Spaces:
Running
Running
File size: 14,874 Bytes
5d959d0 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 | """
Court Session Manager.
Single source of truth for everything that happens in a moot court session.
Every agent reads from and writes to the session object.
Sessions persist to HuggingFace Dataset for durability across container restarts.
Session lifecycle:
created → briefing → rounds → cross_examination → closing → completed
WHY store to HF Dataset?
HF Spaces containers are ephemeral. Without durable storage, all session
data is lost on restart. HF Dataset API gives us free durable storage
using the same HF_TOKEN already in the Space secrets.
"""
import os
import json
import uuid
import logging
from datetime import datetime, timezone
from typing import Optional, Dict, List, Any
from dataclasses import dataclass, field, asdict
logger = logging.getLogger(__name__)
HF_TOKEN = os.getenv("HF_TOKEN")
SESSIONS_REPO = "CaffeinatedCoding/nyayasetu-court-sessions"
# ── In-memory session store ────────────────────────────────────
# Primary store during runtime. HF Dataset is the durable backup.
_sessions: Dict[str, Dict] = {}
# ── Data structures ────────────────────────────────────────────
@dataclass
class TranscriptEntry:
"""A single entry in the court transcript."""
speaker: str # JUDGE | OPPOSING_COUNSEL | REGISTRAR | PETITIONER | RESPONDENT
role_label: str # Display label e.g. "HON'BLE COURT", "RESPONDENT'S COUNSEL"
content: str # The actual text
round_number: int # Which round this belongs to
phase: str # briefing | argument | cross_examination | closing
timestamp: str # ISO timestamp
entry_type: str # argument | question | observation | objection | ruling | document | trap
metadata: Dict = field(default_factory=dict) # extra data e.g. trap_type, precedents_cited
@dataclass
class Concession:
"""A concession made by the user during the session."""
round_number: int
exact_quote: str # The exact text where concession was made
legal_significance: str # What opposing counsel can do with this
exploited: bool = False # Has opposing counsel used this yet
@dataclass
class TrapEvent:
"""A trap set by opposing counsel."""
round_number: int
trap_type: str # admission_trap | precedent_trap | inconsistency_trap
trap_text: str # What opposing counsel said to set the trap
user_fell_in: bool # Whether user fell into the trap
user_response: str = "" # What user said in response
@dataclass
class CourtSession:
"""Complete court session state."""
# Identity
session_id: str
created_at: str
updated_at: str
# Case
case_title: str
user_side: str # petitioner | respondent
user_client: str
opposing_party: str
legal_issues: List[str]
brief_facts: str
jurisdiction: str # supreme_court | high_court | district_court
# Setup
bench_composition: str # single | division | constitutional
difficulty: str # moot | standard | adversarial
session_length: str # brief | standard | extended
show_trap_warnings: bool
# Derived from research session import
imported_from_session: Optional[str] # NyayaSetu research session ID
case_brief: str # Generated case brief text
retrieved_precedents: List[Dict] # Precedents from research session
# Session progress
phase: str # briefing | rounds | cross_examination | closing | completed
current_round: int
max_rounds: int # 3 | 5 | 8
# Transcript
transcript: List[Dict] # List of TranscriptEntry as dicts
# Tracking
concessions: List[Dict] # List of Concession as dicts
trap_events: List[Dict] # List of TrapEvent as dicts
cited_precedents: List[str] # Judgment IDs cited during session
documents_produced: List[Dict] # Documents generated during session
# Arguments tracking for inconsistency detection
user_arguments: List[Dict] # [{round, text, key_claims: []}]
# Analysis (populated at end)
analysis: Optional[Dict]
outcome_prediction: Optional[str]
performance_score: Optional[float]
def create_session(
case_title: str,
user_side: str,
user_client: str,
opposing_party: str,
legal_issues: List[str],
brief_facts: str,
jurisdiction: str,
bench_composition: str,
difficulty: str,
session_length: str,
show_trap_warnings: bool,
imported_from_session: Optional[str] = None,
case_brief: str = "",
retrieved_precedents: Optional[List[Dict]] = None,
) -> str:
"""
Create a new court session. Returns session_id.
"""
session_id = str(uuid.uuid4())
now = datetime.now(timezone.utc).isoformat()
max_rounds_map = {"brief": 3, "standard": 5, "extended": 8}
session = CourtSession(
session_id=session_id,
created_at=now,
updated_at=now,
case_title=case_title,
user_side=user_side,
user_client=user_client,
opposing_party=opposing_party,
legal_issues=legal_issues,
brief_facts=brief_facts,
jurisdiction=jurisdiction,
bench_composition=bench_composition,
difficulty=difficulty,
session_length=session_length,
show_trap_warnings=show_trap_warnings,
imported_from_session=imported_from_session,
case_brief=case_brief,
retrieved_precedents=retrieved_precedents or [],
phase="briefing",
current_round=0,
max_rounds=max_rounds_map.get(session_length, 5),
transcript=[],
concessions=[],
trap_events=[],
cited_precedents=[],
documents_produced=[],
user_arguments=[],
analysis=None,
outcome_prediction=None,
performance_score=None,
)
_sessions[session_id] = asdict(session)
logger.info(f"Session created: {session_id} | {case_title}")
return session_id
def get_session(session_id: str) -> Optional[Dict]:
"""Get session from memory. Returns None if not found."""
return _sessions.get(session_id)
def update_session(session_id: str, updates: Dict) -> bool:
"""Apply updates to session and persist to HF."""
if session_id not in _sessions:
logger.warning(f"Session not found: {session_id}")
return False
_sessions[session_id].update(updates)
_sessions[session_id]["updated_at"] = datetime.now(timezone.utc).isoformat()
# Async persist to HF Dataset
_persist_session(session_id)
return True
def add_transcript_entry(
session_id: str,
speaker: str,
role_label: str,
content: str,
entry_type: str = "argument",
metadata: Optional[Dict] = None,
) -> bool:
"""Add a new entry to the session transcript."""
session = get_session(session_id)
if not session:
return False
entry = asdict(TranscriptEntry(
speaker=speaker,
role_label=role_label,
content=content,
round_number=session["current_round"],
phase=session["phase"],
timestamp=datetime.now(timezone.utc).isoformat(),
entry_type=entry_type,
metadata=metadata or {},
))
session["transcript"].append(entry)
session["updated_at"] = datetime.now(timezone.utc).isoformat()
_persist_session(session_id)
return True
def add_concession(
session_id: str,
exact_quote: str,
legal_significance: str,
) -> bool:
"""Record a concession made by the user."""
session = get_session(session_id)
if not session:
return False
concession = asdict(Concession(
round_number=session["current_round"],
exact_quote=exact_quote,
legal_significance=legal_significance,
))
session["concessions"].append(concession)
session["updated_at"] = datetime.now(timezone.utc).isoformat()
logger.info(f"Concession recorded in session {session_id}: {exact_quote[:80]}")
return True
def add_trap_event(
session_id: str,
trap_type: str,
trap_text: str,
user_fell_in: bool = False,
user_response: str = "",
) -> bool:
"""Record a trap event."""
session = get_session(session_id)
if not session:
return False
trap = asdict(TrapEvent(
round_number=session["current_round"],
trap_type=trap_type,
trap_text=trap_text,
user_fell_in=user_fell_in,
user_response=user_response,
))
session["trap_events"].append(trap)
session["updated_at"] = datetime.now(timezone.utc).isoformat()
return True
def add_user_argument(
session_id: str,
argument_text: str,
key_claims: List[str],
) -> bool:
"""Track user's argument for inconsistency detection."""
session = get_session(session_id)
if not session:
return False
session["user_arguments"].append({
"round": session["current_round"],
"text": argument_text,
"key_claims": key_claims,
"timestamp": datetime.now(timezone.utc).isoformat(),
})
return True
def advance_phase(session_id: str) -> str:
"""
Move session to next phase.
Returns new phase name.
"""
session = get_session(session_id)
if not session:
return ""
phase_progression = {
"briefing": "rounds",
"rounds": "cross_examination",
"cross_examination": "closing",
"closing": "completed",
}
current = session["phase"]
next_phase = phase_progression.get(current, "completed")
update_session(session_id, {"phase": next_phase})
logger.info(f"Session {session_id} advanced: {current} → {next_phase}")
return next_phase
def advance_round(session_id: str) -> int:
"""Increment round counter. Returns new round number."""
session = get_session(session_id)
if not session:
return 0
new_round = session["current_round"] + 1
# Auto-advance phase when max rounds reached
if new_round > session["max_rounds"] and session["phase"] == "rounds":
advance_phase(session_id)
update_session(session_id, {"current_round": new_round})
return new_round
def get_all_sessions() -> List[Dict]:
"""Return all sessions, sorted by updated_at descending."""
sessions = list(_sessions.values())
return sorted(sessions, key=lambda x: x.get("updated_at", ""), reverse=True)
def get_session_transcript_text(session_id: str) -> str:
"""
Return full transcript as formatted text for LLM consumption.
Format matches real court transcript style.
"""
session = get_session(session_id)
if not session:
return ""
lines = [
f"IN THE {session['jurisdiction'].upper().replace('_', ' ')}",
f"Case: {session['case_title']}",
f"Petitioner: {session['user_client'] if session['user_side'] == 'petitioner' else session['opposing_party']}",
f"Respondent: {session['opposing_party'] if session['user_side'] == 'petitioner' else session['user_client']}",
"",
"PROCEEDINGS:",
"",
]
for entry in session["transcript"]:
lines.append(f"{entry['role_label'].upper()}")
lines.append(entry["content"])
lines.append("")
return "\n".join(lines)
def _persist_session(session_id: str):
"""
Persist session to HuggingFace Dataset.
Fails silently — in-memory session is still valid.
Non-critical: if HF upload fails, session continues working offline.
"""
if not HF_TOKEN:
return
try:
from huggingface_hub import HfApi
import threading
def _upload():
try:
api = HfApi(token=HF_TOKEN)
session_data = json.dumps(_sessions[session_id], ensure_ascii=False)
try:
api.create_repo(
repo_id=SESSIONS_REPO,
repo_type="dataset",
private=True,
exist_ok=True
)
except Exception as repo_err:
logger.debug(f"Could not create/access HF repo: {repo_err}")
api.upload_file(
path_or_fileobj=session_data.encode(),
path_in_repo=f"sessions/{session_id}.json",
repo_id=SESSIONS_REPO,
repo_type="dataset",
token=HF_TOKEN
)
except Exception as upload_err:
logger.debug(f"Session upload to HF failed (working offline): {upload_err}")
# Run in background thread — never blocks the response
thread = threading.Thread(target=_upload, daemon=True)
thread.start()
except Exception as e:
logger.debug(f"Session persist setup failed (non-critical): {e}")
def load_sessions_from_hf():
"""
Load all sessions from HF Dataset on startup.
Called once from api/main.py after download_models().
"""
if not HF_TOKEN:
logger.warning("No HF_TOKEN — sessions will not persist across restarts")
return
try:
from huggingface_hub import HfApi, list_repo_files
api = HfApi(token=HF_TOKEN)
try:
files = list(api.list_repo_files(
repo_id=SESSIONS_REPO,
repo_type="dataset",
token=HF_TOKEN
))
except Exception:
logger.info("No existing sessions on HF — starting fresh")
return
session_files = [f for f in files if f.startswith("sessions/") and f.endswith(".json")]
loaded = 0
for filepath in session_files:
try:
from huggingface_hub import hf_hub_download
local_path = hf_hub_download(
repo_id=SESSIONS_REPO,
filename=filepath,
repo_type="dataset",
token=HF_TOKEN
)
with open(local_path) as f:
session_data = json.load(f)
session_id = session_data.get("session_id")
if session_id:
_sessions[session_id] = session_data
loaded += 1
except Exception:
continue
logger.info(f"Loaded {loaded} sessions from HF Dataset")
except Exception as e:
logger.warning(f"Session load from HF failed (non-critical): {e}")
|