py +58 -0
- witnis/core/pr_gate.py +584 -0
- witnis/core/references_python.py +54 -0
- witnis/core/relations_go.py +98 -0
- witnis/core/relations_java.py +370 -0
- witnis/core/relations_js.py +126 -0
- witnis/core/relations_python.py +177 -0
- witnis/core/runtime.py +50 -0
- witnis/core/scene.py +214 -0
- witnis/core/security.py +117 -0
- witnis/core/store.py +785 -0
- witnis/core/tools.py +2625 -0
- witnis/core/verify.py +605 -0
- witnis/core/verify_isolate.py +275 -0
- witnis/mcp_server.py +476 -0
- witnis/sdk.py +83 -0
- witnis/studio/__init__.py +5 -0
- witnis/studio/assets/blast-radius.svg +79 -0
- witnis/studio/assets/caught-lying.svg +23 -0
- witnis/studio/assets/trusted-agent-loop.svg +34 -0
- witnis/studio/assets/witnis-masthead.svg +13 -0
- witnis/studio/home.html +144 -0
- witnis/studio/render.py +212 -0
- witnis/studio/scene_template.html +301 -0
- witnis/studio/template.html +350 -0
- witnis-1.11.0.dist-info/METADATA +91 -0
- witnis-1.11.0.dist-info/RECORD +53 -0
- witnis-1.11.0.dist-info/WHEEL +5 -0
- witnis-1.11.0.dist-info/entry_points.txt +2 -0
- witnis-1.11.0.dist-info/licenses/LICENSE +190 -0
- witnis-1.11.0.dist-info/top_level.txt +1 -0
witnis/__init__.py
ADDED
witnis/__main__.py
ADDED
witnis/bench.py
ADDED
|
@@ -0,0 +1,367 @@
|
|
|
1
|
+
from __future__ import annotations
|
|
2
|
+
|
|
3
|
+
import json
|
|
4
|
+
from dataclasses import dataclass
|
|
5
|
+
from pathlib import Path
|
|
6
|
+
from typing import Any
|
|
7
|
+
|
|
8
|
+
from witnis.core.index import index_repository
|
|
9
|
+
from witnis.core.tools import (
|
|
10
|
+
analyze_impact,
|
|
11
|
+
check_import_graph,
|
|
12
|
+
check_index_health,
|
|
13
|
+
explain_code,
|
|
14
|
+
export_evidence_bundle,
|
|
15
|
+
find_related_tests,
|
|
16
|
+
get_evidence,
|
|
17
|
+
export_relation_graph,
|
|
18
|
+
list_capabilities,
|
|
19
|
+
list_index_snapshots,
|
|
20
|
+
ingest_runtime_traces,
|
|
21
|
+
query_lsp,
|
|
22
|
+
refresh_evidence,
|
|
23
|
+
render_evidence_report,
|
|
24
|
+
repo_overview,
|
|
25
|
+
repo_search,
|
|
26
|
+
repo_status,
|
|
27
|
+
trace_symbol,
|
|
28
|
+
)
|
|
29
|
+
from witnis.compare_baselines import run_baseline_compare
|
|
30
|
+
from witnis.core.verify import create_verification_plan, run_verification
|
|
31
|
+
from witnis.studio import render_studio
|
|
32
|
+
|
|
33
|
+
|
|
34
|
+
@dataclass
|
|
35
|
+
class TaskResult:
|
|
36
|
+
id: str
|
|
37
|
+
passed: bool
|
|
38
|
+
detail: str
|
|
39
|
+
|
|
40
|
+
|
|
41
|
+
def _load_tasks(path: Path) -> dict[str, Any]:
|
|
42
|
+
return json.loads(path.read_text(encoding="utf-8"))
|
|
43
|
+
|
|
44
|
+
|
|
45
|
+
def _check_expect(kind: str, payload: dict[str, Any], expect: dict[str, Any]) -> tuple[bool, str]:
|
|
46
|
+
if kind == "search":
|
|
47
|
+
names = {m.get("name") for m in payload.get("result", {}).get("matches", [])}
|
|
48
|
+
wanted = set(expect.get("any_symbol_names") or [])
|
|
49
|
+
ok = bool(names & wanted)
|
|
50
|
+
return ok, f"names={sorted(names)}"
|
|
51
|
+
|
|
52
|
+
if kind == "trace":
|
|
53
|
+
called_by = payload.get("result", {}).get("called_by") or []
|
|
54
|
+
suffixes = expect.get("called_by_source_suffix") or []
|
|
55
|
+
deco_targets = expect.get("decorated_by_target") or []
|
|
56
|
+
if deco_targets:
|
|
57
|
+
deco = payload.get("result", {}).get("decorated_by") or []
|
|
58
|
+
ok = any(
|
|
59
|
+
any(str(edge.get("target")) == t for t in deco_targets) for edge in deco
|
|
60
|
+
)
|
|
61
|
+
return ok, f"decorated_by={deco}"
|
|
62
|
+
ok = any(
|
|
63
|
+
any(str(edge.get("source", "")).endswith(suf) for suf in suffixes)
|
|
64
|
+
for edge in called_by
|
|
65
|
+
)
|
|
66
|
+
return ok, f"called_by={called_by}"
|
|
67
|
+
|
|
68
|
+
if kind == "impact":
|
|
69
|
+
direct = payload.get("result", {}).get("direct") or []
|
|
70
|
+
suffixes = expect.get("direct_source_suffix") or []
|
|
71
|
+
ok = any(
|
|
72
|
+
any(str(edge.get("source", "")).endswith(suf) for suf in suffixes)
|
|
73
|
+
for edge in direct
|
|
74
|
+
)
|
|
75
|
+
return ok, f"direct={direct}"
|
|
76
|
+
|
|
77
|
+
if kind == "related_tests":
|
|
78
|
+
tests = payload.get("result", {}).get("tests") or []
|
|
79
|
+
needles = expect.get("test_file_contains") or []
|
|
80
|
+
ok = any(
|
|
81
|
+
any(n in str(t.get("file_path")) for n in needles) for t in tests
|
|
82
|
+
)
|
|
83
|
+
return ok, f"tests={tests}"
|
|
84
|
+
|
|
85
|
+
if kind == "evidence":
|
|
86
|
+
count = len(payload.get("evidence") or [])
|
|
87
|
+
need = int(expect.get("min_evidence") or 1)
|
|
88
|
+
return count >= need, f"evidence_count={count}"
|
|
89
|
+
|
|
90
|
+
if kind == "overview":
|
|
91
|
+
result = payload.get("result") or {}
|
|
92
|
+
ok = (
|
|
93
|
+
int(result.get("symbols") or 0) >= int(expect.get("min_symbols") or 0)
|
|
94
|
+
and int(result.get("files") or 0) >= int(expect.get("min_files") or 0)
|
|
95
|
+
)
|
|
96
|
+
return ok, f"files={result.get('files')} symbols={result.get('symbols')}"
|
|
97
|
+
|
|
98
|
+
if kind == "overview_relations":
|
|
99
|
+
sample = payload.get("result", {}).get("relation_sample") or []
|
|
100
|
+
needles = expect.get("relation_target_contains") or []
|
|
101
|
+
ok = any(any(n in str(e.get("target")) for n in needles) for e in sample)
|
|
102
|
+
return ok, f"relations={sample}"
|
|
103
|
+
|
|
104
|
+
if kind == "plan_verify":
|
|
105
|
+
steps = payload.get("result", {}).get("steps") or []
|
|
106
|
+
kinds = {s.get("kind") for s in steps}
|
|
107
|
+
wanted = set(expect.get("step_kinds") or [])
|
|
108
|
+
ok = wanted.issubset(kinds)
|
|
109
|
+
return ok, f"step_kinds={sorted(kinds)}"
|
|
110
|
+
|
|
111
|
+
if kind == "verify_blocked":
|
|
112
|
+
state = payload.get("result", {}).get("state")
|
|
113
|
+
ok = state == expect.get("state")
|
|
114
|
+
return ok, f"state={state}"
|
|
115
|
+
|
|
116
|
+
if kind == "explain_code":
|
|
117
|
+
snippet = payload.get("result", {}).get("snippet") or []
|
|
118
|
+
need = int(expect.get("min_snippet_lines") or 1)
|
|
119
|
+
ok = len(snippet) >= need
|
|
120
|
+
return ok, f"snippet_lines={len(snippet)}"
|
|
121
|
+
|
|
122
|
+
if kind == "mcp_tools":
|
|
123
|
+
names = set(payload.get("result", {}).get("tools") or [])
|
|
124
|
+
wanted = set(expect.get("tool_names") or [])
|
|
125
|
+
ok = wanted.issubset(names)
|
|
126
|
+
return ok, f"tools={sorted(names)}"
|
|
127
|
+
|
|
128
|
+
if kind == "import_graph":
|
|
129
|
+
broken = int(payload.get("result", {}).get("broken_count") or 0)
|
|
130
|
+
max_broken = int(expect.get("max_broken") or 0)
|
|
131
|
+
ok = broken <= max_broken
|
|
132
|
+
return ok, f"broken_count={broken}"
|
|
133
|
+
|
|
134
|
+
if kind == "export_bundle":
|
|
135
|
+
bundle = payload.get("result", {}).get("bundle") or {}
|
|
136
|
+
wanted = set(expect.get("bundle_keys") or [])
|
|
137
|
+
ok = wanted.issubset(set(bundle.keys()))
|
|
138
|
+
return ok, f"bundle_keys={sorted(bundle.keys())}"
|
|
139
|
+
|
|
140
|
+
if kind == "index_health":
|
|
141
|
+
issues = payload.get("result", {}).get("issues") or []
|
|
142
|
+
max_issues = int(expect.get("max_issues") or 0)
|
|
143
|
+
ok = len(issues) <= max_issues
|
|
144
|
+
return ok, f"issues={issues}"
|
|
145
|
+
|
|
146
|
+
if kind == "report":
|
|
147
|
+
md = payload.get("result", {}).get("markdown") or ""
|
|
148
|
+
needles = expect.get("markdown_contains") or []
|
|
149
|
+
ok = all(n in md for n in needles)
|
|
150
|
+
return ok, f"markdown_len={len(md)}"
|
|
151
|
+
|
|
152
|
+
if kind == "status":
|
|
153
|
+
result = payload.get("result") or {}
|
|
154
|
+
wanted = set(expect.get("has_keys") or [])
|
|
155
|
+
ok = wanted.issubset(set(result.keys()))
|
|
156
|
+
return ok, f"keys={sorted(result.keys())}"
|
|
157
|
+
|
|
158
|
+
if kind == "snapshots":
|
|
159
|
+
count = int(payload.get("result", {}).get("count") or 0)
|
|
160
|
+
need = int(expect.get("min_count") or 1)
|
|
161
|
+
ok = count >= need
|
|
162
|
+
return ok, f"count={count}"
|
|
163
|
+
|
|
164
|
+
if kind == "graph":
|
|
165
|
+
graph = payload.get("result", {}).get("graph") or ""
|
|
166
|
+
needles = expect.get("graph_contains") or []
|
|
167
|
+
min_edges = int(expect.get("min_edges") or 0)
|
|
168
|
+
edges = int(payload.get("result", {}).get("edge_count") or 0)
|
|
169
|
+
ok = all(n in graph for n in needles) and edges >= min_edges
|
|
170
|
+
return ok, f"edges={edges}"
|
|
171
|
+
|
|
172
|
+
if kind == "capabilities":
|
|
173
|
+
tools = payload.get("result", {}).get("mcp_tools") or []
|
|
174
|
+
experimental = payload.get("result", {}).get("mcp_experimental") or []
|
|
175
|
+
need = int(expect.get("min_mcp_tools") or 1)
|
|
176
|
+
ok = len(tools) >= need
|
|
177
|
+
exact = expect.get("exact_mcp_tools")
|
|
178
|
+
if exact:
|
|
179
|
+
ok = ok and set(tools) == set(exact)
|
|
180
|
+
min_exp = expect.get("min_mcp_experimental")
|
|
181
|
+
if min_exp is not None:
|
|
182
|
+
ok = ok and len(experimental) >= int(min_exp)
|
|
183
|
+
return ok, f"mcp_tools={len(tools)} experimental={len(experimental)}"
|
|
184
|
+
|
|
185
|
+
if kind == "studio":
|
|
186
|
+
result = payload.get("result") or {}
|
|
187
|
+
ok = True
|
|
188
|
+
if expect.get("has_graph") is True:
|
|
189
|
+
ok = ok and bool(result.get("has_graph"))
|
|
190
|
+
if expect.get("has_report") is True:
|
|
191
|
+
ok = ok and bool(result.get("has_report"))
|
|
192
|
+
return ok, f"studio={result}"
|
|
193
|
+
|
|
194
|
+
if kind == "refresh_evidence":
|
|
195
|
+
result = payload.get("result") or {}
|
|
196
|
+
checked = int(result.get("checked") or 0)
|
|
197
|
+
need = int(expect.get("min_checked") or 0)
|
|
198
|
+
ok = payload.get("status") in {"success", "partial"} and checked >= need
|
|
199
|
+
return ok, f"checked={checked} stale={result.get('marked_stale')}"
|
|
200
|
+
|
|
201
|
+
if kind == "ingest_traces":
|
|
202
|
+
result = payload.get("result") or {}
|
|
203
|
+
accepted = int(result.get("accepted") or 0)
|
|
204
|
+
need = int(expect.get("min_accepted") or 1)
|
|
205
|
+
ok = payload.get("status") == "success" and accepted >= need
|
|
206
|
+
return ok, f"accepted={accepted} added={result.get('relations_added')}"
|
|
207
|
+
|
|
208
|
+
if kind == "baselines":
|
|
209
|
+
result = payload.get("result") or {}
|
|
210
|
+
deltas = result.get("deltas") or {}
|
|
211
|
+
security = result.get("security") or {}
|
|
212
|
+
ok = (
|
|
213
|
+
int(deltas.get("typed_symbols_over_grep") or 0)
|
|
214
|
+
>= int(expect.get("min_typed_symbols") or 1)
|
|
215
|
+
and int(deltas.get("relation_edges_over_grep") or 0)
|
|
216
|
+
>= int(expect.get("min_relation_edges") or 1)
|
|
217
|
+
)
|
|
218
|
+
if expect.get("verify_blocked") is True:
|
|
219
|
+
ok = ok and bool(security.get("verify_blocked_without_auth"))
|
|
220
|
+
return ok, f"deltas={deltas} security={security}"
|
|
221
|
+
|
|
222
|
+
if kind in {"go_search", "fixture_search", "js_search", "java_search"}:
|
|
223
|
+
names = {m.get("name") for m in payload.get("result", {}).get("matches", [])}
|
|
224
|
+
wanted = set(expect.get("any_symbol_names") or [])
|
|
225
|
+
ok = bool(names & wanted)
|
|
226
|
+
return ok, f"names={sorted(names)}"
|
|
227
|
+
|
|
228
|
+
if kind == "lsp":
|
|
229
|
+
result = payload.get("result") or {}
|
|
230
|
+
if expect.get("allow_unavailable") and not result.get("available"):
|
|
231
|
+
return True, "lsp_unavailable_allowed"
|
|
232
|
+
ok = bool(result.get("definitions") or result.get("references"))
|
|
233
|
+
return ok, f"lsp={result}"
|
|
234
|
+
|
|
235
|
+
if kind == "overview_min":
|
|
236
|
+
result = payload.get("result") or {}
|
|
237
|
+
ok = (
|
|
238
|
+
int(result.get("files") or 0) >= int(expect.get("min_files") or 0)
|
|
239
|
+
and int(result.get("symbols") or 0) >= int(expect.get("min_symbols") or 0)
|
|
240
|
+
)
|
|
241
|
+
return ok, f"files={result.get('files')} symbols={result.get('symbols')}"
|
|
242
|
+
|
|
243
|
+
return False, f"unknown kind={kind}"
|
|
244
|
+
|
|
245
|
+
|
|
246
|
+
def run_benchmark(
|
|
247
|
+
repo_root: Path,
|
|
248
|
+
tasks_path: Path,
|
|
249
|
+
db_path: Path,
|
|
250
|
+
) -> dict[str, Any]:
|
|
251
|
+
spec = _load_tasks(tasks_path)
|
|
252
|
+
fixture_rel = spec.get("fixture") or "fixtures/mini_py"
|
|
253
|
+
fixture = (repo_root / fixture_rel).resolve()
|
|
254
|
+
index_repository(fixture, db_path, incremental=False)
|
|
255
|
+
|
|
256
|
+
results: list[TaskResult] = []
|
|
257
|
+
for task in spec.get("tasks") or []:
|
|
258
|
+
kind = task["kind"]
|
|
259
|
+
query = task.get("query") or ""
|
|
260
|
+
expect = task.get("expect") or {}
|
|
261
|
+
if kind == "search":
|
|
262
|
+
payload = repo_search(db_path, query)
|
|
263
|
+
elif kind == "trace":
|
|
264
|
+
payload = trace_symbol(db_path, query)
|
|
265
|
+
elif kind == "impact":
|
|
266
|
+
payload = analyze_impact(db_path, query)
|
|
267
|
+
elif kind == "related_tests":
|
|
268
|
+
payload = find_related_tests(db_path, query)
|
|
269
|
+
elif kind == "evidence":
|
|
270
|
+
payload = get_evidence(db_path, query)
|
|
271
|
+
elif kind in {"overview", "overview_relations"}:
|
|
272
|
+
payload = repo_overview(db_path)
|
|
273
|
+
elif kind == "plan_verify":
|
|
274
|
+
payload = create_verification_plan(db_path, target=query or None)
|
|
275
|
+
elif kind == "verify_blocked":
|
|
276
|
+
payload = run_verification(db_path, execute=False, target=query or None)
|
|
277
|
+
elif kind == "explain_code":
|
|
278
|
+
payload = explain_code(db_path, query)
|
|
279
|
+
elif kind == "mcp_tools":
|
|
280
|
+
from witnis.mcp_server import build_server
|
|
281
|
+
|
|
282
|
+
payload = {
|
|
283
|
+
"result": {
|
|
284
|
+
"tools": [t.name for t in build_server()._tool_manager.list_tools()]
|
|
285
|
+
}
|
|
286
|
+
}
|
|
287
|
+
elif kind == "import_graph":
|
|
288
|
+
payload = check_import_graph(db_path)
|
|
289
|
+
elif kind == "export_bundle":
|
|
290
|
+
payload = export_evidence_bundle(db_path, target=query or None)
|
|
291
|
+
elif kind == "index_health":
|
|
292
|
+
payload = check_index_health(db_path)
|
|
293
|
+
elif kind == "report":
|
|
294
|
+
payload = render_evidence_report(db_path, target=query or None)
|
|
295
|
+
elif kind == "status":
|
|
296
|
+
payload = repo_status(db_path)
|
|
297
|
+
elif kind == "snapshots":
|
|
298
|
+
payload = list_index_snapshots(db_path)
|
|
299
|
+
elif kind == "graph":
|
|
300
|
+
payload = export_relation_graph(db_path, query or "login", fmt="mermaid")
|
|
301
|
+
elif kind == "capabilities":
|
|
302
|
+
payload = list_capabilities()
|
|
303
|
+
elif kind == "studio":
|
|
304
|
+
out = db_path.parent / "bench-studio.html"
|
|
305
|
+
payload = render_studio(db_path, out=out, target=query or "login")
|
|
306
|
+
elif kind == "refresh_evidence":
|
|
307
|
+
payload = refresh_evidence(db_path)
|
|
308
|
+
elif kind == "ingest_traces":
|
|
309
|
+
rel = query or "fixtures/mini_py/sample_trace.jsonl"
|
|
310
|
+
payload = ingest_runtime_traces(db_path, (repo_root / rel).resolve())
|
|
311
|
+
elif kind == "baselines":
|
|
312
|
+
payload = run_baseline_compare(
|
|
313
|
+
(repo_root / (spec.get("fixture") or "fixtures/mini_py")).resolve(),
|
|
314
|
+
db_path,
|
|
315
|
+
query=query or "login",
|
|
316
|
+
)
|
|
317
|
+
elif kind == "go_search":
|
|
318
|
+
go_db = db_path.parent / "bench-go.sqlite"
|
|
319
|
+
index_repository(
|
|
320
|
+
(repo_root / "fixtures" / "mini_go").resolve(),
|
|
321
|
+
go_db,
|
|
322
|
+
incremental=False,
|
|
323
|
+
)
|
|
324
|
+
payload = repo_search(go_db, query or "Login")
|
|
325
|
+
elif kind in {"fixture_search", "js_search", "java_search"}:
|
|
326
|
+
fixture_rel = (
|
|
327
|
+
task.get("fixture")
|
|
328
|
+
or expect.get("fixture")
|
|
329
|
+
or {
|
|
330
|
+
"js_search": "fixtures/mini_js",
|
|
331
|
+
"java_search": "fixtures/mini_java",
|
|
332
|
+
"fixture_search": "fixtures/public_svc",
|
|
333
|
+
}.get(kind, "fixtures/public_svc")
|
|
334
|
+
)
|
|
335
|
+
alt_db = db_path.parent / f"bench-{kind}.sqlite"
|
|
336
|
+
index_repository(
|
|
337
|
+
(repo_root / fixture_rel).resolve(),
|
|
338
|
+
alt_db,
|
|
339
|
+
incremental=False,
|
|
340
|
+
)
|
|
341
|
+
payload = repo_search(alt_db, query)
|
|
342
|
+
elif kind == "overview_min":
|
|
343
|
+
fixture_rel = task.get("fixture") or "fixtures/public_svc"
|
|
344
|
+
alt_db = db_path.parent / "bench-overview.sqlite"
|
|
345
|
+
index_repository(
|
|
346
|
+
(repo_root / fixture_rel).resolve(),
|
|
347
|
+
alt_db,
|
|
348
|
+
incremental=False,
|
|
349
|
+
)
|
|
350
|
+
payload = repo_overview(alt_db)
|
|
351
|
+
elif kind == "lsp":
|
|
352
|
+
payload = query_lsp(db_path, query or "AuthService")
|
|
353
|
+
else:
|
|
354
|
+
results.append(TaskResult(task["id"], False, f"unsupported kind {kind}"))
|
|
355
|
+
continue
|
|
356
|
+
ok, detail = _check_expect(kind, payload, expect)
|
|
357
|
+
results.append(TaskResult(task["id"], ok, detail))
|
|
358
|
+
|
|
359
|
+
passed = sum(1 for r in results if r.passed)
|
|
360
|
+
total = len(results)
|
|
361
|
+
return {
|
|
362
|
+
"fixture": str(fixture),
|
|
363
|
+
"passed": passed,
|
|
364
|
+
"total": total,
|
|
365
|
+
"score": (passed / total) if total else 0.0,
|
|
366
|
+
"results": [{"id": r.id, "passed": r.passed, "detail": r.detail} for r in results],
|
|
367
|
+
}
|