Implemented safe rerun; fixed skill division bug

src/yc_bench/agent/prompt.py

@@ -44,7 +44,7 @@ Run multiple tasks concurrently when possible. Accept → assign → dispatch a
 ## Key Mechanics
 
 - **Salary bumps**: completed tasks raise salary for every assigned employee. More employees assigned = higher payroll growth.
-- **Throughput split**: employees on multiple active tasks split their rate (rate/sqrt(N)). Two tasks run at ~71% each.
+- **Throughput split**: employees on multiple active tasks split their rate (rate/N). Two tasks run at 50% each.
 - **Deadlines**: success before deadline = reward + prestige. Failure = prestige penalty, no reward.
 - **Trust**: completing tasks for a client builds trust → less work per task, access to gated tasks. Working for one client erodes trust with others.
 - **Not all clients are reliable.** Check `client history` for failure patterns.

src/yc_bench/agent/runtime/base.py

@@ -11,4 +11,12 @@ class AgentRuntime(ABC):
     def clear_session(self, session_id):
         raise NotImplementedError
 
+    def save_session_messages(self, session_id: str, path) -> None:
+        """Persist session messages for crash recovery. Override in subclass."""
+        pass
+
+    def restore_session_messages(self, session_id: str, path) -> int:
+        """Restore session messages from file. Returns count. Override in subclass."""
+        return 0
+
 __all__ = ["AgentRuntime"]

src/yc_bench/agent/runtime/litellm_runtime.py

@@ -146,6 +146,32 @@ class LiteLLMRuntime(AgentRuntime):
     def clear_session(self, session_id):
         self._sessions.pop(session_id, None)
 
+    def save_session_messages(self, session_id: str, path) -> None:
+        """Persist current session messages to a JSON file for resume."""
+        session = self._sessions.get(session_id)
+        if session is None:
+            return
+        import json as _json
+        from pathlib import Path
+        Path(path).write_text(_json.dumps(session.messages, separators=(",", ":")))
+
+    def restore_session_messages(self, session_id: str, path) -> int:
+        """Load saved messages into a session. Returns number of messages loaded."""
+        import json as _json
+        from pathlib import Path
+        p = Path(path)
+        if not p.exists():
+            return 0
+        try:
+            messages = _json.loads(p.read_text())
+            session = self._get_or_create_session(session_id)
+            session.messages = messages
+            logger.info("Restored %d messages into session %s.", len(messages), session_id)
+            return len(messages)
+        except Exception as exc:
+            logger.warning("Could not restore session from %s: %s", path, exc)
+            return 0
+
     # ------------------------------------------------------------------
     # Internal helpers
     # ------------------------------------------------------------------

src/yc_bench/core/progress.py

@@ -73,15 +73,13 @@ def _effective_rate_for_task_domain(*, task_id, domain, assignments,
                                     assignment_counts, base_rates):
     """Compute effective rate for one task+domain.
 
-    Throughput split uses sqrt(k) instead of k: two concurrent tasks each run at
-    1/sqrt(2) ≈ 71% speed, not 50%. This makes mild parallelism (2-3 tasks)
-    more efficient than strict sequential.
+    Throughput split is linear: an employee on k concurrent tasks contributes
+    rate/k to each. Total throughput is constant regardless of concurrency —
+    no benefit to parallelizing beyond pipeline flexibility.
 
     Brooks's Law: first 4 employees contribute full rate. Beyond that,
     additional employees contribute at 25% (overcrowding overhead).
     """
-    from math import sqrt
-
     # Collect (employee_id, effective_base) for this task, sorted best-first
     contributions = []
     for a in assignments:
@@ -91,7 +89,7 @@ def _effective_rate_for_task_domain(*, task_id, domain, assignments,
         if k <= 0:
             continue
         base = base_rates.get((a.employee_id, domain), Decimal("0"))
-        split_rate = base / Decimal(str(round(sqrt(k), 6)))
+        split_rate = base / Decimal(k)
         contributions.append(split_rate)
 
     # Sort best contributors first so they get full rate
@@ -264,14 +262,13 @@ def compute_effective_rates(db, company_id):
 
     out = []
     for req in requirements:
-        from math import sqrt
         contributions = []
         for a in assignments_by_task.get(req.task_id, []):
             k = assignment_counts.get(a.employee_id, 0)
             if k <= 0:
                 continue
             base = base_rates.get((a.employee_id, req.domain), Decimal("0"))
-            split_rate = base / Decimal(str(round(sqrt(k), 6)))
+            split_rate = base / Decimal(k)
             contributions.append(split_rate)
 
         contributions.sort(reverse=True)

src/yc_bench/runner/main.py

@@ -277,9 +277,24 @@ def run_benchmark(args):
         # Write live transcript alongside the DB so the streamlit dashboard can read it
         _slug = args.model.replace("/", "_")
         transcript_path = Path("db") / f"{args.config_name}_{args.seed}_{_slug}.transcript.jsonl"
-        if transcript_path.exists():
+        session_messages_path = Path("db") / f"{args.config_name}_{args.seed}_{_slug}.session.json"
+        _is_resume = transcript_path.exists() and session_messages_path.exists()
+        if not _is_resume and transcript_path.exists():
             transcript_path.unlink()
 
+        # Restore session messages on resume
+        if _is_resume:
+            n_restored = runtime.restore_session_messages(session_id, session_messages_path)
+            if n_restored > 0:
+                # Restore run_state turn count from transcript
+                try:
+                    prior_lines = transcript_path.read_text().strip().split("\n")
+                    prior_turns = len([l for l in prior_lines if l.strip()])
+                    run_state.turn_count = prior_turns
+                    logger.info("Resumed: %d prior turns, %d session messages.", prior_turns, n_restored)
+                except Exception:
+                    pass
+
         def _write_live_transcript(snapshot, rs, commands):
             """Append one JSONL line per turn for the streamlit dashboard."""
             if not rs.transcript:
@@ -296,6 +311,8 @@ def run_benchmark(args):
             }, separators=(",", ":"))
             with open(transcript_path, "a") as f:
                 f.write(line + "\n")
+            # Save session messages after every turn for crash recovery
+            runtime.save_session_messages(session_id, session_messages_path)
 
         if use_live:
             from .dashboard import BenchmarkDashboard