Package portal and close UI API gaps

ansible/roles/cezen-backend/files/main.py

@@ -2630,6 +2630,32 @@ async def guardrail_log(limit: int = 100, admin: dict = Depends(admin_only)):
     db.close()
     return [dict(r) for r in rows]
 
+@app.get("/api/guardrails/flags")
+async def guardrail_flags(limit: int = 20, admin: dict = Depends(admin_only)):
+    db = get_db()
+    rows = db.execute(
+        """SELECT gl.*, gr.label
+           FROM guardrail_log gl
+           LEFT JOIN guardrail_rules gr ON gr.id=gl.rule_id
+           ORDER BY gl.logged_at DESC LIMIT ?""",
+        (limit,)
+    ).fetchall()
+    db.close()
+    flags = []
+    for r in rows:
+        action = r["action"] or ""
+        severity = "high" if action == "block" else "medium" if action == "redact" else "low"
+        flags.append({
+            "id": r["id"],
+            "severity": severity,
+            "excerpt": r["pattern"],
+            "user": r["username"],
+            "ts": r["logged_at"],
+            "rule": r["label"] or r["pattern"] or r["rule_type"],
+            "action": action,
+        })
+    return {"flags": flags}
+
 # ── Model Benchmarking ────────────────────────────────────────────────────────
 
 class BenchmarkRun(BaseModel):
@@ -3294,6 +3320,80 @@ async def delete_doc_job(job_id: int, user: dict = Depends(current_user)):
 
 # ── Multimodal Chat ────────────────────────────────────────────────────────────
 
+@app.post("/api/chat")
+async def chat_completion(body: dict, user: dict = Depends(current_user)):
+    import urllib.request as _ur
+    model = body.get("model") or "llama3"
+    message = (body.get("message") or body.get("prompt") or "").strip()
+    system_prompt = (body.get("system_prompt") or "").strip()
+    if not message:
+        raise HTTPException(400, "message required")
+
+    messages = []
+    if system_prompt:
+        messages.append({"role": "system", "content": system_prompt})
+    messages.append({"role": "user", "content": message})
+
+    routed = _select_router_model(message, model, {"content_type": "text", "user_role": user.get("role", "")})
+    options = {}
+    if body.get("temperature") is not None:
+        options["temperature"] = float(body.get("temperature"))
+    if body.get("max_tokens") is not None:
+        options["num_predict"] = int(body.get("max_tokens"))
+    elif routed.get("max_tokens"):
+        options["num_predict"] = int(routed["max_tokens"])
+
+    payload = {"model": routed["model"], "messages": messages, "stream": False}
+    if options:
+        payload["options"] = options
+
+    started = datetime.now(timezone.utc)
+    try:
+        req = _ur.Request(
+            f"{OLLAMA_URL}/api/chat",
+            data=json.dumps(payload).encode(),
+            headers={"Content-Type": "application/json"},
+            method="POST"
+        )
+        with _ur.urlopen(req, timeout=300) as resp:
+            data = json.loads(resp.read())
+        response_text = data.get("message", {}).get("content", "")
+    except Exception as e:
+        db = get_db()
+        try:
+            db.execute(
+                "INSERT INTO query_log (user_id,username,model,prompt_tokens,completion_tokens,duration_ms,status,logged_at) VALUES (?,?,?,?,?,?,?,?)",
+                (user["sub"], user["username"], routed["model"], _estimate_tokens(message), 0, 0, "error", utcnow())
+            )
+            db.commit()
+        finally:
+            db.close()
+        raise HTTPException(502, f"Ollama error: {e}")
+
+    duration_ms = int((datetime.now(timezone.utc) - started).total_seconds() * 1000)
+    prompt_tokens = _estimate_tokens(system_prompt + " " + message)
+    completion_tokens = _estimate_tokens(response_text)
+    db = get_db()
+    try:
+        db.execute(
+            "INSERT INTO query_log (user_id,username,model,prompt_tokens,completion_tokens,duration_ms,status,logged_at) VALUES (?,?,?,?,?,?,?,?)",
+            (user["sub"], user["username"], routed["model"], prompt_tokens, completion_tokens, duration_ms, "ok", utcnow())
+        )
+        db.commit()
+    finally:
+        db.close()
+    return {
+        "response": response_text,
+        "model": routed["model"],
+        "route": routed.get("route", "default"),
+        "usage": {
+            "prompt_tokens": prompt_tokens,
+            "completion_tokens": completion_tokens,
+            "total_tokens": prompt_tokens + completion_tokens,
+        },
+        "duration_ms": duration_ms,
+    }
+
 @app.post("/api/multimodal/chat")
 async def multimodal_chat(body: dict, user: dict = Depends(current_user)):
     """
@@ -4191,6 +4291,73 @@ async def list_schedule_runs(job_id: int, user=Depends(current_user)):
     db.close()
     return [dict(r) for r in rows]
 
+@app.get("/api/jobs/failed")
+async def failed_jobs(limit: int = 10, admin: dict = Depends(admin_only)):
+    db = get_db()
+    items = []
+    for r in db.execute(
+        "SELECT id,orig_name,error_msg,created_at,finished_at FROM doc_jobs WHERE status='error' ORDER BY COALESCE(finished_at,created_at) DESC LIMIT ?",
+        (limit,)
+    ).fetchall():
+        items.append({"id": f"doc:{r['id']}", "type": "doc", "name": r["orig_name"], "error": r["error_msg"] or "Document job failed", "ts": r["finished_at"] or r["created_at"]})
+    for r in db.execute(
+        "SELECT id,agent_name,error_msg,created_at,finished_at FROM agent_runs WHERE status='error' ORDER BY COALESCE(finished_at,created_at) DESC LIMIT ?",
+        (limit,)
+    ).fetchall():
+        items.append({"id": f"agent:{r['id']}", "type": "agent", "name": r["agent_name"], "error": r["error_msg"] or "Agent run failed", "ts": r["finished_at"] or r["created_at"]})
+    for r in db.execute(
+        "SELECT id,job_id,job_name,error_msg,started_at,finished_at FROM scheduled_job_runs WHERE status='error' ORDER BY COALESCE(finished_at,started_at) DESC LIMIT ?",
+        (limit,)
+    ).fetchall():
+        items.append({"id": f"sched:{r['job_id']}", "type": "sched", "name": r["job_name"], "error": r["error_msg"] or "Scheduled job failed", "ts": r["finished_at"] or r["started_at"]})
+    for r in db.execute(
+        "SELECT id,name,status,created_at,finished_at FROM training_jobs WHERE status='error' ORDER BY COALESCE(finished_at,created_at) DESC LIMIT ?",
+        (limit,)
+    ).fetchall():
+        items.append({"id": f"training:{r['id']}", "type": "train", "name": r["name"], "error": "Training job failed", "ts": r["finished_at"] or r["created_at"]})
+    db.close()
+    items.sort(key=lambda x: x.get("ts") or "", reverse=True)
+    return {"jobs": items[:limit]}
+
+@app.post("/api/jobs/{job_key:path}/retry")
+async def retry_failed_job(job_key: str, admin: dict = Depends(admin_only)):
+    if ":" not in job_key:
+        raise HTTPException(400, "job id must include type prefix")
+    kind, raw_id = job_key.split(":", 1)
+    try:
+        jid = int(raw_id)
+    except ValueError:
+        raise HTTPException(400, "invalid job id")
+
+    db = get_db()
+    try:
+        if kind == "doc":
+            row = db.execute("SELECT id FROM doc_jobs WHERE id=?", (jid,)).fetchone()
+            if not row:
+                raise HTTPException(404, "Document job not found")
+            db.execute("UPDATE doc_jobs SET status='pending', error_msg=NULL, finished_at=NULL WHERE id=?", (jid,))
+            db.commit()
+            asyncio.create_task(_run_doc_job(jid))
+            return {"ok": True, "message": "Document job queued"}
+        if kind == "agent":
+            row = db.execute("SELECT id FROM agent_runs WHERE id=?", (jid,)).fetchone()
+            if not row:
+                raise HTTPException(404, "Agent run not found")
+            db.execute("UPDATE agent_runs SET status='pending', error_msg=NULL, finished_at=NULL WHERE id=?", (jid,))
+            db.commit()
+            asyncio.create_task(_execute_agent(jid))
+            return {"ok": True, "message": "Agent run queued"}
+        if kind == "sched":
+            row = db.execute("SELECT id FROM scheduled_jobs WHERE id=?", (jid,)).fetchone()
+            if not row:
+                raise HTTPException(404, "Scheduled job not found")
+            loop = asyncio.get_event_loop()
+            loop.run_in_executor(None, _run_scheduled_job_sync, jid)
+            return {"ok": True, "message": "Scheduled job triggered"}
+    finally:
+        db.close()
+    raise HTTPException(400, f"Retry not supported for job type: {kind}")
+
 
 # ── Notifications ─────────────────────────────────────────────────────────────
 

ansible/roles/cezen-nginx/tasks/main.yml

@@ -3,7 +3,9 @@
 
 - name: Install Nginx
   apt:
-    name: nginx
+    name:
+      - nginx
+      - rsync
     state: present
     update_cache: yes
 
@@ -17,7 +19,7 @@
 
 - name: Sync portal static files
   synchronize:
-    src: "{{ playbook_dir }}/../../../cezen-portal/"
+    src: "{{ playbook_dir }}/../cezen-portal/"
     dest: /opt/cezen/portal/
     delete: yes
     recursive: yes