File size: 8,475 Bytes
d61821a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
"""Outcome-blind preflight for the frozen Study 2 execution revision."""

from __future__ import annotations

from dataclasses import asdict
from hashlib import sha256
import importlib.metadata
import json
import platform
from pathlib import Path
import shutil
import subprocess
import sys
import time
from typing import Any

from agent_harness.lm_studio import LMStudioClient
from agent_harness.lm_studio_embeddings import LMStudioEmbeddingClient
from agent_harness.lm_studio_management import (
    LMStudioResidencyManager,
    LMStudioServer,
)
from agent_harness.pilot import research_code_revision
from agent_harness.specs import (
    load_embeddings,
    load_models,
    load_repositories,
    validate_configuration_tree,
)
from agent_harness.study2_experiment import tokenizer_for


PACKAGES = (
    "faiss-cpu",
    "numpy",
    "pytest",
    "scipy",
    "statsmodels",
    "tokenizers",
    "tree-sitter",
    "tree-sitter-go",
    "tree-sitter-python",
)


def _command(arguments: list[str], cwd: Path | None = None) -> dict[str, Any]:
    result = subprocess.run(
        arguments,
        cwd=cwd,
        text=True,
        capture_output=True,
        check=False,
        timeout=120,
    )
    return {
        "command": arguments,
        "returncode": result.returncode,
        "stdout": result.stdout,
        "stderr": result.stderr,
    }


def _tool_probe(client: LMStudioClient, model_key: str) -> dict[str, Any]:
    response = client.chat_completions(
        model_key,
        [
            {
                "role": "system",
                "content": "This is a runtime preflight. Use the required tool exactly once.",
            },
            {
                "role": "user",
                "content": "Call preflight_echo with marker MODEL_TOOL_OK. Do not answer in prose.",
            },
        ],
        tools=[
            {
                "type": "function",
                "function": {
                    "name": "preflight_echo",
                    "description": "Return the requested preflight marker.",
                    "parameters": {
                        "type": "object",
                        "properties": {"marker": {"type": "string"}},
                        "required": ["marker"],
                        "additionalProperties": False,
                    },
                },
            }
        ],
        max_tokens=2_048,
        seed=0,
    )
    try:
        message = response["choices"][0]["message"]
        calls = message["tool_calls"]
        function = calls[0]["function"]
        arguments = function["arguments"]
        decoded = arguments if isinstance(arguments, dict) else json.loads(arguments)
    except (KeyError, IndexError, TypeError, json.JSONDecodeError) as exc:
        raise RuntimeError("model did not return a valid preflight tool call") from exc
    if function.get("name") != "preflight_echo" or decoded != {"marker": "MODEL_TOOL_OK"}:
        raise RuntimeError(f"unexpected tool preflight payload: {function}")
    serialized = json.dumps(response, sort_keys=True, separators=(",", ":"))
    return {
        "tool_name": function["name"],
        "arguments": decoded,
        "finish_reason": response["choices"][0].get("finish_reason"),
        "usage": response.get("usage", {}),
        "response_sha256": sha256(serialized.encode()).hexdigest(),
    }


def run(root: Path) -> dict[str, Any]:
    revision = research_code_revision(root)
    errors, warnings = validate_configuration_tree(root)
    if errors or warnings:
        raise RuntimeError(f"configuration failed: errors={errors}, warnings={warnings}")
    disk = shutil.disk_usage(root)
    if disk.free < 50 * 1024**3:
        raise RuntimeError(f"less than 50 GiB free before Study 2: {disk.free} bytes")

    models = load_models(root)
    embedding = load_embeddings(root)["EMB002"]
    repositories = load_repositories(root)
    server = LMStudioServer(port=1234)
    server_state = server.ensure_running()
    residency = LMStudioResidencyManager(
        models["M002"].base_url,
        models["M002"].api_token_env,
        timeout_seconds=1_800,
    )
    report: dict[str, Any] = {
        "schema_version": 1,
        "study": "Study 2",
        "outcome_blind": True,
        "research_code_revision": revision,
        "started_unix": time.time(),
        "platform": platform.platform(),
        "python": sys.version,
        "torch_used": False,
        "disk": {"total": disk.total, "used": disk.used, "free": disk.free},
        "lms_version": _command([str(server.cli_path), "--version"]),
        "server_start": server_state,
        "dependencies": {
            package: importlib.metadata.version(package) for package in PACKAGES
        },
        "repositories": {},
        "models": {},
        "embedding": {},
    }
    try:
        residency.unload_all()
        for repository_id, repository in sorted(repositories.items()):
            observed = _command(
                ["git", "rev-parse", "HEAD"], cwd=root / repository.local_path
            )
            if observed["returncode"] or observed["stdout"].strip() != repository.pinned_head:
                raise RuntimeError(f"repository head mismatch: {repository_id}: {observed}")
            report["repositories"][repository_id] = {
                "spec": asdict(repository),
                "observed_head": observed["stdout"].strip(),
            }

        for model_id in ("M002", "M003"):
            model = models[model_id]
            transition = residency.ensure_exclusive(
                model.expected_inference_key, model.context_length
            )
            client = LMStudioClient(model, timeout_seconds=1_800)
            discovery, resolved = client.resolve()
            tokenizer = tokenizer_for(model)
            report["models"][model_id] = {
                "spec": asdict(model),
                "config_hash": model.config_hash,
                "transition": transition.to_dict(),
                "resolved": resolved.to_dict(),
                "discovery_errors": discovery.endpoint_errors,
                "tokenizer_path": str(tokenizer.path),
                "tokenizer_sha256": tokenizer.sha256,
                "tool_probe": _tool_probe(client, resolved.inference_key),
            }
            report["models"][model_id]["unload"] = residency.unload_all().to_dict()

        transition = residency.ensure_exclusive(
            embedding.model_key, embedding.loaded_context_length
        )
        embedding_client = LMStudioEmbeddingClient(embedding, timeout_seconds=1_800)
        record = embedding_client.resolve()
        probe = embedding_client.probe()
        report["embedding"] = {
            "spec": asdict(embedding),
            "config_hash": embedding.config_hash,
            "transition": transition.to_dict(),
            "resolved": record,
            "probe": probe.to_dict(),
            "unload": residency.unload_all().to_dict(),
        }
        report["passed"] = True
        return report
    finally:
        cleanup_errors: list[str] = []
        try:
            report["final_unload"] = residency.unload_all().to_dict()
        except Exception as cleanup_error:
            cleanup_errors.append(f"unload_all: {cleanup_error}")
        try:
            status = server.status()
            report["server_stop"] = (
                server.stop()
                if status["running"]
                else {"action": "already_stopped", "status": status}
            )
        except Exception as cleanup_error:
            cleanup_errors.append(f"server_stop: {cleanup_error}")
        report["cleanup_errors"] = cleanup_errors
        report["finished_unix"] = time.time()


def main() -> None:
    root = Path(__file__).resolve().parents[1]
    report: dict[str, Any] = {}
    output = root / "results" / "reports" / "study2_preflight.json"
    output.parent.mkdir(parents=True, exist_ok=True)
    try:
        report = run(root)
    except Exception as exc:
        report = {**report, "passed": False, "error": repr(exc)}
        output.write_text(
            json.dumps(report, indent=2, sort_keys=True, default=str) + "\n",
            encoding="utf-8",
        )
        raise
    output.write_text(
        json.dumps(report, indent=2, sort_keys=True, default=str) + "\n",
        encoding="utf-8",
    )
    print(json.dumps({"passed": True, "report": str(output)}, indent=2))


if __name__ == "__main__":
    main()