fix(arguments): route "חשב טיעונים" through the legal-analyst agent

The /aggregate-arguments endpoint ran an in-container BackgroundTask that
called claude_session (the local `claude` CLI) — which does not exist in the
FastAPI container. The button silently produced nothing, and on `force` it
destructively DELETEd existing arguments *before* the doomed LLM call.

Replace the inline task with the established delegation pattern used by
"חלץ עובדות שמאיות" (extract-appraiser-facts): a cheap in-container DB
pre-check (no_claims / exists), then a Paperclip wakeup of the company's
legal-analyst, which runs mcp__legal-ai__aggregate_claims_to_arguments
locally (where the CLI lives) and reports back. `force` now runs locally too,
so delete+recompute are atomic on the host — no more destructive failure.

Frontend: AggregateArgumentsResult becomes a discriminated union
(queued | no_claims | exists | skipped) and the toast is status-accurate
instead of the misleading fixed "refresh in a minute".

Invariants: G12 (Paperclip touch confined to paperclip_client behind the
agent_platform_port), G2 (replaces the broken path, no parallel capability),
engineering §6 (explicit statuses, no silent swallow).

UI change is logic/toast only (no visual-layout change) — within the
Claude-Design-gate bug-fix exemption. Richer inline status panels deferred
to the gate.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

web/agent_platform_port.py

@@ -56,6 +56,7 @@ from web.paperclip_client import (
     restore_project as pc_restore_project,
     update_project_name as pc_update_project_name,
     wake_analyst_for_appraiser_facts as pc_wake_analyst_for_appraiser_facts,
+    wake_analyst_for_argument_aggregation as pc_wake_analyst_for_argument_aggregation,
     wake_ceo_agent as pc_wake_ceo,
     wake_ceo_for_feedback_fold as pc_wake_ceo_for_feedback_fold,
     wake_curator_for_final as pc_wake_curator_for_final,
@@ -99,6 +100,7 @@ __all__ = [
     "pc_wake_curator_for_final",
     "pc_wake_for_precedent_extraction",
     "pc_wake_analyst_for_appraiser_facts",
+    "pc_wake_analyst_for_argument_aggregation",
     # comments / interactions
     "pc_post_comment",
     "pc_get_issue_comments",

web/app.py

@@ -75,6 +75,7 @@ from web.agent_platform_port import (
     pc_respond_to_interaction,
     pc_restore_project,
     pc_wake_analyst_for_appraiser_facts,
+    pc_wake_analyst_for_argument_aggregation,
     rename_case_project,
     pc_wake_ceo,
     pc_wake_ceo_for_feedback_fold,
@@ -2454,41 +2455,78 @@ async def api_get_claims(case_number: str):
 # the FastAPI container it short-circuits with status="llm_unavailable".
 
 @app.post("/api/cases/{case_number}/aggregate-arguments")
-async def api_aggregate_arguments(
-    case_number: str,
-    background_tasks: BackgroundTasks,
-    force: bool = False,
-):
-    """Aggregate raw claims into distinct legal arguments via Claude.
+async def api_aggregate_arguments(case_number: str, force: bool = False):
+    """Queue claim→argument aggregation by waking the legal-analyst agent.
 
-    Runs as a BackgroundTask because the LLM pass can take 30-90 seconds.
+    The aggregation itself calls `claude_session.query_json()`, which shells
+    out to the local `claude` CLI — present on the agent host, **absent in
+    this FastAPI container**. Running it inline as a BackgroundTask (the old
+    behaviour) silently produced nothing, and on `force` it destructively
+    deleted the existing arguments *before* the doomed LLM call. So we
+    delegate to the analyst exactly like `extract-appraiser-facts`: create a
+    child Paperclip issue, assign it to the company's analyst, and trigger a
+    wakeup. The analyst runs the MCP tool locally and posts results.
+
+    Cheap in-container pre-checks (DB only, no LLM) short-circuit before any
+    agent is spun up:
+      - `no_claims`  — there are no raw claims to aggregate yet.
+      - `exists`     — arguments already computed and `force` is False.
+
+    Response shape:
+        {"status": "queued", "sub_issue_id", "analyst_id", "main_issue_id"}
+      or {"status": "no_claims"|"exists", "total", "message"}
+      or {"status": "skipped", "reason": "no_api_key"|"no_analyst"|"no_issue"}
     """
     case = await db.get_case_by_number(case_number)
     if not case:
         raise HTTPException(404, f"תיק {case_number} לא נמצא")
 
-    async def _run() -> None:
-        try:
-            from legal_mcp.services import argument_aggregator
-            result = await argument_aggregator.aggregate_claims_to_arguments(
-                UUID(case["id"]), force=force,
-            )
-            logger.info(
-                "aggregate_arguments[%s] finished: %s",
-                case_number, result,
-            )
-        except Exception as e:  # noqa: BLE001
-            logger.exception(
-                "aggregate_arguments[%s] failed: %s", case_number, e,
-            )
+    case_id = UUID(case["id"])
+    pool = await db.get_pool()
+    async with pool.acquire() as conn:
+        claim_count = await conn.fetchval(
+            "SELECT COUNT(*) FROM claims WHERE case_id = $1", case_id,
+        )
+        existing_args = await conn.fetchval(
+            "SELECT COUNT(*) FROM legal_arguments WHERE case_id = $1", case_id,
+        )
 
-    background_tasks.add_task(_run)
-    return {
-        "status": "started",
-        "case_number": case_number,
-        "force": force,
-        "message": "Aggregation started in background. Poll /legal-arguments for results.",
-    }
+    if not claim_count:
+        return {
+            "status": "no_claims",
+            "total": 0,
+            "message": (
+                "אין טענות גולמיות בתיק. הרץ קודם חילוץ טענות (extract_claims) "
+                "ואז חשב טיעונים."
+            ),
+        }
+    if existing_args and not force:
+        return {
+            "status": "exists",
+            "total": existing_args,
+            "message": (
+                f"כבר קיימים {existing_args} טיעונים מאוגדים. השתמש בכפתור "
+                "החישוב-מחדש כדי לחשב מחדש (מוחק ובונה מחדש)."
+            ),
+        }
+
+    # Route to the analyst of the correct company by case-number prefix.
+    prefix = case_number[:1]
+    company_id = (
+        PAPERCLIP_COMPANIES["licensing"] if prefix == "1"
+        else PAPERCLIP_COMPANIES["betterment"] if prefix in ("8", "9")
+        else ""
+    )
+
+    try:
+        result = await pc_wake_analyst_for_argument_aggregation(
+            case_number, company_id=company_id, force=force,
+        )
+    except Exception as e:
+        logger.exception("analyst wakeup failed for argument aggregation %s", case_number)
+        raise HTTPException(500, f"לא ניתן לשלוח לאנליטיקאי: {e}")
+
+    return result
 
 
 @app.get("/api/cases/{case_number}/legal-arguments")

web/paperclip_client.py

@@ -1371,3 +1371,121 @@ async def wake_analyst_for_appraiser_facts(
         "analyst_id": analyst_id,
         "main_issue_id": main_issue_id,
     }
+
+
+async def wake_analyst_for_argument_aggregation(
+    case_number: str,
+    company_id: str,
+    force: bool = False,
+) -> dict:
+    """Wake the legal-analyst to aggregate raw claims into legal arguments.
+
+    Triggered by the chair clicking "חשב טיעונים" in the case page. The
+    FastAPI container cannot run `aggregate_claims_to_arguments` directly —
+    the aggregator calls `claude_session.query_json()`, which only works
+    where the local `claude` CLI is present (the MCP server / agent runner
+    on the host), **not in this container**. So instead of an in-container
+    BackgroundTask (which fails silently and, on `force`, destructively
+    deletes the existing arguments before the doomed LLM call), we create a
+    child issue under the case's main Paperclip issue, assign it to the
+    analyst of the correct company, and trigger a wakeup. The analyst's
+    HEARTBEAT picks up the issue, runs the MCP tool locally, and reports
+    back via a comment.
+
+    Mirrors ``wake_analyst_for_appraiser_facts`` — same delegation shape.
+
+    Returns a dict shaped for the FastAPI endpoint to serialize as-is:
+        {"status": "queued", "sub_issue_id", "analyst_id", "main_issue_id"}
+      or {"status": "skipped", "reason": "..."} for non-fatal early outs.
+    """
+    if not PAPERCLIP_BOARD_API_KEY:
+        logger.warning(
+            "PAPERCLIP_BOARD_API_KEY not set — cannot queue analyst wakeup "
+            "for argument aggregation on %s",
+            case_number,
+        )
+        return {"status": "skipped", "reason": "no_api_key"}
+
+    analyst_id = ANALYST_AGENTS.get(company_id)
+    if not analyst_id:
+        logger.info("No analyst configured for company %s — skipping", company_id)
+        return {"status": "skipped", "reason": "no_analyst", "company_id": company_id}
+
+    issues = await get_case_issues(case_number)
+    if not issues:
+        logger.warning(
+            "No Paperclip issues found for case %s — cannot queue analyst", case_number,
+        )
+        return {"status": "skipped", "reason": "no_issue"}
+
+    main_issue = next((i for i in issues if i.get("status") == "in_progress"), None) or issues[0]
+    main_issue_id = main_issue["id"]
+
+    force_clause = ", force=True" if force else ""
+    rerun_note = (
+        "זהו חישוב-מחדש (force) — הטיעונים הקיימים יימחקו ויחושבו מחדש.\n\n"
+        if force else ""
+    )
+    description = (
+        f"חיים ביקש חישוב טיעונים משפטיים בתיק {case_number}.\n\n"
+        f"{rerun_note}"
+        f"הרץ `mcp__legal-ai__aggregate_claims_to_arguments(case_number=\"{case_number}\"{force_clause})` "
+        f"וכתוב comment בעברית עם תוצאת האיגוד — כמה טיעונים מובחנים נוצרו לכל צד "
+        f"(עוררים / משיבים / ועדה / מבקשי-היתר). אם אין טענות גולמיות בתיק, דווח "
+        f"ב-comment שצריך להריץ קודם חילוץ טענות (`extract_claims`) וסגור את ה-issue כ-blocked."
+    )
+    child_resp = await pc_request(
+        "POST",
+        f"/api/issues/{main_issue_id}/children",
+        json={
+            "title": f"[ערר {case_number}] חישוב טיעונים משפטיים",
+            "description": description,
+            "status": "in_progress",
+            # Paperclip ISSUE_PRIORITIES = critical|high|medium|low — "normal"
+            # is NOT a valid enum value (Zod 400 → surfaces as 500).
+            "priority": "medium",
+            "assigneeAgentId": analyst_id,
+        },
+        raise_on_error=True,
+    )
+    sub_issue = child_resp.json()
+    sub_issue_id = sub_issue["id"]
+
+    # Tag plugin_state so the case page surfaces this sub-issue too.
+    try:
+        conn = await asyncpg.connect(PAPERCLIP_DB_URL)
+        try:
+            await _link_case_to_issue(conn, sub_issue_id, case_number)
+        finally:
+            await conn.close()
+    except Exception as e:
+        logger.warning("plugin_state link failed for sub_issue=%s: %s", sub_issue_id, e)
+
+    wake_resp = await pc_request(
+        "POST",
+        f"/api/agents/{analyst_id}/wakeup",
+        json={
+            "source": "on_demand",
+            "triggerDetail": "manual",
+            "reason": f"aggregate_arguments_{case_number}",
+            # "assignment" is the generic mutation the HEARTBEAT recognises;
+            # task intent lives in the child-issue description, not the payload.
+            "payload": {
+                "issueId": sub_issue_id,
+                "mutation": "assignment",
+                "caseNumber": case_number,
+            },
+        },
+        raise_on_error=True,
+    )
+    logger.info(
+        "Analyst wakeup for argument aggregation on case %s: sub_issue=%s "
+        "analyst=%s force=%s wake=%s",
+        case_number, sub_issue_id, analyst_id, force, wake_resp.status_code,
+    )
+    return {
+        "status": "queued",
+        "sub_issue_id": sub_issue_id,
+        "analyst_id": analyst_id,
+        "main_issue_id": main_issue_id,
+    }