File size: 6,864 Bytes
fc1a684
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from __future__ import annotations

import json
import argparse
import os
from pathlib import Path
from typing import Any, Dict, List

from conv_data_gen.generators.conversation.core_simulator import (
    ConversationSimulatorWithTools,
)
from conv_data_gen.generators.user.user_persona_generator import (
    UserPersonaGenerator,
    PersonaSample,
)
from concurrent.futures import ThreadPoolExecutor, as_completed


def _project_root() -> Path:
    return Path(__file__).resolve().parent.parent


def _default_assets_dir() -> Path:
    return _project_root() / "tests" / "assets"


def _default_output_dir() -> Path:
    out = _project_root() / "tests" / "output" / "user_randomizer"
    out.mkdir(parents=True, exist_ok=True)
    return out


def _load_proxy_json(path: Path) -> Dict[str, str]:
    obj = json.loads(path.read_text(encoding="utf-8"))
    if isinstance(obj, dict):
        return {str(k): str(v) for k, v in obj.items()}
    raise ValueError("Proxy JSON must be an object/dict")


def _read_text_file(path: Path) -> str:
    return path.read_text(encoding="utf-8")


def run_user_randomizer_test(
    n_variants: int,
    bot_prompt_path: Path,
    proxy_json_path: Path,
    out_dir: Path,
) -> Path:
    out_dir.mkdir(parents=True, exist_ok=True)
    proxy = _load_proxy_json(proxy_json_path)

    bot_text = _read_text_file(bot_prompt_path)

    # Single generator reused for sampling + formatting
    gen = UserPersonaGenerator(
        company_name="sarvam-finance",
        stakeholder="customer",
        use_case="loan",
        bot_prompt_text=bot_text,
    )

    mapping: List[Dict[str, Any]] = []

    def _work(i: int) -> Dict[str, Any]:
        # Sample attributes
        sample = gen._randomizer.sample_persona()

        # Compose persona text using reusable generator
        proxy_for_format = {str(k): str(v) for k, v in proxy.items()}
        persona = PersonaSample(
            knobs=sample.get("knobs", {}),
            knob_descriptions=sample.get("knob_descriptions", {}),
            language_style=sample.get("language_style", {}),
            language_descriptions=sample.get("language_descriptions", {}),
            demographics=sample.get("demographics", {}),
            demographic_descriptions=sample.get(
                "demographic_descriptions", {}
            ),
            interaction=sample.get("interaction", {}),
        )
        user_text = gen._compose_user_message_for_proxy(
            persona, proxy_for_format
        )

        # Persist persona
        # Run simulation
        sim = ConversationSimulatorWithTools(
            user_prompt_path=None,
            user_prompt_text=user_text,
            bot_prompt_path=str(bot_prompt_path),
            max_turns=10,
            output_dir=str(out_dir),
            base_filename=f"persona_{i}",
            conversation_direction="agent_to_user",
        )
        transcript = sim.generate()

        conv_base = (out_dir / f"persona_{i}").as_posix()
        # Prepend sampled knobs to the conversation text file
        try:
            conv_txt_path = out_dir / f"persona_{i}.txt"
            if conv_txt_path.exists():
                txt = conv_txt_path.read_text(encoding="utf-8")
                knobs_lines: List[str] = ["USER KNOBS (sampled):"]
                for k, v in sample.get("knobs", {}).items():
                    knobs_lines.append(f"- {k}: {v}")
                header = "\n".join(knobs_lines) + "\n\n"
                conv_txt_path.write_text(header + txt, encoding="utf-8")
        except Exception:
            pass

        # Enrich conversation JSON with user attributes
        conv_json_path = out_dir / f"persona_{i}.json"
        try:
            if conv_json_path.exists():
                meta = json.loads(conv_json_path.read_text(encoding="utf-8"))
                meta["user_attributes"] = {
                    "knobs": sample.get("knobs", {}),
                    "language_style": sample.get("language_style", {}),
                    "demographics": sample.get("demographics", {}),
                    "interaction": sample.get("interaction", {}),
                }
                conv_json_path.write_text(
                    json.dumps(meta, ensure_ascii=False, indent=2),
                    encoding="utf-8",
                )
        except Exception:
            pass
        return {
            "persona_index": i,
            "conversation_base": conv_base,
            "user_prompt_path": "",
            "bot_prompt_path": str(bot_prompt_path),
            "sampled_attributes": {
                "knobs": sample.get("knobs", {}),
                "language_style": sample.get("language_style", {}),
                "demographics": sample.get("demographics", {}),
                "interaction": sample.get("interaction", {}),
            },
            "total_turns": len(transcript),
        }

    cpu = os.cpu_count() or 2
    max_workers = max(1, min(16, cpu * 4))
    workers = min(max_workers, max(1, n_variants))

    futures = []
    with ThreadPoolExecutor(max_workers=workers) as ex:
        for i in range(1, n_variants + 1):
            futures.append(ex.submit(_work, i))
        for fut in as_completed(futures):
            try:
                res = fut.result()
                if isinstance(res, dict):
                    mapping.append(res)
            except Exception:
                # Skip failed runs but continue others
                pass

    # Stable order by persona index
    mapping.sort(key=lambda r: int(r.get("persona_index", 0)))

    # Write compact JSON map: conversation file base -> user attributes
    map_path = out_dir / "conversation_user_attributes_map.json"
    with open(map_path, "w", encoding="utf-8") as f:
        json.dump(mapping, f, ensure_ascii=False, indent=2)

    return map_path


def _parse_args() -> argparse.Namespace:
    p = argparse.ArgumentParser(
        description=(
            "Generate conversations to test the effect of user attributes."
        )
    )
    p.add_argument("--num", type=int, default=10, help="Number of variants")
    p.add_argument(
        "--bot",
        type=Path,
        default=_default_assets_dir() / "bot.txt",
        help="Path to bot prompt text file",
    )
    p.add_argument(
        "--proxy",
        type=Path,
        default=_default_assets_dir() / "proxy.json",
        help="Path to base user proxy JSON file",
    )
    p.add_argument(
        "--out",
        type=Path,
        default=_default_output_dir(),
        help="Output directory for personas, conversations, and map",
    )
    return p.parse_args()


if __name__ == "__main__":
    args = _parse_args()
    path = run_user_randomizer_test(
        n_variants=max(1, int(args.num)),
        bot_prompt_path=args.bot,
        proxy_json_path=args.proxy,
        out_dir=args.out,
    )
    print(str(path))